Resilience-Ledger/Agentic-AI-Governance-Query.html at main · JakeTOpenSource/Resilience-Ledger · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
<!DOCTYPE html>
<html lang="en"><head><meta charset="utf-8"/>
<meta name="viewport" content="width=device-width, initial-scale=1"/>
<title>Delta Atlas - Explore</title>
<style>
 :root{--bg:#16301f;--panel:#2c4a38;--line:#4a6b56;--txt:#fafcf7;--dim:#d3decf;--accent:#e3c258;--win:#8fe0a5;}
 *{box-sizing:border-box;} body{margin:0;background:var(--bg);color:var(--txt);
   font-family:-apple-system,BlinkMacSystemFont,"Segoe UI",Roboto,Arial,sans-serif;font-size:14px;line-height:1.5;}
 header{padding:16px 22px;border-bottom:1px solid var(--line);position:sticky;top:0;background:var(--bg);z-index:5;}
 header h1{margin:0;font-size:18px;} header p{margin:4px 0 0;color:var(--dim);font-size:12.5px;max-width:920px;}
 .wrap{display:flex;gap:0;align-items:flex-start;}
 .side{width:300px;flex:none;border-right:1px solid var(--line);padding:16px 16px 60px;height:calc(100vh - 78px);overflow:auto;position:sticky;top:78px;}
 main{flex:1;padding:16px 22px 80px;min-width:0;}
 .facet{margin-bottom:15px;} .facet h4{margin:0 0 7px;font-size:11px;text-transform:uppercase;letter-spacing:.5px;color:var(--dim);}
 .chip{display:inline-block;cursor:pointer;border:1px solid var(--line);background:var(--panel);color:var(--txt);
   padding:4px 9px;border-radius:20px;font-size:11.5px;margin:2px 4px 2px 0;user-select:none;}
 .chip.on{border-color:var(--accent);color:var(--accent);background:rgba(217,181,69,0.08);}
 .chip .ct{color:var(--dim);font-size:10px;margin-left:4px;}
 .presets{display:flex;flex-wrap:wrap;gap:7px;margin:6px 0 14px;}
 .preset{cursor:pointer;border:1px solid var(--line);background:rgba(217,181,69,0.06);color:var(--txt);
   padding:6px 11px;border-radius:8px;font-size:12px;}
 .preset:hover{border-color:var(--accent);}
 .qline{font-family:ui-monospace,Menlo,Consolas,monospace;font-size:12.5px;color:var(--accent);
   background:var(--panel);border:1px solid var(--line);border-radius:8px;padding:9px 12px;margin-bottom:12px;word-break:break-word;}
 .count{font-size:13px;color:var(--dim);margin-bottom:10px;}
 .count b{color:var(--txt);font-size:15px;}
 .shape{display:flex;flex-wrap:wrap;gap:14px;background:var(--panel);border:1px solid var(--line);border-radius:10px;padding:11px 14px;margin-bottom:14px;}
 .shape .grp{font-size:11.5px;} .shape .gl{color:var(--dim);text-transform:uppercase;font-size:10px;letter-spacing:.4px;}
 .bar{height:6px;border-radius:3px;background:var(--accent);display:inline-block;vertical-align:middle;margin-right:6px;}
 .winners{background:linear-gradient(180deg,rgba(95,125,80,0.08),rgba(95,125,80,0));border:1px solid #4a6b56;
   border-radius:10px;padding:11px 14px;margin-bottom:16px;}
 .winners h3{margin:0 0 4px;font-size:12.5px;color:var(--win);} .winners .sub{color:var(--dim);font-size:11px;margin-bottom:8px;}
 .wrow{display:flex;align-items:center;gap:9px;margin:4px 0;font-size:12.5px;}
 .wscore{font-family:ui-monospace,monospace;color:var(--win);font-size:11.5px;min-width:30px;}
 .card{border:1px solid var(--line);border-radius:10px;padding:11px 14px;margin:8px 0;background:var(--panel);}
 .card h3{margin:0 0 3px;font-size:14.5px;}
 .card .fn{color:var(--dim);font-style:italic;margin:2px 0 7px;font-size:12.5px;}
 .tag{display:inline-block;font-size:10.5px;padding:1px 7px;border-radius:20px;border:1px solid var(--line);margin:1px 4px 1px 0;color:var(--dim);}
 .tag.p{color:var(--accent);border-color:#bcd0e6;} .tag.w{color:var(--win);border-color:#4a6b56;}
 .rel{margin-top:7px;font-size:11.5px;color:var(--dim);}
 .rel b{color:var(--txt);font-weight:600;}
 .adv{border-top:1px solid var(--line);margin-top:10px;padding-top:12px;} .adv h4{margin:0 0 2px;font-size:12.5px;color:var(--txt);} .advhelp{font-size:11.5px;color:var(--dim);margin-bottom:4px;} .advlab{display:block;font-size:10.5px;text-transform:uppercase;letter-spacing:.4px;color:var(--dim);margin-top:8px;}
 select{width:100%;background:var(--panel);color:var(--txt);border:1px solid var(--line);border-radius:7px;padding:6px 8px;font-size:12px;margin-top:5px;}
 .clear{cursor:pointer;color:#1e2b17;background:var(--accent);font-weight:600;font-size:12px;margin-top:10px;padding:5px 12px;border-radius:16px;display:inline-block;}
 .none{color:var(--dim);font-style:italic;padding:20px 0;}
 .jump{cursor:pointer;color:var(--accent);border-bottom:1px dotted transparent;}
 .jump:hover{border-bottom-color:var(--accent);}
 .seclbl2{font-size:11px;text-transform:uppercase;letter-spacing:.5px;color:var(--dim);margin:16px 0 8px;border-bottom:1px solid var(--line);padding-bottom:5px;}
 .method-d{color:#f2cd55;} .method-c{color:var(--win);}
 .sdot{display:inline-block;width:9px;height:9px;border-radius:50%;margin-right:7px;vertical-align:middle;} .slegend{font-size:11px;color:var(--dim);display:inline-flex;align-items:center;gap:2px;flex-wrap:wrap;} .slegend .sdot{margin:0 3px 0 8px;}

 /* mobile: stack the two-column tool layout so nothing is trapped in a fixed-height pane */
 @media (max-width:720px){
  .wrap{flex-direction:column;min-height:0;}
  .in,.side{width:100%!important;min-width:0!important;height:auto!important;max-height:none!important;position:static!important;overflow:visible!important;border-right:none;border-bottom:1px solid var(--line);}
  main{overflow:visible!important;height:auto!important;min-height:0!important;}
 }

 /* canonical typography: one baseline across every page (added 2026-07-11) */
 body{font-family:-apple-system,BlinkMacSystemFont,"Segoe UI",Roboto,Arial,sans-serif;font-size:14.5px;font-weight:450;line-height:1.6;}
 textarea,input,select{font-family:inherit;font-size:14px;line-height:1.5;}
 button{font-family:inherit;}
 ::placeholder{color:var(--dim);opacity:1;}
</style></head><body>
<header>
 <h1>Delta Atlas - Explore</h1>
 <p><b>See how the words connect by meaning.</b> Browse every term, or tick boxes on the left (like <i>a risk</i>, <i>hard to undo</i>, <i>no human in charge</i>) to narrow it. Each term shows what it does and which other terms it <b>couples with</b>: what defends what, what needs what, what is a kind of what. Try a <b>Try:</b> example below.</p>
 <div style="margin-top:7px"><span class="slegend"><span class="sdot" style="background:#8fe0a5"></span>Verified (solid)<span class="sdot" style="background:transparent;border:3px solid #f2cd55"></span>Draft (ring)<span class="sdot" style="background:transparent;border:2px dashed #ff9d80"></span>Thin (dashed)</span></div>
</header>
<div class="wrap">
 <div class="side" id="side"></div>
 <main>
   <div class="presets" id="presets"></div>
   <div class="count" id="count"></div>
   <div id="results"></div>
 </main>
</div>
<script>
function sdot(s){var m={reviewed:['background:#8fe0a5','Verified (human-reviewed)'],candidate:['background:transparent;border:3px solid #f2cd55','Draft (not yet reviewed)'],stub:['background:transparent;border:2px dashed #ff9d80','Thin (placeholder)']};var x=m[s]||m.candidate;return '<span class="sdot" style="'+x[0]+'" title="'+x[1]+'"></span>';}

const DATA={"terms":[{"id":"artificial-intelligence","function_statement":"Software that performs tasks we associate with human intelligence: learning from data, spotting patterns, using language, or making decisions.","names":{"plain":["Artificial Intelligence"],"technical":[{"name":"Artificial Intelligence","source":"STANFORD-HAI"},{"name":"AI","source":"STANFORD-HAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","ISO-22989"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"traditional-ai","function_statement":"Useful contrast: it explains why modern AI behaves less predictably, having learned patterns instead of being told rules.","names":{"plain":["Traditional AI","symbolic AI","rule-based AI"],"technical":[{"name":"Traditional AI","source":"STANFORD-HAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"machine-learning","function_statement":"A method where a system learns patterns from data instead of following explicit rules, then applies them to new inputs.","names":{"plain":["Machine Learning"],"technical":[{"name":"Machine Learning","source":"STANFORD-HAI"},{"name":"ML","source":"STANFORD-HAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","ISO-22989"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"deep-learning","function_statement":"The breakthrough that made image recognition, speech, and language models work well.","names":{"plain":["Deep Learning"],"technical":[{"name":"Deep Learning","source":"STANFORD-HAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","ISO-22989"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"neural-network","function_statement":"The basic machinery deep learning runs on.","names":{"plain":["Neural Network"],"technical":[{"name":"Neural Network","source":"STANFORD-HAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","ISO-22989"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"parameter-weights","function_statement":"The numbers a model adjusts during training to store what it has learned. Model size is often quoted as a parameter count.","names":{"plain":["Parameter / Weights","Parameter","Weights"],"technical":[{"name":"Parameter / Weights","source":"STANFORD-HAI"},{"name":"Parameter","source":"OTHER"},{"name":"Weights","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","HF","OTHER","FOUND","ARXIV"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived","prov":["Statistical Modeling / Deep Learning","Rosenblatt Perceptron / Matrix Parameters"]},{"id":"model","function_statement":"A trained system that maps inputs to outputs. The artifact you run, deploy, and govern once training is finished.","names":{"plain":["Model"],"technical":[{"name":"Model","source":"STANFORD-HAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","ISO","NIST","ISO-22989"],"status":"reviewed","prov":["ISO/IEC 22989 / NIST AI RMF"],"attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"foundation-model","function_statement":"A large model trained broadly on general data, then adapted to many tasks rather than trained from scratch for each one.","names":{"plain":["Foundation Model"],"technical":[{"name":"Foundation Model","source":"STANFORD-HAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","ISO-22989"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"large-language-model","function_statement":"The core of chatbots and most agentic systems; NSA/CISA guidance treats agentic AI as primarily LLM-based.","names":{"plain":["Large Language Model"],"technical":[{"name":"Large Language Model","source":"NSA-CISA-AGENTIC"},{"name":"LLM","source":"NSA-CISA-AGENTIC"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["NSA-CISA-AGENTIC","NIST","EU","STANFORD-HAI"],"status":"reviewed","prov":["NIST AI RMF / EU AI Act"],"attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"generative-ai","function_statement":"Names the capability that distinguishes today's tools from older predict-and-classify AI.","names":{"plain":["Generative AI"],"technical":[{"name":"Generative AI","source":"STANFORD-HAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","ISO-22989"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"transformer","function_statement":"The architecture that made LLMs possible. You rarely need the internals, but the word comes up constantly.","names":{"plain":["Transformer"],"technical":[{"name":"Transformer","source":"STANFORD-HAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","ARXIV","HF"],"status":"reviewed","prov":["Vaswani et al. (Attention Is All You Need)"],"attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"context-window","function_statement":"The fixed amount of text, measured in tokens, a model can consider at once. Input beyond it is dropped or must be retrieved.","names":{"plain":["Context Window"],"technical":[{"name":"Context Window","source":"STANFORD-HAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","HF"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"embeddings","function_statement":"Data represented as numeric vectors whose distances reflect similarity in meaning, which enables search and comparison by meaning.","names":{"plain":["Embeddings","Embedding"],"technical":[{"name":"Embeddings","source":"STANFORD-HAI"},{"name":"Embedding","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","OWASP-LLM","HF","OTHER","OPENAI"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived","prov":["Word2Vec / OpenAI"]},{"id":"inference","function_statement":"Running a trained model to produce an output. The operating phase of AI, as opposed to training.","names":{"plain":["Inference"],"technical":[{"name":"Inference","source":"STANFORD-HAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","FOUND","NVIDIA","ISO-22989"],"status":"reviewed","prov":["Statistics / NVIDIA TensorRT"],"attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"training-data","function_statement":"Quality, bias, and provenance of this data shape everything the model later does. A core governance concern.","names":{"plain":["Training Data"],"technical":[{"name":"Training Data","source":"STANFORD-HAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","ISO-22989","NSA-CISA-DATA"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"fine-tuning","function_statement":"Further training a pretrained model on focused data so it adapts to a specific task or domain.","names":{"plain":["Fine-tuning"],"technical":[{"name":"Fine-tuning","source":"STANFORD-HAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","FOUND","HF"],"status":"reviewed","prov":["Deep Learning"],"attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"prompt","function_statement":"The text input given to a model. It carries both the request and its context, so its wording strongly shapes the output.","names":{"plain":["Prompt"],"technical":[{"name":"Prompt","source":"STANFORD-HAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"human","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","NIST","OTHER"],"status":"reviewed","prov":["NIST AI RMF / Generative AI"],"attributes_method":"curated","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"prompt-engineering","function_statement":"The practice of designing prompts, examples, and context to get reliable outputs from a model without retraining it.","names":{"plain":["Prompt Engineering"],"technical":[{"name":"Prompt Engineering","source":"STANFORD-HAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"human","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","HF"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"retrieval-augmented-generation","function_statement":"A method where the model first retrieves relevant documents, then writes its answer from them, so it can cite current or private facts instead of relying on memory. Cuts down made-up answers.","names":{"plain":["Retrieval-Augmented Generation"],"technical":[{"name":"Retrieval-Augmented Generation","source":"STANFORD-HAI"},{"name":"RAG","source":"STANFORD-HAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","META"],"status":"candidate","prov":["Meta AI (Retrieval-Augmented)"],"attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"reasoning-model","function_statement":"Better at math, logic, and multi-step tasks; underpins more capable agents.","names":{"plain":["Reasoning Model"],"technical":[{"name":"Reasoning Model","source":"STANFORD-HAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","ARXIV"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"multimodal-ai","function_statement":"Expands what agents can perceive and act on (reading a screenshot, describing a photo).","names":{"plain":["Multimodal AI"],"technical":[{"name":"Multimodal AI","source":"STANFORD-HAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","HF"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"agent","function_statement":"An AI system that pursues a goal over multiple steps, choosing actions and calling tools, rather than answering a single prompt.","names":{"plain":["Agent"],"technical":[{"name":"Agent","source":"STANFORD-HAI"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","OTHER","NIST","OWASP-AGENTIC"],"status":"reviewed","prov":["Russell & Norvig / NIST AI RMF"],"attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"agentic-ai","function_statement":"AI that pursues goals over multiple steps with limited human oversight, often by coordinating one or more agents and subagents. Agency is a spectrum, from a single scoped agent to many acting together, not an on/off switch.","names":{"plain":["Agentic AI"],"technical":[{"name":"Agentic AI","source":"STANFORD-HAI"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","NIST-AI-RMF","OWASP-AGENTIC"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"autonomy","function_statement":"The dial that most determines risk. More autonomy = more speed and scale, but also more ways to go wrong before anyone notices.","names":{"plain":["Autonomy"],"technical":[{"name":"Autonomy","source":"STANFORD-HAI"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","EU","ISO-22989","OWASP-AGENTIC"],"status":"reviewed","prov":["EU AI Act"],"attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"tool-use-function-calling","function_statement":"What lets an agent affect the real world. Also the main place it can cause real damage, so tools are a key control point.","names":{"plain":["Tool Use / Function Calling"],"technical":[{"name":"Tool Use / Function Calling","source":"STANFORD-HAI"},{"name":"Function Calling","source":"STANFORD-HAI"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","OWASP-AGENTIC"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"orchestration","function_statement":"The layer that coordinates models, tools, and steps into a workflow, and where approvals and limits are enforced.","names":{"plain":["Orchestration"],"technical":[{"name":"Orchestration","source":"STANFORD-HAI"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","OASIS","CNCF","OWASP-AGENTIC"],"status":"reviewed","prov":["OASIS TOSCA / CNCF"],"attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"planning","function_statement":"The capability that turns a single answer into a multi-step task. Powerful, and harder to predict or audit.","names":{"plain":["Planning"],"technical":[{"name":"Planning","source":"STANFORD-HAI"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","OTHER"],"status":"reviewed","prov":["Russell & Norvig / STRIPS Systems"],"attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"goal-objective","function_statement":"The thing everything else is optimized toward. Most agentic failures trace back to a goal that was vague, wrong, or quietly outranked the rules.","names":{"plain":["Goal / Objective","Goal","Objective"],"technical":[{"name":"Goal / Objective","source":"STANFORD-HAI"},{"name":"Goal","source":"OTHER"},{"name":"Objective","source":"OTHER"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","OWASP-AGENTIC","OTHER"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived","prov":["Russell & Norvig"]},{"id":"workflow","function_statement":"A fixed series of steps the AI follows to finish a task. Because the path is set ahead of time, it is predictable and easy to check, which is why it is the safest place to start.","names":{"plain":["Workflow"],"technical":[{"name":"Workflow","source":"STANFORD-HAI"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","OTHER"],"status":"reviewed","prov":["Workflow Management Coalition"],"attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"multi-agent-system","function_statement":"More capable, but introduces structural risk: one agent's error can cascade through the others (NSA/CISA).","names":{"plain":["Multi-Agent System","Multi-Agent"],"technical":[{"name":"Multi-Agent System","source":"NSA-CISA-AGENTIC"},{"name":"MAS","source":"NSA-CISA-AGENTIC"},{"name":"Multi-Agent","source":"OTHER"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["NSA-CISA-AGENTIC","OWASP-AGENTIC","MIT","FIPA","ARXIV"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived","prov":["FIPA Standards"]},{"id":"model-context-protocol","function_statement":"An open standard for connecting agents to external tools and data through one uniform interface.","names":{"plain":["Model Context Protocol","tool/connector protocol"],"technical":[{"name":"Model Context Protocol","source":"STANFORD-HAI"},{"name":"MCP","source":"STANFORD-HAI"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","OWASP-AGENTIC"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"memory","function_statement":"Lets agents stay coherent over long tasks, and creates a new place where bad or poisoned data can persist.","names":{"plain":["Memory"],"technical":[{"name":"Memory","source":"STANFORD-HAI"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","OTHER"],"status":"reviewed","prov":["Computer Architecture / Agent Frameworks"],"attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"human-in-the-loop","function_statement":"A setup where a person must review or approve an AI's action before it takes effect, so higher-risk decisions are not left fully to the machine.","names":{"plain":["Human-in-the-Loop"],"technical":[{"name":"Human-in-the-Loop","source":"STANFORD-HAI"},{"name":"HITL","source":"STANFORD-HAI"}]},"cluster":"Agentic Mechanics","purpose":"center","attributes":{"actor":"human","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","NIST","ISO","ISO-22989"],"status":"reviewed","prov":["NIST AI RMF / ISO/IEC 22989","NIST AI RMF / ISO/IEC 42001"],"attributes_method":"preset","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"human-on-the-loop","function_statement":"A lighter-touch oversight model for higher autonomy: faster, but relies on the human actually noticing in time.","names":{"plain":["Human-on-the-Loop"],"technical":[{"name":"Human-on-the-Loop","source":"STANFORD-HAI"},{"name":"HOTL","source":"STANFORD-HAI"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"human","accountability":"shared","cost_substrate":"non-resetting","reversibility":"undefined","observability":"undefined"},"sources":["STANFORD-HAI","NATO","ISO-22989"],"status":"reviewed","prov":["NATO Autonomous Frameworks"],"attributes_method":"preset","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"excessive-agency","function_statement":"Names a top agentic risk category and points straight at the fix: scope it down, limit its tools, add approval.","names":{"plain":["Excessive Agency"],"technical":[{"name":"Excessive Agency","source":"OWASP-LLM"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"non-resetting","reversibility":"irreversible","observability":"latent"},"sources":["OWASP-LLM","NIST-AI-RMF","OWASP-AGENTIC","MIT"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"guardrail","function_statement":"An enforced limit on what a model or agent may do, checked against a rule before an action is allowed to proceed.","names":{"plain":["Guardrail"],"technical":[{"name":"Guardrail","source":"NIST-AI-RMF"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST-AI-RMF","NVIDIA","OWASP"],"status":"reviewed","prov":["NVIDIA NeMo / OWASP"],"attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"ontology-policy-gate","function_statement":"A control that first checks whether an action is even a permitted type, before checking whether it is carried out correctly.","names":{"plain":["Ontology / Policy Gate"],"technical":[{"name":"Ontology / Policy Gate","source":"NIST-AI-RMF"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST-AI-RMF"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"validation-gate","function_statement":"Catches bad results before they take effect. A deterministic Check the agent cannot talk its way past.","names":{"plain":["Validation Gate","enforcement gate"],"technical":[{"name":"Validation Gate","source":"NIST-AI-RMF"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST-AI-RMF"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"deterministic-constraint","function_statement":"A fixed rule the model cannot override, so it holds even when the model misbehaves. It is the enforcing core of a harness or tether: the part that actually says no.","names":{"plain":["Deterministic Constraint"],"technical":[{"name":"Deterministic Constraint","source":"NIST-AI-RMF"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST-AI-RMF"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"alignment","function_statement":"The overarching aim of safety work. Misalignment is the root of many failure modes downstream.","names":{"plain":["Alignment"],"technical":[{"name":"Alignment","source":"STANFORD-HAI"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","OPENAI","NIST"],"status":"reviewed","prov":["OpenAI / NIST AI RMF"],"attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"ai-safety","function_statement":"The discipline that contains most other terms in this cluster.","names":{"plain":["AI Safety"],"technical":[{"name":"AI Safety","source":"STANFORD-HAI"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"sandboxing","function_statement":"Lets an agent fail safely. In Ledger terms an Absorb function: it contains survivable error instead of letting it spread.","names":{"plain":["Sandboxing"],"technical":[{"name":"Sandboxing","source":"NIST-AI-RMF"}]},"cluster":"Control & Safety","purpose":"absorb","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST-AI-RMF","OWASP-AGENTIC"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"least-privilege","function_statement":"Giving an agent only the minimum access and permissions its task needs, so a compromise or mistake can do limited damage.","names":{"plain":["Least Privilege"],"technical":[{"name":"Least Privilege","source":"NSA-CISA-AGENTIC"}]},"cluster":"Control & Safety","purpose":"capability","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NSA-CISA-AGENTIC","OWASP-AGENTIC"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"zero-trust","function_statement":"A security model that verifies every action rather than trusting it by default, applied here to each step an agent takes.","names":{"plain":["Zero Trust"],"technical":[{"name":"Zero Trust","source":"NSA-CISA-AGENTIC"}]},"cluster":"Control & Safety","purpose":"capability","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NSA-CISA-AGENTIC","OWASP-AGENTIC"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"defense-in-depth","function_statement":"If one control fails, others still hold. An Absorb function at the system level.","names":{"plain":["Defense-in-Depth"],"technical":[{"name":"Defense-in-Depth","source":"NIST-AI-RMF"}]},"cluster":"Control & Safety","purpose":"absorb","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST-AI-RMF","OWASP-AGENTIC"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"red-teaming","function_statement":"Turns unknown risks into known, fixable ones. A Check run adversarially.","names":{"plain":["Red Teaming"],"technical":[{"name":"Red Teaming","source":"NIST-AI-RMF"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"human","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST-AI-RMF"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"evaluation-benchmark","function_statement":"A standardized test of a model or agent's performance, used to judge fitness for a task and to detect decline over time.","names":{"plain":["Evaluation / Benchmark"],"technical":[{"name":"Evaluation / Benchmark","source":"STANFORD-HAI"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","NIST-AI-RMF"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"kill-switch-circuit-breaker","function_statement":"The last-resort Reset: halts runaway behavior. Circuit breakers can trip automatically when limits are crossed.","names":{"plain":["Kill Switch / Circuit Breaker"],"technical":[{"name":"Kill Switch / Circuit Breaker","source":"NIST-AI-RMF"}]},"cluster":"Control & Safety","purpose":"absorb","attributes":{"actor":"human","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST-AI-RMF","OWASP-AGENTIC"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"rate-limiting","function_statement":"Slows damage so humans can catch it, an Absorb function that buys reaction time.","names":{"plain":["Rate Limiting"],"technical":[{"name":"Rate Limiting","source":"NIST-AI-RMF"}]},"cluster":"Control & Safety","purpose":"absorb","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST-AI-RMF","OWASP-AGENTIC"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"scoping","function_statement":"The cheapest, most effective control. NSA/CISA advise starting with low-risk, tightly scoped, non-sensitive use cases.","names":{"plain":["Scoping"],"technical":[{"name":"Scoping","source":"NSA-CISA-AGENTIC"}]},"cluster":"Control & Safety","purpose":"capability","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NSA-CISA-AGENTIC","OWASP-AGENTIC"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"graceful-degradation","function_statement":"The Ledger's Absorb function by its field-standard name: survivable error doesn't become a terminal stop.","names":{"plain":["Graceful Degradation","fail-soft","fault tolerance"],"technical":[{"name":"Graceful Degradation","source":"NIST-AI-RMF"}]},"cluster":"Control & Safety","purpose":"absorb","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST-AI-RMF"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"ai-governance","function_statement":"The framework everything in this cluster lives inside. NSA/CISA stress agentic AI usually fits existing governance, not a brand-new discipline.","names":{"plain":["AI Governance"],"technical":[{"name":"AI Governance","source":"NSA-CISA-AGENTIC"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NSA-CISA-AGENTIC","NIST-AI-RMF","ISO-42001"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"nist-ai-risk-management-framework","function_statement":"Gives organizations a shared structure and vocabulary for AI risk. Govern sets policy; Map finds risks in context; Measure rates trustworthiness; Manage decides what to do about them.","names":{"plain":["NIST AI Risk Management Framework"],"technical":[{"name":"NIST AI Risk Management Framework","source":"NIST-AI-RMF"},{"name":"AI RMF","source":"NIST-AI-RMF"},{"name":"NIST AI RMF","source":"NIST-AI-RMF"},{"name":"RMF","source":"NIST-AI-RMF"},{"name":"NIST","source":"NIST-AI-RMF"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST-AI-RMF"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"responsible-ai","function_statement":"The values layer that governance operationalizes.","names":{"plain":["Responsible AI"],"technical":[{"name":"Responsible AI","source":"STANFORD-HAI"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","NIST-AI-RMF"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"ethical-ai","function_statement":"Close cousin of Responsible AI; emphasizes the values themselves over the process.","names":{"plain":["Ethical AI"],"technical":[{"name":"Ethical AI","source":"STANFORD-HAI"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"accountability","function_statement":"One of NSA/CISA's five agentic risk categories. Hard with autonomous agents precisely because the actor isn't a person, so accountability has to be assigned deliberately.","names":{"plain":["Accountability"],"technical":[{"name":"Accountability","source":"NSA-CISA-AGENTIC"}]},"cluster":"Oversight & Governance","purpose":"center","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"undefined","observability":"undefined"},"sources":["NSA-CISA-AGENTIC","NIST","ISO","NIST-AI-RMF","ISO-42001"],"status":"reviewed","prov":["NIST AI RMF / ISO/IEC 42001"],"attributes_method":"preset","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"audit-trail","function_statement":"The Ledger's Reset/Return function: it makes a valid action distinguishable from damage, and makes accountability possible after the fact.","names":{"plain":["Audit Trail","audit log"],"technical":[{"name":"Audit Trail","source":"NIST-AI-RMF"}]},"cluster":"Oversight & Governance","purpose":"reset","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST-AI-RMF","ISO-42001"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Termination State","sysdyn_method":"derived"},{"id":"model-card","function_statement":"A short document describing a model's purpose, training data, limitations, and intended use.","names":{"plain":["Model Card"],"technical":[{"name":"Model Card","source":"NIST-AI-RMF"}]},"cluster":"Oversight & Governance","purpose":"reset","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST-AI-RMF","ISO-42001"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Termination State","sysdyn_method":"derived"},{"id":"system-card","function_statement":"Documentation of a deployed system as a whole: the model plus the surrounding components and safeguards.","names":{"plain":["System Card"],"technical":[{"name":"System Card","source":"NIST-AI-RMF"}]},"cluster":"Oversight & Governance","purpose":"reset","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST-AI-RMF","ISO-42001"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Termination State","sysdyn_method":"derived"},{"id":"transparency","function_statement":"Making how a system works and how it reaches its decisions open to inspection, so outsiders can see what it does rather than take it on faith.","names":{"plain":["Transparency"],"technical":[{"name":"Transparency","source":"NIST-AI-RMF"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST-AI-RMF","EU","NIST","ISO-22989","ISO-42001"],"status":"reviewed","prov":["EU AI Act / NIST AI RMF"],"attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"black-box","function_statement":"A system whose outputs can be observed but whose internal reasoning cannot be seen or readily understood.","names":{"plain":["Black Box"],"technical":[{"name":"Black Box","source":"NIST-AI-RMF"}]},"cluster":"Oversight & Governance","purpose":"capability","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST-AI-RMF","ISO-42001"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"compliance","function_statement":"Conformance with the laws, regulations, and policies that apply to a system. These are the rules that cannot be opted out of without eventual consequence.","names":{"plain":["Compliance"],"technical":[{"name":"Compliance","source":"NIST-AI-RMF"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST-AI-RMF","ISO"],"status":"reviewed","prov":["ISO/IEC 42001"],"attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"risk-assessment","function_statement":"The process of identifying, rating, and prioritizing the ways a system could cause harm.","names":{"plain":["Risk Assessment"],"technical":[{"name":"Risk Assessment","source":"NIST-AI-RMF"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST-AI-RMF","ISO-42001"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"ai-lifecycle","function_statement":"Governance attaches at every stage; risks and controls differ by stage. A useful mental map for newcomers.","names":{"plain":["AI Lifecycle"],"technical":[{"name":"AI Lifecycle","source":"NIST-AI-RMF"}]},"cluster":"Oversight & Governance","purpose":"capability","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST-AI-RMF","ISO-22989"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"shadow-ai","function_statement":"AI tools used inside an organization without approval or oversight, operating outside its governance.","names":{"plain":["Shadow AI"],"technical":[{"name":"Shadow AI","source":"NIST-AI-RMF"}]},"cluster":"Oversight & Governance","purpose":"capability","attributes":{"actor":"org","accountability":"org","cost_substrate":"non-resetting","reversibility":"reversible","observability":"silent"},"sources":["NIST-AI-RMF","ISO-42001"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"human-centered-ai","function_statement":"The orienting philosophy behind keeping a human meaningfully in the loop.","names":{"plain":["Human-Centered AI"],"technical":[{"name":"Human-Centered AI","source":"STANFORD-HAI"},{"name":"HAI","source":"STANFORD-HAI"}]},"cluster":"Oversight & Governance","purpose":"capability","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"ground-truth","function_statement":"The fixed reference everything is measured from. The Ledger's whole point: a system can be perfectly self-consistent and still wrong, so it needs an external ground truth, ultimately verified by a human in contact with reality.","names":{"plain":["Ground Truth"],"technical":[{"name":"Ground Truth","source":"NSA-CISA-DATA"}]},"cluster":"Data & Provenance","purpose":"center","attributes":{"actor":"human","accountability":"human","cost_substrate":"non-resetting","reversibility":"undefined","observability":"undefined"},"sources":["NSA-CISA-DATA","STANFORD-HAI"],"status":"reviewed","attributes_method":"preset","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"grounding","function_statement":"Tying a model's output to verified external sources so its claims can be checked, which reduces fabrication.","names":{"plain":["Grounding"],"technical":[{"name":"Grounding","source":"NSA-CISA-DATA"}]},"cluster":"Data & Provenance","purpose":"check","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["NSA-CISA-DATA","NIST","OTHER","STANFORD-HAI"],"status":"candidate","prov":["NIST AI RMF / Cognitive Science"],"attributes_method":"derived","tier":"technical","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"provenance","function_statement":"The recorded origin and history of data or a model, used to judge whether it can be trusted.","names":{"plain":["Provenance"],"technical":[{"name":"Provenance","source":"NSA-CISA-AGENTIC"}]},"cluster":"Data & Provenance","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NSA-CISA-AGENTIC","W3C","NSA-CISA-DATA"],"status":"reviewed","prov":["W3C PROV-O"],"attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"attestation","function_statement":"Verifiable, often cryptographic, evidence of a system's identity or state, used to establish trust between parties.","names":{"plain":["Attestation"],"technical":[{"name":"Attestation","source":"NSA-CISA-DATA"}]},"cluster":"Data & Provenance","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NSA-CISA-DATA"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"data-lineage","function_statement":"The detailed, technical form of provenance. Essential for auditing and debugging bad outputs.","names":{"plain":["Data Lineage"],"technical":[{"name":"Data Lineage","source":"NSA-CISA-DATA"}]},"cluster":"Data & Provenance","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NSA-CISA-DATA"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"synthetic-data","function_statement":"Useful when real data is scarce or sensitive, but its quality and biases need their own scrutiny.","names":{"plain":["Synthetic Data"],"technical":[{"name":"Synthetic Data","source":"STANFORD-HAI"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","HF"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"vector-database","function_statement":"The memory store behind most RAG and agent-memory systems.","names":{"plain":["Vector Database"],"technical":[{"name":"Vector Database","source":"STANFORD-HAI"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","OWASP-LLM"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"knowledge-graph","function_statement":"Gives agents a reliable, queryable map of how things relate; a structured alternative or complement to RAG.","names":{"plain":["Knowledge Graph"],"technical":[{"name":"Knowledge Graph","source":"STANFORD-HAI"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","HF"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"watermarking","function_statement":"Helps trace provenance of outputs and flag synthetic media. Imperfect, but raises the cost of undetected fakery.","names":{"plain":["Watermarking"],"technical":[{"name":"Watermarking","source":"NSA-CISA-DATA"}]},"cluster":"Data & Provenance","purpose":"check","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["NSA-CISA-DATA"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"sybil-resistance","function_statement":"In multi-agent settings, stops fake \"agreement\" from many sock-puppet sources. The Ledger names this directly: convergence counts only in proportion to how independent the sources truly are.","names":{"plain":["Sybil Resistance"],"technical":[{"name":"Sybil Resistance","source":"NSA-CISA-DATA"}]},"cluster":"Data & Provenance","purpose":"check","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["NSA-CISA-DATA"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"supply-chain","function_statement":"Each dependency is a way in for risk (OWASP LLM03). Governing agents means governing what they're built from and connect to.","names":{"plain":["Supply Chain"],"technical":[{"name":"Supply Chain","source":"OWASP-LLM"},{"name":"AI","source":"OWASP-LLM"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"non-resetting","reversibility":"reversible","observability":"latent"},"sources":["OWASP-LLM","NSA-CISA-DATA"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"hallucination","function_statement":"When an AI states something false as if it were true: fluent, confident output that isn't grounded in real data. Also called contextual mismatch or semantic drift.","names":{"plain":["Hallucination"],"technical":[{"name":"Hallucination","source":"STANFORD-HAI"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"undefined","reversibility":"undefined","observability":"latent"},"sources":["STANFORD-HAI","NIST","OWASP-LLM","MIT"],"status":"reviewed","prov":["NIST AI RMF"],"attributes_method":"preset","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"prompt-injection","function_statement":"The top-ranked LLM risk. Dangerous for agents because the injected instruction can trigger real tool actions.","names":{"plain":["Prompt Injection"],"technical":[{"name":"Prompt Injection","source":"STANFORD-HAI"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"undefined","reversibility":"undefined","observability":"latent"},"sources":["STANFORD-HAI","OWASP-LLM","NIST-AML"],"status":"reviewed","attributes_method":"preset","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"jailbreak","function_statement":"Close to prompt injection; specifically aimed at defeating guardrails. A reason guardrails alone aren't enough.","names":{"plain":["Jailbreak"],"technical":[{"name":"Jailbreak","source":"OWASP-LLM"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"non-resetting","reversibility":"irreversible","observability":"latent"},"sources":["OWASP-LLM","NIST-AML"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"agent-goal-hijack","function_statement":"Combines prompt injection with autonomy: multi-step execution amplifies the damage before a human can step in.","names":{"plain":["Agent Goal Hijack"],"technical":[{"name":"Agent Goal Hijack","source":"OWASP-LLM"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"non-resetting","reversibility":"irreversible","observability":"latent"},"sources":["OWASP-LLM","OWASP-AGENTIC","MIT"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Accelerator","sysdyn_method":"derived"},{"id":"goal-misalignment-goal-misgeneralization","function_statement":"A quiet, systemic failure: every individual step can look fine while the overall direction is wrong.","names":{"plain":["Goal Misalignment / Goal Misgeneralization"],"technical":[{"name":"Goal Misalignment / Goal Misgeneralization","source":"OWASP-LLM"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"non-resetting","reversibility":"irreversible","observability":"silent"},"sources":["OWASP-LLM","MIT"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"reward-hacking","function_statement":"When a system maximizes its given metric in unintended ways that satisfy the measure but not the real goal.","names":{"plain":["Reward Hacking"],"technical":[{"name":"Reward Hacking","source":"OWASP-LLM"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"non-resetting","reversibility":"irreversible","observability":"silent"},"sources":["OWASP-LLM","MIT"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Accelerator","sysdyn_method":"derived"},{"id":"deceptive-alignment","function_statement":"The hardest failure to detect: it's invisible to checks that only test individual claims. Maps to the Ledger's ordering drift: Purpose quietly outranking the Check, with every local test still green.","names":{"plain":["Deceptive Alignment"],"technical":[{"name":"Deceptive Alignment","source":"OWASP-LLM"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"undefined","reversibility":"undefined","observability":"silent"},"sources":["OWASP-LLM","MIT"],"status":"reviewed","attributes_method":"preset","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"model-drift","function_statement":"The gradual decline in a model's accuracy as real-world data drifts away from the data it was trained on.","names":{"plain":["Model Drift"],"technical":[{"name":"Model Drift","source":"STANFORD-HAI"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"org","cost_substrate":"non-resetting","reversibility":"reversible","observability":"latent"},"sources":["STANFORD-HAI","NIST-AI-RMF","NIST-AML","NSA-CISA-DATA","MIT"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"cascading-failure","function_statement":"When one part's failure sets off others in a chain, so a small fault spreads into a system-wide breakdown. Common where many agents or services depend on each other.","names":{"plain":["Cascading Failure"],"technical":[{"name":"Cascading Failure","source":"NSA-CISA-AGENTIC"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"non-resetting","reversibility":"irreversible","observability":"latent"},"sources":["NSA-CISA-AGENTIC","MIT"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Accelerator","sysdyn_method":"derived"},{"id":"tool-poisoning","function_statement":"Turns the agent's own capabilities against it. A supply-chain attack aimed at the action layer.","names":{"plain":["Tool Poisoning"],"technical":[{"name":"Tool Poisoning","source":"OWASP-LLM"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"non-resetting","reversibility":"irreversible","observability":"latent"},"sources":["OWASP-LLM","NIST-AML","OWASP-AGENTIC"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"data-poisoning","function_statement":"An attack that corrupts training data so the resulting model behaves incorrectly or carries hidden flaws.","names":{"plain":["Data Poisoning"],"technical":[{"name":"Data Poisoning","source":"OWASP-LLM"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"non-resetting","reversibility":"irreversible","observability":"latent"},"sources":["OWASP-LLM","NIST-AML","NSA-CISA-DATA"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"sensitive-information-disclosure","function_statement":"A leading real-world harm, especially for agents with access to internal data. Drives least-privilege and scoping.","names":{"plain":["Sensitive Information Disclosure","data leakage"],"technical":[{"name":"Sensitive Information Disclosure","source":"OWASP-LLM"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"org","cost_substrate":"non-resetting","reversibility":"irreversible","observability":"latent"},"sources":["OWASP-LLM","NIST-AML"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"privilege","function_statement":"Names the single most cited agentic risk and its fix: don't over-grant access.","names":{"plain":["Privilege"],"technical":[{"name":"Privilege","source":"NSA-CISA-AGENTIC"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"non-resetting","reversibility":"irreversible","observability":"latent"},"sources":["NSA-CISA-AGENTIC","OWASP-AGENTIC"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"bias","function_statement":"A core fairness and legal risk; often traces back to training data.","names":{"plain":["Bias"],"technical":[{"name":"Bias","source":"STANFORD-HAI"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"non-resetting","reversibility":"irreversible","observability":"latent"},"sources":["STANFORD-HAI","NIST","ISO","ISO-22989","NIST-AI-RMF","NIST-AML","MIT"],"status":"reviewed","prov":["NIST AI RMF / ISO/IEC TR 24027"],"attributes_method":"derived","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"automation-bias","function_statement":"The human tendency to over-trust automated output and stop checking it, which weakens human oversight.","names":{"plain":["Automation Bias"],"technical":[{"name":"Automation Bias","source":"OWASP-LLM"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"human","accountability":"human","cost_substrate":"undefined","reversibility":"undefined","observability":"latent"},"sources":["OWASP-LLM","MIT"],"status":"reviewed","attributes_method":"preset","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"overfitting","function_statement":"Explains why a model can ace tests yet fail in the real world.","names":{"plain":["Overfitting"],"technical":[{"name":"Overfitting","source":"STANFORD-HAI"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD-HAI","FOUND"],"status":"candidate","prov":["Core ML Theory"],"attributes_method":"curated","tier":"technical","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"emergent-behavior","function_statement":"Cuts both ways: useful new skills, but also unanticipated risks. A reason evaluation must be ongoing.","names":{"plain":["Emergent Behavior"],"technical":[{"name":"Emergent Behavior","source":"OWASP-LLM"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"non-resetting","reversibility":"irreversible","observability":"silent"},"sources":["OWASP-LLM","MIT"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Termination State","sysdyn_method":"derived"},{"id":"reinforcement-learning","function_statement":"A model learns by trial and error, rewarded for actions that move it toward a goal.","names":{"plain":["Reinforcement learning","learning by reward"],"technical":[{"name":"Reinforcement Learning","source":"STANFORD-HAI"},{"name":"RL","source":"STANFORD-HAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"undefined","reversibility":"undefined","observability":"undefined"},"sources":["STANFORD-HAI","ARXIV","ISO-22989"],"status":"reviewed","prov":["Sutton & Barto Textbook"],"attributes_method":"preset","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"rlhf","function_statement":"Humans rank a model's outputs and that preference signal trains the model toward what people actually want.","names":{"plain":["Learning from human feedback","training on human preference"],"technical":[{"name":"Reinforcement Learning from Human Feedback","source":"STANFORD-HAI"},{"name":"RLHF","source":"STANFORD-HAI"}]},"cluster":"Agentic Mechanics","purpose":"center","attributes":{"actor":"hybrid","accountability":"shared","cost_substrate":"non-resetting","reversibility":"undefined","observability":"latent"},"sources":["STANFORD-HAI","OPENAI"],"status":"reviewed","prov":["OpenAI (Human Feedback)"],"attributes_method":"preset","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"harness","function_statement":"Wraps a model to direct, constrain, and watch its actions: the control surface that turns a raw model into a governable agent.","names":{"plain":["Harness"],"technical":[]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"tether","function_statement":"Binds an agent to its goal, its ground truth, and human oversight, so it cannot drift loose.","names":{"plain":["Tether","leash to reality"],"technical":[]},"cluster":"Control & Safety","purpose":"center","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"buffer","function_statement":"Takes survivable error and deforms instead of breaking, holding room a node cannot shed.","names":{"plain":["Buffer"],"technical":[]},"cluster":"Control & Safety","purpose":"absorb","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER","FOUND"],"status":"reviewed","prov":["Reinforcement Learning"],"attributes_method":"derived","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"floor","function_statement":"Compares every claim to a fixed reference held still during the check.","names":{"plain":["Floor"],"technical":[]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"return","function_statement":"Keeps an append-only record so the system returns to baseline rather than locking in a changed state.","names":{"plain":["Return"],"technical":[]},"cluster":"Oversight & Governance","purpose":"reset","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Termination State","sysdyn_method":"derived"},{"id":"grace","function_statement":"Absorbs cost before a stop, so a recoverable error does not end the run.","names":{"plain":["Grace"],"technical":[]},"cluster":"Control & Safety","purpose":"absorb","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"forgiveness","function_statement":"Absorbs cost after a stop, making a valid intervention distinguishable from damage.","names":{"plain":["Forgiveness"],"technical":[]},"cluster":"Control & Safety","purpose":"reset","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Termination State","sysdyn_method":"derived"},{"id":"coherence","function_statement":"The whole structure holding its orientation as it moves, not merely keeping its facts: sustained coupling between what a system intends, what it does, and what is real.","names":{"plain":["Coherence","continuity of orientation"],"technical":[]},"cluster":"Oversight & Governance","purpose":"center","attributes":{"actor":"undefined","accountability":"undefined","cost_substrate":"non-resetting","reversibility":"undefined","observability":"undefined"},"sources":["RESILIENCE-LEDGER"],"status":"reviewed","attributes_method":"preset","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"acid","function_statement":"Database guarantees that keep transactions reliable.","names":{"plain":[],"technical":[{"name":"ACID","source":"ISO"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO"],"prov":["ANSI / ISO"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"awq","function_statement":"Activation-aware quantization that protects key weights.","names":{"plain":[],"technical":[{"name":"AWQ","source":"MIT"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["MIT"],"prov":["MIT"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"accuracy","function_statement":"The share of predictions a model gets right.","names":{"plain":["Accuracy"],"technical":[{"name":"Accuracy","source":"NIST"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"ai","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST","ISO"],"prov":["NIST AI RMF / ISO/IEC 22989"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"action","function_statement":"A choice the agent makes that changes its environment.","names":{"plain":["Action"],"technical":[{"name":"Action","source":"ARXIV"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Sutton & Barto"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"activation","function_statement":"A neuron's function that decides how strongly it fires.","names":{"plain":["Activation"],"technical":[{"name":"Activation","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Rosenblatt Perceptron"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"actor-critic","function_statement":"An RL method pairing a policy (actor) with a value estimate (critic).","names":{"plain":[],"technical":[{"name":"Actor-Critic","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Sutton & Barto"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"adapter","function_statement":"A small inserted module that adapts a frozen model.","names":{"plain":["Adapter"],"technical":[{"name":"Adapter","source":"HF"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["HF","OTHER"],"prov":["Hugging Face / PEFT"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"agentic","function_statement":"Describes a system that acts on goals with autonomy rather than only answering, often working through agents and subagents that each handle part of the task.","names":{"plain":["Agentic"],"technical":[{"name":"Agentic","source":"OPENAI"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OPENAI","ANTHROPIC","ARXIV"],"prov":["OpenAI / Anthropic"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"airflow","function_statement":"Schedules data pipelines as code with dependencies and retries, so recurring workflows run and recover without hand-tending.","names":{"plain":["Airflow"],"technical":[{"name":"Airflow","source":"APACHE"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["APACHE"],"prov":["Apache"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"algorithm","function_statement":"A defined procedure for solving a problem.","names":{"plain":["Algorithm"],"technical":[{"name":"Algorithm","source":"ISO"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO","ISO-22989"],"prov":["ISO/IEC 22989"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"anonymization","function_statement":"Removing identity so data cannot be traced to a person.","names":{"plain":["Anonymization"],"technical":[{"name":"Anonymization","source":"EU"}]},"cluster":"Data & Provenance","purpose":"check","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["EU","ISO"],"prov":["EU GDPR / ISO/IEC 27701"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"architecture","function_statement":"The overall design of a model or system.","names":{"plain":["Architecture"],"technical":[{"name":"Architecture","source":"ISO"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO","IEEE"],"prov":["ISO/IEC 42001 / IEEE"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"artifact","function_statement":"A produced file such as a trained model or dataset.","names":{"plain":["Artifact"],"technical":[{"name":"Artifact","source":"LF"}]},"cluster":"Oversight & Governance","purpose":"reset","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["LF"],"prov":["MLflow / Linux Foundation"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Termination State","sysdyn_method":"derived"},{"id":"attention","function_statement":"Lets a model weigh how much each part of the input matters to each other part.","names":{"plain":["Attention"],"technical":[{"name":"Attention","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Vaswani et al. (Transformer)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"audio","function_statement":"Sound as a data type a model processes.","names":{"plain":["Audio"],"technical":[{"name":"Audio","source":"ITU"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ITU"],"prov":["ITU-T"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"auditing","function_statement":"Reviewing a system against rules or expectations.","names":{"plain":["Auditing"],"technical":[{"name":"Auditing","source":"ISO"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"human","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO","NIST"],"prov":["ISO/IEC 42001 / NIST AI RMF"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"augmentation","function_statement":"Expanding data by making varied copies of examples.","names":{"plain":["Augmentation"],"technical":[{"name":"Augmentation","source":"FOUND"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["Core ML"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"automation","function_statement":"Running a task without human effort each time.","names":{"plain":["Automation"],"technical":[{"name":"Automation","source":"ISO"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO","ISA","ISO-22989"],"prov":["ISO / ISA"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"autonomous","function_statement":"Acting without step-by-step human direction.","names":{"plain":["Autonomous"],"technical":[{"name":"Autonomous","source":"SAE"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["SAE","NIST"],"prov":["SAE J3016 / NIST AI RMF"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"autoregressive","function_statement":"Generating output one token at a time, each from the last.","names":{"plain":["Autoregressive"],"technical":[{"name":"Autoregressive","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Transformer (GPT)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"bleu","function_statement":"Scores machine translation by overlap with reference text.","names":{"plain":[],"technical":[{"name":"BLEU","source":"IBM"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"ai","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["IBM"],"prov":["IBM"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"backpropagation","function_statement":"Propagates the error backward through a network to adjust its weights.","names":{"plain":["Backpropagation"],"technical":[{"name":"Backpropagation","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Rumelhart & Hinton"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"bagging","function_statement":"Training many models on data samples and averaging them.","names":{"plain":["Bagging"],"technical":[{"name":"Bagging","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Leo Breiman"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"balancing","function_statement":"Evening out class sizes so a model learns fairly.","names":{"plain":["Balancing"],"technical":[{"name":"Balancing","source":"OTHER"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["SMOTE"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"bandwidth","function_statement":"How much data can move per unit time.","names":{"plain":["Bandwidth"],"technical":[{"name":"Bandwidth","source":"IEEE"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["IEEE","ITU"],"prov":["IEEE / ITU-T"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"batch","function_statement":"A group of examples processed together in one training step.","names":{"plain":["Batch"],"technical":[{"name":"Batch","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["SGD Optimization"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"bidirectional","function_statement":"Reading a sequence in both directions for context.","names":{"plain":["Bidirectional"],"technical":[{"name":"Bidirectional","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["BERT Paper"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"boosting","function_statement":"Training models in sequence, each fixing the last's errors.","names":{"plain":["Boosting"],"technical":[{"name":"Boosting","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["AdaBoost"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"boundary","function_statement":"The limit of what a system may do.","names":{"plain":["Boundary"],"technical":[{"name":"Boundary","source":"FOUND"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND","OTHER"],"prov":["SVM / Decision Theory"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"clip","function_statement":"A model linking images and text in one space.","names":{"plain":[],"technical":[{"name":"CLIP","source":"OPENAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OPENAI"],"prov":["OpenAI"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"cuda","function_statement":"NVIDIA's platform for running compute on GPUs.","names":{"plain":[],"technical":[{"name":"CUDA","source":"NVIDIA"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["NVIDIA"],"prov":["NVIDIA"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"catalog","function_statement":"Indexes an organization's datasets with ownership and lineage attached, so people can find and trust data instead of duplicating it.","names":{"plain":["Catalog"],"technical":[{"name":"Catalog","source":"W3C"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["W3C"],"prov":["W3C DCAT"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"chain-of-thought","function_statement":"Prompting a model to reason step by step before answering.","names":{"plain":[],"technical":[{"name":"Chain-of-Thought","source":"GOOGLE"},{"name":"CoT","source":"GOOGLE"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["GOOGLE","ARXIV"],"prov":["Google"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"chroma","function_statement":"Stores embeddings for retrieval with near-zero setup, so a RAG prototype can run locally before any infrastructure exists.","names":{"plain":["Chroma"],"technical":[{"name":"Chroma","source":"OTHER"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Vector DB"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"chunking","function_statement":"Splitting documents into pieces sized for retrieval.","names":{"plain":["Chunking"],"technical":[{"name":"Chunking","source":"OTHER"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["LangChain / LlamaIndex"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"classification","function_statement":"Sorting inputs into discrete categories.","names":{"plain":["Classification"],"technical":[{"name":"Classification","source":"ISO"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO"],"prov":["ISO/IEC 22989"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"cleaning","function_statement":"Fixing errors and gaps so data is usable.","names":{"plain":["Cleaning"],"technical":[{"name":"Cleaning","source":"OTHER"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Data Engineering"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"cluster","function_statement":"A group of machines working as one, or a group of similar data.","names":{"plain":["Cluster"],"technical":[{"name":"Cluster","source":"IEEE"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["IEEE","OTHER"],"prov":["IEEE / Core Infra"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"clustering","function_statement":"Grouping similar items without labels.","names":{"plain":["Clustering"],"technical":[{"name":"Clustering","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["K-Means"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"cognitive","function_statement":"Describes work that needs reasoning, understanding, or judgment rather than rote processing; the kind of labor AI increasingly takes on.","names":{"plain":["Cognitive"],"technical":[{"name":"Cognitive","source":"IEEE"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["IEEE"],"prov":["IEEE 7000"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"collaboration","function_statement":"Agents or people working together toward a goal.","names":{"plain":["Collaboration"],"technical":[{"name":"Collaboration","source":"FIPA"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FIPA","ARXIV"],"prov":["FIPA"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"conflict","function_statement":"When goals or actions of agents collide.","names":{"plain":["Conflict"],"technical":[{"name":"Conflict","source":"FIPA"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FIPA"],"prov":["FIPA"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"consistency","function_statement":"Data staying coherent across a system.","names":{"plain":["Consistency"],"technical":[{"name":"Consistency","source":"OTHER"}]},"cluster":"Data & Provenance","purpose":"check","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER","ISO"],"prov":["CAP Theorem / ANSI"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"constraint","function_statement":"A rule limiting an agent's actions.","names":{"plain":["Constraint"],"technical":[{"name":"Constraint","source":"NIST"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST"],"prov":["NIST AI RMF"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"context","function_statement":"The surrounding information a model uses to respond.","names":{"plain":["Context"],"technical":[{"name":"Context","source":"ARXIV"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Vaswani et al. (Transformer)"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"contrastive","function_statement":"Learning by pulling similar items together, pushing others apart.","names":{"plain":["Contrastive"],"technical":[{"name":"Contrastive","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["LeCun (Self-Supervised)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"control","function_statement":"Keeping a system within intended bounds.","names":{"plain":["Control"],"technical":[{"name":"Control","source":"IEEE"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["IEEE","NIST"],"prov":["IEEE / NIST SP 800-53"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"convergence","function_statement":"When training settles and the loss stops improving.","names":{"plain":["Convergence"],"technical":[{"name":"Convergence","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["Optimization Math"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"critic","function_statement":"A component that scores or critiques an agent's output.","names":{"plain":["Critic"],"technical":[{"name":"Critic","source":"ARXIV"}]},"cluster":"Agentic Mechanics","purpose":"check","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Sutton & Barto"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"cross-entropy","function_statement":"Measures the gap between predicted and true distributions.","names":{"plain":[],"technical":[{"name":"Cross-Entropy","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV","FOUND"],"prov":["Shannon Theory / Core ML"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"dpo","function_statement":"Direct Preference Optimization: aligns a model from preferences without a separate reward model.","names":{"plain":[],"technical":[{"name":"DPO","source":"STANFORD"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["STANFORD"],"prov":["Stanford University"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"dqn","function_statement":"Deep Q-Network: Q-learning carried by a neural network.","names":{"plain":[],"technical":[{"name":"DQN","source":"GOOGLE"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["GOOGLE"],"prov":["DeepMind"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"dvc","function_statement":"Data Version Control: git-style versioning for data.","names":{"plain":[],"technical":[{"name":"DVC","source":"MLOPS"}]},"cluster":"Oversight & Governance","purpose":"reset","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["MLOPS"],"prov":["Open-Source MLOps"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Termination State","sysdyn_method":"derived"},{"id":"dataops","function_statement":"Practices for reliable, fast data pipelines.","names":{"plain":[],"technical":[{"name":"DataOps","source":"LF"}]},"cluster":"Oversight & Governance","purpose":"capability","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["LF"],"prov":["Linux Foundation"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"database","function_statement":"An organized store for structured data.","names":{"plain":["Database"],"technical":[{"name":"Database","source":"ISO"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO"],"prov":["ANSI/X3/SPARC"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"dataset","function_statement":"A collection of data used to train or test.","names":{"plain":["Dataset"],"technical":[{"name":"Dataset","source":"ISO"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO","W3C","ISO-22989"],"prov":["ISO/IEC 22989 / W3C"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"decision","function_statement":"A choice a model or agent commits to.","names":{"plain":["Decision"],"technical":[{"name":"Decision","source":"FOUND"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["Markov Decision Process"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"decoding","function_statement":"Turning model probabilities into output tokens.","names":{"plain":["Decoding"],"technical":[{"name":"Decoding","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND","OTHER"],"prov":["Information Theory / LLM"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"decomposition","function_statement":"Breaking a task into smaller subtasks.","names":{"plain":["Decomposition"],"technical":[{"name":"Decomposition","source":"FOUND"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND","ARXIV"],"prov":["Linear Algebra"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"deduplication","function_statement":"Removing duplicate records from a dataset.","names":{"plain":["Deduplication"],"technical":[{"name":"Deduplication","source":"OTHER"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["MinHash"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"deepspeed","function_statement":"A library for training large models efficiently.","names":{"plain":[],"technical":[{"name":"DeepSpeed","source":"MSFT"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["MSFT"],"prov":["Microsoft"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"degradation","function_statement":"A gradual decline in a system's performance.","names":{"plain":["Degradation"],"technical":[{"name":"Degradation","source":"MLOPS"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["MLOPS","ISO"],"prov":["MLOps / ISO/IEC 42001"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"delegation","function_statement":"Handing a task to another agent or tool.","names":{"plain":["Delegation"],"technical":[{"name":"Delegation","source":"FIPA"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FIPA","ARXIV"],"prov":["FIPA"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"delta","function_statement":"The change between two versions or values.","names":{"plain":["Delta"],"technical":[{"name":"Delta","source":"FOUND"}]},"cluster":"Oversight & Governance","purpose":"reset","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["Differential Calculus"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Termination State","sysdyn_method":"derived"},{"id":"denoising","function_statement":"Removing noise to recover clean data.","names":{"plain":["Denoising"],"technical":[{"name":"Denoising","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["Diffusion Models"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"density","function_statement":"How tightly packed information or connections are: how concentrated data points are in a space, how much meaning a representation packs into few numbers, or how many links a graph has between its parts. The opposite is sparse.","names":{"plain":["Density"],"technical":[{"name":"Density","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["Statistics / Probability"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"deployment","function_statement":"Putting a trained model into production use.","names":{"plain":["Deployment"],"technical":[{"name":"Deployment","source":"MLOPS"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["MLOPS","ISO","ISO-22989"],"prov":["MLOps / ISO/IEC 42001"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"detection","function_statement":"Surfacing a problem or pattern as it occurs.","names":{"plain":["Detection"],"technical":[{"name":"Detection","source":"FOUND"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND","ISO-42001"],"prov":["Signal Processing"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"diffusion","function_statement":"A generative method that denoises random noise into data.","names":{"plain":["Diffusion"],"technical":[{"name":"Diffusion","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Generative Models"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"dimension","function_statement":"One axis of a data space.","names":{"plain":["Dimension"],"technical":[{"name":"Dimension","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["Linear Algebra"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"discount","function_statement":"How much future reward is valued against immediate reward.","names":{"plain":["Discount"],"technical":[{"name":"Discount","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Bellman Equations"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"distillation","function_statement":"Training a small model to imitate a larger one.","names":{"plain":["Distillation"],"technical":[{"name":"Distillation","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Hinton et al."],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"docker","function_statement":"Packaging software with its dependencies so it runs anywhere.","names":{"plain":["Docker"],"technical":[{"name":"Docker","source":"OCI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OCI"],"prov":["OCI"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"drift","function_statement":"Data or behavior moving away from what a model was trained on.","names":{"plain":["Drift"],"technical":[{"name":"Drift","source":"MLOPS"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"non-resetting","reversibility":"irreversible","observability":"latent"},"sources":["MLOPS","NIST"],"prov":["MLOps / NIST AI RMF"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"dropout","function_statement":"Randomly drops units during training to reduce overfitting.","names":{"plain":["Dropout"],"technical":[{"name":"Dropout","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Srivastava & Hinton"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"etl","function_statement":"Extract, transform, load: moving data into a usable form.","names":{"plain":[],"technical":[{"name":"ETL","source":"OTHER"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Data Warehousing"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"efficient","function_statement":"Achieving results with less compute or data.","names":{"plain":["Efficient"],"technical":[{"name":"Efficient","source":"GOOGLE"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["GOOGLE","OTHER"],"prov":["Google (EfficientNet) / PEFT"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"elasticity","function_statement":"Scaling resources up and down with demand.","names":{"plain":["Elasticity"],"technical":[{"name":"Elasticity","source":"NIST"}]},"cluster":"Oversight & Governance","purpose":"capability","attributes":{"actor":"ai","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST"],"prov":["NIST SP 800-145"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"elicitation","function_statement":"Drawing out a model's capabilities or knowledge.","names":{"plain":["Elicitation"],"technical":[{"name":"Elicitation","source":"NIST"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST","OTHER"],"prov":["NIST AI RMF / Preference Alignment"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"encoding","function_statement":"Turning data into a numeric form a model can use.","names":{"plain":["Encoding"],"technical":[{"name":"Encoding","source":"ISO"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO","FOUND"],"prov":["ISO/IEC 8859 / Information Theory"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"engine","function_statement":"Runs the core computation of a system; the part everything else configures, feeds, or wraps.","names":{"plain":["Engine"],"technical":[{"name":"Engine","source":"NVIDIA"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["NVIDIA","APACHE"],"prov":["NVIDIA TensorRT / Apache"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"ensemble","function_statement":"Combining several models for a stronger prediction.","names":{"plain":["Ensemble"],"technical":[{"name":"Ensemble","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Machine Learning Foundations"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"entropy","function_statement":"Measures how unpredictable data is, which bounds how far it can be compressed and how confidently it can be modeled.","names":{"plain":["Entropy"],"technical":[{"name":"Entropy","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Shannon Theory"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"environment","function_statement":"The world the agent acts in and gets feedback from.","names":{"plain":["Environment"],"technical":[{"name":"Environment","source":"OPENAI"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OPENAI","OTHER"],"prov":["OpenAI Gym / Gymnasium"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"epoch","function_statement":"One full pass over the training data.","names":{"plain":["Epoch"],"technical":[{"name":"Epoch","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["Deep Learning Training"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"ethics","function_statement":"Principles for what AI should and should not do.","names":{"plain":["Ethics"],"technical":[{"name":"Ethics","source":"IEEE"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["IEEE","UNESCO"],"prov":["IEEE 7000 / UNESCO"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"evolution","function_statement":"Gradual change in a system or model over time.","names":{"plain":["Evolution"],"technical":[{"name":"Evolution","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Genetic Algorithms"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"exception","function_statement":"An error condition a program handles.","names":{"plain":["Exception"],"technical":[{"name":"Exception","source":"ISO"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO","OTHER"],"prov":["ISO/IEC C++ / Java Specs"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"execution","function_statement":"Carrying out a planned action.","names":{"plain":["Execution"],"technical":[{"name":"Execution","source":"OTHER"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER","OWASP-AGENTIC"],"prov":["Operating Systems"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"explainability","function_statement":"The degree to which a model's outputs, and the reasoning behind them, can be understood and explained by people.","names":{"plain":["Explainability","XAI","Explainable AI"],"technical":[{"name":"Explainability","source":"DARPA"},{"name":"XAI","source":"OTHER"},{"name":"Explainable AI","source":"OTHER"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["DARPA","NIST","ISO-22989","STANFORD-HAI","NIST-AI-RMF"],"prov":["DARPA XAI / NIST AI RMF","DARPA Explainable AI / NIST"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"exploitation","function_statement":"Choosing the best known action to maximize reward.","names":{"plain":["Exploitation"],"technical":[{"name":"Exploitation","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Multi-Armed Bandit"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"exploration","function_statement":"Trying new actions to learn more about the environment.","names":{"plain":["Exploration"],"technical":[{"name":"Exploration","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Multi-Armed Bandit"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"fsdp","function_statement":"Fully Sharded Data Parallel: splits a model across GPUs.","names":{"plain":[],"technical":[{"name":"FSDP","source":"META"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["META"],"prov":["PyTorch (Fully Sharded Data Parallel)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"fallbacks","function_statement":"Backup actions used when the main one fails.","names":{"plain":["Fallbacks"],"technical":[{"name":"Fallbacks","source":"W3C"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["W3C","OTHER"],"prov":["W3C / Architecture Patterns"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"fault","function_statement":"A flaw that can make a system fail.","names":{"plain":["Fault"],"technical":[{"name":"Fault","source":"IEEE"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["IEEE"],"prov":["IEEE Reliability Society"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"feature","function_statement":"A measurable input a model learns from.","names":{"plain":["Feature"],"technical":[{"name":"Feature","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND","ISO"],"prov":["Core ML / ISO/IEC 22989"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"feedforward","function_statement":"A plain layer that transforms each position independently.","names":{"plain":["Feedforward"],"technical":[{"name":"Feedforward","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Rosenblatt Perceptron"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"few-shot","function_statement":"Giving a few examples in the prompt to guide output.","names":{"plain":[],"technical":[{"name":"Few-Shot","source":"ARXIV"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV","HF"],"prov":["GPT-3 Paper"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"filtering","function_statement":"Removing unwanted items from data or output.","names":{"plain":["Filtering"],"technical":[{"name":"Filtering","source":"FOUND"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND","OTHER"],"prov":["Signal Processing / Collaborative"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"forest","function_statement":"Votes many randomized decision trees together so individual trees' errors cancel out; a strong, hard-to-overfit baseline for tabular data.","names":{"plain":["Forest"],"technical":[{"name":"Forest","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Random Forests (Breiman)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"gguf","function_statement":"A file format for running quantized models locally.","names":{"plain":[],"technical":[{"name":"GGUF","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["GGML Project (Gerganov)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"gptq","function_statement":"A method for quantizing model weights after training.","names":{"plain":[],"technical":[{"name":"GPTQ","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Post-Training Quantization"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"gpu","function_statement":"A chip that runs the parallel math AI training needs.","names":{"plain":[],"technical":[{"name":"GPU","source":"IEEE"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["IEEE"],"prov":["IEEE Computer Society"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"gru","function_statement":"Gates what a recurrent network keeps or forgets using fewer parameters than an LSTM, trading some capacity for speed.","names":{"plain":[],"technical":[{"name":"GRU","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Cho et al."],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"generation","function_statement":"Producing new output from a model.","names":{"plain":["Generation"],"technical":[{"name":"Generation","source":"NIST"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST","OPENAI"],"prov":["NIST AI RMF / OpenAI"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"generative","function_statement":"Able to produce new content.","names":{"plain":["Generative"],"technical":[{"name":"Generative","source":"EU"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["EU","NIST"],"prov":["EU AI Act / NIST AI RMF"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"generator","function_statement":"The part of a model that produces output.","names":{"plain":["Generator"],"technical":[{"name":"Generator","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["GAN (Goodfellow)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"git","function_statement":"Records every change to code with full history, so work can branch, merge, and roll back to any known-good state.","names":{"plain":["Git"],"technical":[{"name":"Git","source":"OTHER"}]},"cluster":"Oversight & Governance","purpose":"reset","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Software Freedom Conservancy"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Termination State","sysdyn_method":"derived"},{"id":"governance","function_statement":"The rules and oversight guiding how AI is built and used.","names":{"plain":["Governance"],"technical":[{"name":"Governance","source":"ISO"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO","NIST"],"prov":["ISO/IEC 42001 / NIST AI RMF"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"gradient","function_statement":"The slope that tells training which way to nudge weights to reduce error.","names":{"plain":["Gradient"],"technical":[{"name":"Gradient","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["Deep Learning Optimization"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"hbm","function_statement":"Feeds an accelerator data fast enough to keep large-model training and inference from starving; the memory bandwidth that often sets the real speed limit of AI hardware.","names":{"plain":[],"technical":[{"name":"HBM","source":"JEDEC"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["JEDEC"],"prov":["JEDEC (High Bandwidth Memory)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"hitc","function_statement":"Human-in-the-Command: a human directs the agent's actions.","names":{"plain":[],"technical":[{"name":"HITC","source":"DOD"}]},"cluster":"Control & Safety","purpose":"center","attributes":{"actor":"human","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["DOD","OWASP-AGENTIC"],"prov":["US DoD Directive"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"hotl","function_statement":"Human-on-the-Loop: a human monitors and can intervene.","names":{"plain":[],"technical":[{"name":"HOTL","source":"NATO"}]},"cluster":"Control & Safety","purpose":"center","attributes":{"actor":"human","accountability":"shared","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NATO","OTHER"],"prov":["NATO / UK MoD"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"hacking","function_statement":"Games the letter of a metric instead of the goal it stands for; shorthand for reward hacking.","names":{"plain":["Hacking"],"technical":[{"name":"Hacking","source":"NIST"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"non-resetting","reversibility":"irreversible","observability":"latent"},"sources":["NIST","OWASP"],"prov":["NIST SP 800-53 / OWASP"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Accelerator","sysdyn_method":"derived"},{"id":"hashing","function_statement":"Mapping data to a fixed-size fingerprint.","names":{"plain":["Hashing"],"technical":[{"name":"Hashing","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Core CS"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"huggingface","function_statement":"Hosts open models, datasets, and demos so anyone can download, fine-tune, or publish them; the public commons where most open AI work is shared.","names":{"plain":[],"technical":[{"name":"HuggingFace","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Model Registry Standard"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"human-in-the-command","function_statement":"A human directs and authorizes the agent's actions.","names":{"plain":[],"technical":[{"name":"Human-in-the-Command","source":"DOD"},{"name":"HITC","source":"DOD"}]},"cluster":"Control & Safety","purpose":"center","attributes":{"actor":"human","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["DOD","OWASP-AGENTIC"],"prov":["US DoD Ethical Principles"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"hyperparameter","function_statement":"A setting chosen before training, not learned.","names":{"plain":["Hyperparameter"],"technical":[{"name":"Hyperparameter","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["Core ML Optimization"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"hyperplane","function_statement":"The boundary a classifier draws between classes.","names":{"plain":["Hyperplane"],"technical":[{"name":"Hyperplane","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["Linear Algebra / SVM"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"imbalance","function_statement":"When some classes have far fewer examples than others.","names":{"plain":["Imbalance"],"technical":[{"name":"Imbalance","source":"OTHER"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["SMOTE (Chawla)"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"imputation","function_statement":"Filling in missing values with estimates.","names":{"plain":["Imputation"],"technical":[{"name":"Imputation","source":"FOUND"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND","OTHER"],"prov":["Statistics / Multiple Imputation"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"index","function_statement":"A structure that speeds up search over data.","names":{"plain":["Index"],"technical":[{"name":"Index","source":"ISO"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO"],"prov":["ANSI/NISO Z39.4"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"ingestion","function_statement":"Bringing data into a system for use.","names":{"plain":["Ingestion"],"technical":[{"name":"Ingestion","source":"APACHE"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["APACHE"],"prov":["Apache Data Pipelines"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"initialization","function_statement":"Setting starting weights before training.","names":{"plain":["Initialization"],"technical":[{"name":"Initialization","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["He / Glorot Weight Init"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"integration","function_statement":"Connecting systems so they work together.","names":{"plain":["Integration"],"technical":[{"name":"Integration","source":"ISO"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO"],"prov":["ISO/IEC 12207"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"interpretability","function_statement":"How well a model's inner workings can be understood.","names":{"plain":["Interpretability"],"technical":[{"name":"Interpretability","source":"DARPA"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["DARPA","NIST","ISO-22989","STANFORD-HAI"],"prov":["DARPA XAI / NIST AI RMF"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"kto","function_statement":"Kahneman-Tversky Optimization: aligns a model from simple good/bad signals.","names":{"plain":[],"technical":[{"name":"KTO","source":"OTHER"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER","ARXIV"],"prov":["Contextual AI (Kahneman-Tversky)"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"kafka","function_statement":"A streaming platform for moving data in real time.","names":{"plain":["Kafka"],"technical":[{"name":"Kafka","source":"APACHE"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["APACHE"],"prov":["Apache"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"kahneman-tversky","function_statement":"An alignment method using human-style gain/loss judgments.","names":{"plain":[],"technical":[{"name":"Kahneman-Tversky","source":"OTHER"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER","ARXIV"],"prov":["Behavioral Economics"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"kernel","function_statement":"A function that lets a model separate non-linear data.","names":{"plain":["Kernel"],"technical":[{"name":"Kernel","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["SVM (Schölkopf)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"kubernetes","function_statement":"Orchestrating containers across many machines.","names":{"plain":["Kubernetes"],"technical":[{"name":"Kubernetes","source":"CNCF"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["CNCF"],"prov":["CNCF"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"llmops","function_statement":"Practices for running language models in production.","names":{"plain":[],"technical":[{"name":"LLMOps","source":"LF"}]},"cluster":"Oversight & Governance","purpose":"capability","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["LF"],"prov":["Linux Foundation"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"lsh","function_statement":"Locality-sensitive hashing: fast approximate nearest-neighbor search.","names":{"plain":[],"technical":[{"name":"LSH","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Locality-Sensitive Hashing"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"lstm","function_statement":"A recurrent unit with gates that remember long-range dependencies.","names":{"plain":[],"technical":[{"name":"LSTM","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Hochreiter & Schmidhuber"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"label","function_statement":"The correct answer attached to a training example.","names":{"plain":["Label"],"technical":[{"name":"Label","source":"ISO"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO"],"prov":["ISO/IEC 22989"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"lake","function_statement":"A store for raw data of any kind.","names":{"plain":["Lake"],"technical":[{"name":"Lake","source":"OTHER"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Data Lakes / Enterprise Architecture"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"lakehouse","function_statement":"A store combining a lake's flexibility with a warehouse's structure.","names":{"plain":["Lakehouse"],"technical":[{"name":"Lakehouse","source":"DATABRICKS"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["DATABRICKS","OTHER"],"prov":["Databricks / Delta Lake"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"latency","function_statement":"The delay before a system responds.","names":{"plain":["Latency"],"technical":[{"name":"Latency","source":"IEEE"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["IEEE"],"prov":["IEEE Computer Society"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"latent","function_statement":"Holds the model's compressed internal representation of data; not directly readable, but where what the model has learned actually lives.","names":{"plain":["Latent"],"technical":[{"name":"Latent","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND","OTHER"],"prov":["Statistics / LDA"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"layer","function_statement":"One stage of a neural network's processing.","names":{"plain":["Layer"],"technical":[{"name":"Layer","source":"ISO"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO","FOUND"],"prov":["ISO/IEC OSI / Deep Learning"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"lineage","function_statement":"The traceable origin and transformations of data.","names":{"plain":["Lineage"],"technical":[{"name":"Lineage","source":"W3C"}]},"cluster":"Data & Provenance","purpose":"check","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["W3C"],"prov":["W3C PROV-O"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"lora","function_statement":"Low-Rank Adaptation: fine-tunes by training small added matrices.","names":{"plain":[],"technical":[{"name":"LoRA","source":"MSFT"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["MSFT"],"prov":["Microsoft (Low-Rank Adaptation)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"logging","function_statement":"Recording events a system produces.","names":{"plain":["Logging"],"technical":[{"name":"Logging","source":"W3C"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"ai","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["W3C","IETF"],"prov":["W3C / IETF Syslog"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"logits","function_statement":"A model's raw scores before they become probabilities.","names":{"plain":["Logits"],"technical":[{"name":"Logits","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND","OTHER"],"prov":["Statistics / Logistic Regression"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"long-term","function_statement":"Memory or planning that spans many steps.","names":{"plain":[],"technical":[{"name":"Long-Term","source":"OTHER"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Cognitive Psychology / Memory Buffers"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"loss","function_statement":"A number measuring how wrong a prediction is; training minimizes it.","names":{"plain":["Loss"],"technical":[{"name":"Loss","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND","OTHER"],"prov":["Optimization Theory / Statistical Decision"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"mdp","function_statement":"The formal model of an agent acting through states for reward.","names":{"plain":[],"technical":[{"name":"MDP","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Markov Decision Process (Bellman)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"mlops","function_statement":"Practices for deploying and maintaining ML in production.","names":{"plain":[],"technical":[{"name":"MLOps","source":"LF"}]},"cluster":"Oversight & Governance","purpose":"capability","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["LF","ISO"],"prov":["Linux Foundation / ISO/IEC 42001"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"management","function_statement":"Coordinating resources, models, or workflows.","names":{"plain":["Management"],"technical":[{"name":"Management","source":"ISO"}]},"cluster":"Oversight & Governance","purpose":"capability","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO","ISO-42001"],"prov":["ISO/IEC 42001"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"manifold","function_statement":"The lower-dimensional surface data tends to lie on.","names":{"plain":["Manifold"],"technical":[{"name":"Manifold","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["Differential Geometry"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"markov","function_statement":"The property that the future depends only on the present state.","names":{"plain":["Markov"],"technical":[{"name":"Markov","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["Probability Theory"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"masking","function_statement":"Hiding part of the input so a model learns to fill it.","names":{"plain":["Masking"],"technical":[{"name":"Masking","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV","OTHER"],"prov":["BERT / Transformer"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"matrix","function_statement":"A grid of numbers used in model math.","names":{"plain":["Matrix"],"technical":[{"name":"Matrix","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["Linear Algebra"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"megatron","function_statement":"A framework for training very large models across GPUs.","names":{"plain":["Megatron"],"technical":[{"name":"Megatron","source":"NVIDIA"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["NVIDIA"],"prov":["NVIDIA"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"metadata","function_statement":"Data describing other data.","names":{"plain":["Metadata"],"technical":[{"name":"Metadata","source":"ISO"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO"],"prov":["ISO/IEC 11179"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"milvus","function_statement":"Stores and searches embeddings at billion-vector scale, built for production clusters where retrieval speed under heavy load is the constraint.","names":{"plain":["Milvus"],"technical":[{"name":"Milvus","source":"LF"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["LF"],"prov":["LF AI & Data Foundation"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"modality","function_statement":"A type of data such as text, image, or audio.","names":{"plain":["Modality"],"technical":[{"name":"Modality","source":"OPENAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OPENAI","GOOGLE"],"prov":["Multi-Modal ML (OpenAI / Google)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"monitoring","function_statement":"Watching a deployed model's behavior over time.","names":{"plain":["Monitoring"],"technical":[{"name":"Monitoring","source":"ISO"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"ai","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO","MLOPS"],"prov":["ISO/IEC 42001 / MLOps"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"multimodal","function_statement":"Working across several data types at once.","names":{"plain":["Multimodal"],"technical":[{"name":"Multimodal","source":"EU"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["EU","OTHER"],"prov":["EU AI Act / Model Specs"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"npu","function_statement":"A chip specialized for neural network math.","names":{"plain":[],"technical":[{"name":"NPU","source":"IEEE"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["IEEE"],"prov":["IEEE Computer Society"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"negotiation","function_statement":"Agents resolving differences to reach agreement.","names":{"plain":["Negotiation"],"technical":[{"name":"Negotiation","source":"FIPA"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FIPA","ARXIV"],"prov":["FIPA Systems"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"neuron","function_statement":"A unit that sums weighted inputs and passes them through an activation.","names":{"plain":["Neuron"],"technical":[{"name":"Neuron","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["McCulloch & Pitts"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"node","function_statement":"A point in a network or a machine in a cluster.","names":{"plain":["Node"],"technical":[{"name":"Node","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Graph Theory"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"noise","function_statement":"Random variation that obscures the signal in data.","names":{"plain":["Noise"],"technical":[{"name":"Noise","source":"FOUND"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["Signal Processing / Diffusion"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"normalization","function_statement":"Rescales values so training stays stable.","names":{"plain":["Normalization"],"technical":[{"name":"Normalization","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Batch Norm (Ioffe & Szegedy)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"observability","function_statement":"How well a system's internal state can be seen from outside.","names":{"plain":["Observability"],"technical":[{"name":"Observability","source":"OTHER"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"ai","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER","CNCF"],"prov":["Kalman (Control Theory) / CNCF"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"ollama","function_statement":"A tool for running language models locally.","names":{"plain":["Ollama"],"technical":[{"name":"Ollama","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Edge Inference Engine"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"one-shot","function_statement":"Giving one example in the prompt to guide output.","names":{"plain":[],"technical":[{"name":"One-Shot","source":"OTHER"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER","ARXIV"],"prov":["Computer Vision Foundations"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"optimization","function_statement":"Adjusting parameters to minimize a loss.","names":{"plain":["Optimization"],"technical":[{"name":"Optimization","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["Convex Optimization (Boyd)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"outlier","function_statement":"A data point far from the rest.","names":{"plain":["Outlier"],"technical":[{"name":"Outlier","source":"FOUND"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["Statistics (Hawkins)"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"overlap","function_statement":"Sharing text between adjacent chunks to keep context.","names":{"plain":["Overlap"],"technical":[{"name":"Overlap","source":"FOUND"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND","OTHER"],"prov":["Signal Processing / RAG Chunks"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"peft","function_statement":"Parameter-efficient fine-tuning: adapt a model by training few parameters.","names":{"plain":[],"technical":[{"name":"PEFT","source":"HF"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["HF"],"prov":["Hugging Face Library"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"pomdp","function_statement":"An MDP where the agent sees its state only partially.","names":{"plain":[],"technical":[{"name":"POMDP","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Partially Observable MDP"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"ppo","function_statement":"Proximal Policy Optimization: a stable policy-gradient RL method.","names":{"plain":[],"technical":[{"name":"PPO","source":"OPENAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OPENAI"],"prov":["OpenAI (Proximal Policy Optimization)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"parallelism","function_statement":"Running parts of a computation at the same time.","names":{"plain":["Parallelism"],"technical":[{"name":"Parallelism","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Distributed AI Execution"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"parquet","function_statement":"A columnar file format for analytics data.","names":{"plain":["Parquet"],"technical":[{"name":"Parquet","source":"APACHE"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["APACHE"],"prov":["Apache Columnar Storage"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"partitioning","function_statement":"Dividing data into parts for storage or speed.","names":{"plain":["Partitioning"],"technical":[{"name":"Partitioning","source":"ISO"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO","OTHER"],"prov":["ANSI SQL / Database Storage"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"perceptron","function_statement":"The simplest neural unit: weighted inputs through an activation.","names":{"plain":["Perceptron"],"technical":[{"name":"Perceptron","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Rosenblatt"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"perplexity","function_statement":"A measure of how surprised a model is by text; lower is better.","names":{"plain":["Perplexity"],"technical":[{"name":"Perplexity","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND","OTHER"],"prov":["Information Theory / NLP"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"persistence","function_statement":"Keeping data after a process ends.","names":{"plain":["Persistence"],"technical":[{"name":"Persistence","source":"OTHER"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Software Patterns (Fowler)"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"pinecone","function_statement":"A managed vector database for similarity search.","names":{"plain":["Pinecone"],"technical":[{"name":"Pinecone","source":"OTHER"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Vector Database Standard"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"pipeline","function_statement":"An automated sequence of data and model steps.","names":{"plain":["Pipeline"],"technical":[{"name":"Pipeline","source":"ISO"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO","MLOPS"],"prov":["ISO/IEC 12207 / MLOps"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"plan","function_statement":"An ordered set of steps toward a goal.","names":{"plain":["Plan"],"technical":[{"name":"Plan","source":"NIST"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST","OTHER"],"prov":["NIST AI RMF / Automated Planning"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"policy","function_statement":"An agent's rule mapping each situation to an action.","names":{"plain":["Policy"],"technical":[{"name":"Policy","source":"FOUND"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND","NIST"],"prov":["Reinforcement Learning / NIST AI RMF"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"positional","function_statement":"Encoding that tells a model the order of tokens.","names":{"plain":["Positional"],"technical":[{"name":"Positional","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Vaswani et al. (Transformer)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"pre-training","function_statement":"Training a model broadly before specializing it.","names":{"plain":["Pre-training","Pretraining"],"technical":[{"name":"Pre-training","source":"OPENAI"},{"name":"Pretraining","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OPENAI","HF","ARXIV"],"prov":["OpenAI GPT Framework","BERT / GPT Foundations"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"precision","function_statement":"Of the items flagged positive, the share that truly are.","names":{"plain":["Precision"],"technical":[{"name":"Precision","source":"ISO"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"ai","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO"],"prov":["ISO/IEC 22989"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"prediction","function_statement":"A model's estimated output for an input.","names":{"plain":["Prediction"],"technical":[{"name":"Prediction","source":"ISO"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO"],"prov":["ISO/IEC 22989"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"preference","function_statement":"A human signal of which output is better.","names":{"plain":["Preference"],"technical":[{"name":"Preference","source":"NIST"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"human","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST","OTHER"],"prov":["NIST AI RMF / RLHF Alignment"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"prefix","function_statement":"Tokens prepended to steer a model's output.","names":{"plain":["Prefix"],"technical":[{"name":"Prefix","source":"OTHER"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER","ARXIV"],"prov":["Prefix-Tuning (PEFT)"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"privacy","function_statement":"Protecting personal information from misuse.","names":{"plain":["Privacy"],"technical":[{"name":"Privacy","source":"EU"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"org","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["EU","NIST","ISO-22989"],"prov":["EU GDPR / NIST SP 800-53"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"projection","function_statement":"Mapping data into fewer dimensions.","names":{"plain":["Projection"],"technical":[{"name":"Projection","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["Linear Algebra"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"prompting","function_statement":"Phrasing an input to steer a model's output.","names":{"plain":["Prompting"],"technical":[{"name":"Prompting","source":"OTHER"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"human","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER","ARXIV"],"prov":["Fine-Tuning Alternative"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"proximal","function_statement":"Trains a policy by reinforcement learning while clipping each update so behavior cannot lurch too far at once; the stability method (PPO) behind much preference tuning.","names":{"plain":["Proximal"],"technical":[{"name":"Proximal","source":"OPENAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OPENAI"],"prov":["OpenAI PPO"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"pruning","function_statement":"Cutting unneeded weights to shrink a model.","names":{"plain":["Pruning"],"technical":[{"name":"Pruning","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Optimal Brain Damage (LeCun)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"pseudonymization","function_statement":"Replacing identifiers with reversible tokens.","names":{"plain":["Pseudonymization"],"technical":[{"name":"Pseudonymization","source":"EU"}]},"cluster":"Data & Provenance","purpose":"check","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["EU","ISO"],"prov":["EU GDPR / ISO/IEC 27701"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"q-learning","function_statement":"Learning the value of actions so the agent can pick the best.","names":{"plain":[],"technical":[{"name":"Q-Learning","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Watkins (Reinforcement Learning)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"q-network","function_statement":"A neural network that estimates action values.","names":{"plain":[],"technical":[{"name":"Q-Network","source":"GOOGLE"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["GOOGLE"],"prov":["DeepMind (Deep Q-Learning)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"qlora","function_statement":"LoRA applied to a quantized model, cutting memory cost.","names":{"plain":[],"technical":[{"name":"QLoRA","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Quantized Low-Rank Adaptation"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"quantization","function_statement":"Shrinking a model by storing its weights at lower precision.","names":{"plain":["Quantization"],"technical":[{"name":"Quantization","source":"IEEE"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["IEEE","FOUND"],"prov":["IEEE / Deep Learning Compression"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"quantized","function_statement":"Stored at lower precision to save memory.","names":{"plain":["Quantized"],"technical":[{"name":"Quantized","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["GGUF / GGML Specifications"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"rlaif","function_statement":"Reinforcement learning where an AI model, not a human, gives the preference signal.","names":{"plain":[],"technical":[{"name":"RLAIF","source":"ANTHROPIC"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"latent"},"sources":["ANTHROPIC","ARXIV"],"prov":["Anthropic (Constitutional AI)"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"react","function_statement":"Interleaving reasoning with tool actions in a loop.","names":{"plain":[],"technical":[{"name":"ReAct","source":"GOOGLE"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["GOOGLE","ARXIV"],"prov":["Google (Reason and Act)"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"relu","function_statement":"An activation that passes positives and zeros out negatives.","names":{"plain":[],"technical":[{"name":"ReLU","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Rectified Linear Units (Hinton)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"reasoning","function_statement":"Working through a problem in steps before answering.","names":{"plain":["Reasoning"],"technical":[{"name":"Reasoning","source":"NIST"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST","OTHER"],"prov":["NIST AI RMF / Cognitive Architecture"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"recall","function_statement":"Of the truly positive items, the share the model caught.","names":{"plain":["Recall"],"technical":[{"name":"Recall","source":"ISO"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"ai","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO"],"prov":["ISO/IEC 22989 Metrics"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"recurrent","function_statement":"A network that carries state across a sequence.","names":{"plain":["Recurrent"],"technical":[{"name":"Recurrent","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["RNN Architectures"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"redaction","function_statement":"Removing sensitive content from data or output.","names":{"plain":["Redaction"],"technical":[{"name":"Redaction","source":"NIST"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST","EU"],"prov":["NIST SP 800-53 / GDPR"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"redundancy","function_statement":"Spare capacity that keeps a system running.","names":{"plain":["Redundancy"],"technical":[{"name":"Redundancy","source":"IEEE"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["IEEE","OTHER"],"prov":["IEEE / Enterprise Architecture"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"reflect","function_statement":"An agent reviewing its own work to improve it.","names":{"plain":["Reflect"],"technical":[{"name":"Reflect","source":"OTHER"}]},"cluster":"Agentic Mechanics","purpose":"check","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER","ARXIV"],"prov":["Reflexion Agentic Framework"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"registry","function_statement":"A catalog of models or artifacts with versions.","names":{"plain":["Registry"],"technical":[{"name":"Registry","source":"LF"}]},"cluster":"Oversight & Governance","purpose":"reset","attributes":{"actor":"ai","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["LF","HF"],"prov":["Linux Foundation / Hugging Face"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Termination State","sysdyn_method":"derived"},{"id":"regression","function_statement":"Predicting a continuous number from inputs.","names":{"plain":["Regression"],"technical":[{"name":"Regression","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Statistical Modeling"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"regularization","function_statement":"Penalizing complexity so a model generalizes better.","names":{"plain":["Regularization"],"technical":[{"name":"Regularization","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER","FOUND"],"prov":["Tikhonov / Core ML"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"retraining","function_statement":"Training a model again on fresh data.","names":{"plain":["Retraining"],"technical":[{"name":"Retraining","source":"ISO"}]},"cluster":"Oversight & Governance","purpose":"reset","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO","MLOPS"],"prov":["ISO/IEC 42001 / MLOps"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Termination State","sysdyn_method":"derived"},{"id":"retrieval","function_statement":"Fetching relevant external documents to ground an answer.","names":{"plain":["Retrieval"],"technical":[{"name":"Retrieval","source":"OTHER"}]},"cluster":"Data & Provenance","purpose":"check","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER","ARXIV"],"prov":["Information Retrieval / Vector DB"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"reward","function_statement":"The signal telling a learning agent how good an action was.","names":{"plain":["Reward"],"technical":[{"name":"Reward","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Sutton & Barto Bellman Framework"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"risk","function_statement":"The chance and severity of harm.","names":{"plain":["Risk"],"technical":[{"name":"Risk","source":"ISO"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"undefined","cost_substrate":"non-resetting","reversibility":"irreversible","observability":"latent"},"sources":["ISO","NIST","ISO-22989"],"prov":["ISO 31000 / NIST AI RMF"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"smote","function_statement":"Synthesizing minority examples to balance a dataset.","names":{"plain":[],"technical":[{"name":"SMOTE","source":"OTHER"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Synthetic Minority Over-sampling"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"safety","function_statement":"Keeping a system from causing harm.","names":{"plain":["Safety"],"technical":[{"name":"Safety","source":"NIST"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST","ISO","OPENAI","ISO-22989","NIST-AI-RMF"],"prov":["NIST AI RMF / ISO/IEC 42001 / OpenAI"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"sampling","function_statement":"Picking the next token from the model's probabilities.","names":{"plain":["Sampling"],"technical":[{"name":"Sampling","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND","OTHER"],"prov":["Statistics / Nucleus Sampling"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"scalability","function_statement":"How well a system grows with load.","names":{"plain":["Scalability"],"technical":[{"name":"Scalability","source":"NIST"}]},"cluster":"Oversight & Governance","purpose":"capability","attributes":{"actor":"ai","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST","OTHER"],"prov":["NIST Cloud Computing / Software Patterns"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"scaling","function_statement":"Growing a model or system in size or capacity.","names":{"plain":["Scaling"],"technical":[{"name":"Scaling","source":"OPENAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OPENAI"],"prov":["OpenAI Scaling Laws"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"schema","function_statement":"The defined structure of data.","names":{"plain":["Schema"],"technical":[{"name":"Schema","source":"W3C"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["W3C","OTHER"],"prov":["W3C XML / Database Standards"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"scraping","function_statement":"Extracting data from web pages.","names":{"plain":["Scraping"],"technical":[{"name":"Scraping","source":"W3C"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["W3C","OTHER"],"prov":["W3C / Legal Precedents"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"search","function_statement":"Finding relevant items in data or a space.","names":{"plain":["Search"],"technical":[{"name":"Search","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Russell & Norvig / Vector Search"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"security","function_statement":"Protecting systems and data from attack.","names":{"plain":["Security"],"technical":[{"name":"Security","source":"ISO"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO","NIST","ISO-22989"],"prov":["ISO/IEC 27001 / NIST SP 800-53"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"self-reflection","function_statement":"An agent critiques its own output and revises it.","names":{"plain":[],"technical":[{"name":"Self-Reflection","source":"ANTHROPIC"}]},"cluster":"Agentic Mechanics","purpose":"check","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ANTHROPIC","ARXIV"],"prov":["Anthropic Constitutional / Reflexion Paper"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"semantic","function_statement":"Based on meaning rather than exact words.","names":{"plain":["Semantic"],"technical":[{"name":"Semantic","source":"W3C"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["W3C","OTHER"],"prov":["W3C Semantic Web / Embeddings"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"sensitive","function_statement":"Data needing special protection, such as health or identity.","names":{"plain":["Sensitive"],"technical":[{"name":"Sensitive","source":"EU"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["EU"],"prov":["EU GDPR Article 9"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"server","function_statement":"A machine that hosts a service or model.","names":{"plain":["Server"],"technical":[{"name":"Server","source":"W3C"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["W3C","OTHER"],"prov":["W3C HTTP / Triton Inference"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"session","function_statement":"A bounded span of interaction with a system.","names":{"plain":["Session"],"technical":[{"name":"Session","source":"IETF"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["IETF"],"prov":["IETF RFC 2616"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"shaping","function_statement":"Designing extra rewards to guide an agent's learning.","names":{"plain":["Shaping"],"technical":[{"name":"Shaping","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Reward Shaping (Ng & Russell)"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"sharded","function_statement":"Split across machines to fit or to scale.","names":{"plain":["Sharded"],"technical":[{"name":"Sharded","source":"META"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["META"],"prov":["PyTorch FSDP"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"sharding","function_statement":"Splitting a model or data across machines.","names":{"plain":["Sharding"],"technical":[{"name":"Sharding","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Database Scalability Patterns"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"short-term","function_statement":"Memory holding the current context only.","names":{"plain":[],"technical":[{"name":"Short-Term","source":"OTHER"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Cognitive Models / Interaction Buffers"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"spark","function_statement":"A framework for large-scale data processing.","names":{"plain":["Spark"],"technical":[{"name":"Spark","source":"APACHE"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["APACHE"],"prov":["Apache Framework"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"sparsity","function_statement":"How many of a model's values are zero.","names":{"plain":["Sparsity"],"technical":[{"name":"Sparsity","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Network Compression"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"speech","function_statement":"Processing spoken language.","names":{"plain":["Speech"],"technical":[{"name":"Speech","source":"ITU"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ITU"],"prov":["ITU-T Coding"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"speech-to-text","function_statement":"Converting spoken audio into written text.","names":{"plain":[],"technical":[{"name":"Speech-to-Text","source":"W3C"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["W3C","OPENAI"],"prov":["W3C Speech API / OpenAI Whisper"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"staleness","function_statement":"How out of date a model or its data has become.","names":{"plain":["Staleness"],"technical":[{"name":"Staleness","source":"MLOPS"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["MLOPS","GOOGLE"],"prov":["MLOps Monitoring / Google TFX"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"standardization","function_statement":"Rescaling features to a common mean and spread.","names":{"plain":["Standardization"],"technical":[{"name":"Standardization","source":"ISO"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO","FOUND"],"prov":["ISO/IEC / Core Statistics"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"state","function_statement":"The agent's current situation, the basis for its next action.","names":{"plain":["State"],"technical":[{"name":"State","source":"OTHER"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER","W3C"],"prov":["MDP / W3C REST State"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"stochasticity","function_statement":"Built-in randomness in a model's behavior.","names":{"plain":["Stochasticity"],"technical":[{"name":"Stochasticity","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["Stochastic Gradient Descent Theory"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"store","function_statement":"A system that holds data or features.","names":{"plain":["Store"],"technical":[{"name":"Store","source":"LF"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["LF"],"prov":["Linux Foundation Feast"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"stratification","function_statement":"Splitting data so each part keeps the same class mix.","names":{"plain":["Stratification"],"technical":[{"name":"Stratification","source":"FOUND"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND","OTHER"],"prov":["Statistics / Cross-Validation"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"streaming","function_statement":"Processing data continuously as it arrives.","names":{"plain":["Streaming"],"technical":[{"name":"Streaming","source":"APACHE"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["APACHE","OTHER"],"prov":["Apache Kafka / Flink"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"supervised","function_statement":"Learning from labeled examples.","names":{"plain":["Supervised"],"technical":[{"name":"Supervised","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND","ISO","ISO-22989"],"prov":["Core ML / ISO/IEC 22989"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"supervision","function_statement":"Human oversight of a model or agent.","names":{"plain":["Supervision"],"technical":[{"name":"Supervision","source":"FIPA"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"human","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["FIPA","ISO-42001"],"prov":["FIPA Governance Systems"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"synthesis","function_statement":"Generating new content such as speech or images.","names":{"plain":["Synthesis"],"technical":[{"name":"Synthesis","source":"ITU"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ITU","OTHER"],"prov":["ITU-T Audio / Speech Gen"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"synthetic","function_statement":"Artificially generated data used in place of real data.","names":{"plain":["Synthetic"],"technical":[{"name":"Synthetic","source":"NIST"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["NIST","EU"],"prov":["NIST AI RMF / GDPR"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"tanh","function_statement":"An activation that squashes values to between minus one and one.","names":{"plain":["Tanh"],"technical":[{"name":"Tanh","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Hyperbolic Tangent Activation"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"telemetry","function_statement":"The signals a running system emits about itself.","names":{"plain":["Telemetry"],"technical":[{"name":"Telemetry","source":"CNCF"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"ai","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["CNCF"],"prov":["CNCF OpenTelemetry Framework"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"tensor","function_statement":"A multi-dimensional array, the basic unit of model math.","names":{"plain":["Tensor"],"technical":[{"name":"Tensor","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND","OTHER"],"prov":["Linear Algebra / TensorFlow"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"tensorrt","function_statement":"NVIDIA's runtime for fast model inference.","names":{"plain":[],"technical":[{"name":"TensorRT","source":"NVIDIA"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["NVIDIA"],"prov":["NVIDIA Inference Engine"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"testing","function_statement":"Checking a system behaves as intended.","names":{"plain":["Testing"],"technical":[{"name":"Testing","source":"ISO"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO","ISO-22989"],"prov":["ISO/IEC 12207 Quality Eng"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"text-to-speech","function_statement":"Converting written text into spoken audio.","names":{"plain":[],"technical":[{"name":"Text-to-Speech","source":"ITU"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ITU","W3C"],"prov":["ITU-T H.Series / W3C"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"throughput","function_statement":"How much work a system completes per unit time.","names":{"plain":["Throughput"],"technical":[{"name":"Throughput","source":"IEEE"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["IEEE"],"prov":["IEEE Computer Society"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"token","function_statement":"A chunk of text a model reads or writes. Usage limits and pricing follow from it: most AI services bill and budget by tokens.","names":{"plain":["Token"],"technical":[{"name":"Token","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV","OTHER","STANFORD-HAI","HF"],"prov":["Vaswani et al. / Byte-Pair Encoding"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"tokenization","function_statement":"Splitting text into tokens for a model.","names":{"plain":["Tokenization"],"technical":[{"name":"Tokenization","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER","OPENAI"],"prov":["Chomsky Linguistics / OpenAI Tiktoken"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"tolerance","function_statement":"Keeps a system running despite component failures, absorbing faults instead of turning them into outages.","names":{"plain":["Tolerance"],"technical":[{"name":"Tolerance","source":"IEEE"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["IEEE","NIST"],"prov":["IEEE Reliability / NIST Controls"],"status":"candidate","attributes_method":"derived","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"tool","function_statement":"An external function an agent can call.","names":{"plain":["Tool"],"technical":[{"name":"Tool","source":"OPENAI"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OPENAI","ARXIV"],"prov":["OpenAI Function Calling Specs"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"top-k","function_statement":"Sampling from the k most likely tokens.","names":{"plain":[],"technical":[{"name":"Top-K","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["NLP Model Sampling"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"top-p","function_statement":"Sampling from the smallest set of tokens covering probability p.","names":{"plain":[],"technical":[{"name":"Top-P","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Nucleus Sampling"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"tracing","function_statement":"Following a request through a system's steps.","names":{"plain":["Tracing"],"technical":[{"name":"Tracing","source":"CNCF"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"ai","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["CNCF"],"prov":["CNCF OpenTelemetry"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"tracking","function_statement":"Recording experiment runs and their results.","names":{"plain":["Tracking"],"technical":[{"name":"Tracking","source":"LF"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"ai","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["LF","OTHER"],"prov":["MLflow / Weights & Biases"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"training","function_statement":"Adjusting a model's weights from data.","names":{"plain":["Training"],"technical":[{"name":"Training","source":"ISO"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO","ARXIV"],"prov":["ISO/IEC 22989 Lifecycle"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"trajectory","function_statement":"The sequence of states and actions an agent takes.","names":{"plain":["Trajectory"],"technical":[{"name":"Trajectory","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV"],"prov":["Sutton & Barto Reinforcement Learning"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"tree-of-thoughts","function_statement":"Exploring multiple reasoning branches and keeping the best.","names":{"plain":[],"technical":[{"name":"Tree-of-Thoughts","source":"OTHER"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER","ARXIV"],"prov":["Prompt Engineering Search"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"trigger","function_statement":"An event that starts an automated action.","names":{"plain":["Trigger"],"technical":[{"name":"Trigger","source":"OTHER"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER","MLOPS"],"prov":["Software Patterns / MLOps Pipeline Controls"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"triton","function_statement":"NVIDIA's server for hosting models at scale.","names":{"plain":["Triton"],"technical":[{"name":"Triton","source":"OPENAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OPENAI"],"prov":["OpenAI Triton GPU Programming"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"tuning","function_statement":"Adjusting a trained model for a specific task.","names":{"plain":["Tuning"],"technical":[{"name":"Tuning","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND"],"prov":["Deep Learning Fine-Tuning"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"unmasking","function_statement":"Predicting masked tokens during training.","names":{"plain":["Unmasking"],"technical":[{"name":"Unmasking","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["NLP Masked Language Modeling"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"unsupervised","function_statement":"Finding structure in data without labels.","names":{"plain":["Unsupervised"],"technical":[{"name":"Unsupervised","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["FOUND","ISO","ISO-22989"],"prov":["Core ML / ISO/IEC 22989"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"vram","function_statement":"A GPU's onboard memory, limiting model size.","names":{"plain":[],"technical":[{"name":"VRAM","source":"JEDEC"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["JEDEC"],"prov":["JEDEC Memory Standards"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"validation","function_statement":"Checking that data or a model meets requirements.","names":{"plain":["Validation"],"technical":[{"name":"Validation","source":"ISO"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ISO","ISO-22989","NIST-AI-RMF"],"prov":["ISO/IEC 22989 Verification"],"status":"reviewed","attributes_method":"curated","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"value","function_statement":"The expected long-run reward from a state.","names":{"plain":["Value"],"technical":[{"name":"Value","source":"ARXIV"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["ARXIV","FOUND"],"prov":["Bellman Equations / Reinforcement Learning"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"variance","function_statement":"How much predictions scatter; high variance is overfitting.","names":{"plain":["Variance"],"technical":[{"name":"Variance","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Bias-Variance Tradeoff Equations"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"versioning","function_statement":"Tracking changes to data, models, or code over time.","names":{"plain":["Versioning"],"technical":[{"name":"Versioning","source":"OTHER"}]},"cluster":"Oversight & Governance","purpose":"reset","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["DVC Framework"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Termination State","sysdyn_method":"derived"},{"id":"vision","function_statement":"Processing and understanding images.","names":{"plain":["Vision"],"technical":[{"name":"Vision","source":"IEEE"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["IEEE","OTHER"],"prov":["IEEE Computer Society / Core CV"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"vocabulary","function_statement":"The full set of tokens a model knows.","names":{"plain":["Vocabulary"],"technical":[{"name":"Vocabulary","source":"OTHER"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["NLP Model Standards"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"wandb","function_statement":"Weights & Biases: a tool for tracking experiments.","names":{"plain":[],"technical":[{"name":"WandB","source":"OTHER"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Weights & Biases"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"warehouse","function_statement":"A structured store for analytics-ready data.","names":{"plain":["Warehouse"],"technical":[{"name":"Warehouse","source":"OTHER"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Data Warehousing Standards"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"weaviate","function_statement":"Stores embeddings alongside structured object data, so retrieval can combine meaning-based search with exact filters.","names":{"plain":["Weaviate"],"technical":[{"name":"Weaviate","source":"OTHER"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER"],"prov":["Vector Engine Specification"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"weightsbiases","function_statement":"Tracks experiments, metrics, and model versions across training runs so results stay reproducible and comparable.","names":{"plain":[],"technical":[{"name":"WeightsBiases","source":"MLOPS"}]},"cluster":"Oversight & Governance","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["MLOPS"],"prov":["MLOps Metric Monitoring"],"status":"candidate","attributes_method":"curated","tier":"technical","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"whisper","function_statement":"OpenAI's speech-to-text model.","names":{"plain":["Whisper"],"technical":[{"name":"Whisper","source":"OPENAI"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OPENAI"],"prov":["OpenAI Audio Architecture"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"zero-shot","function_statement":"Asking with no examples, relying on prior training.","names":{"plain":[],"technical":[{"name":"Zero-Shot","source":"OTHER"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"shared","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["OTHER","ARXIV"],"prov":["Generative Model Task Transfer"],"status":"reviewed","attributes_method":"derived","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"vllm","function_statement":"A fast engine for serving large language models.","names":{"plain":[],"technical":[{"name":"vLLM","source":"BERKELEY"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":["BERKELEY"],"prov":["UC Berkeley PagedAttention Engine"],"status":"candidate","attributes_method":"derived","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"resilience-ledger","function_statement":"Measures how a system absorbs load, checks against a reference, and resets, scored from a neutral center rather than its own story.","names":{"plain":["Resilience Ledger"],"technical":[{"name":"Resilience Ledger","source":"RESILIENCE-LEDGER"},{"name":"Ledger","source":"RESILIENCE-LEDGER"}]},"cluster":"Oversight & Governance","purpose":"check","tier":"core","attributes":{"actor":"human","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER"],"prov":["Resilience Ledger (author framework)"],"status":"reviewed","attributes_method":"curated","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"coherence-ledger","function_statement":"Audits any framework for coherence and resilience to its stated goal, scoring structure over narrative.","names":{"plain":["Coherence Ledger"],"technical":[{"name":"Coherence Ledger","source":"RESILIENCE-LEDGER"}]},"cluster":"Oversight & Governance","purpose":"check","tier":"core","attributes":{"actor":"human","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER"],"prov":["Resilience Ledger (author framework)"],"status":"reviewed","attributes_method":"curated","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"sovereign-zero","function_statement":"The neutral, function-first reference a translation calibrates against, so no single vocabulary captures the meaning.","names":{"plain":["Sovereign Zero"],"technical":[{"name":"Sovereign Zero","source":"RESILIENCE-LEDGER"},{"name":"Neutral Center","source":"RESILIENCE-LEDGER"}]},"cluster":"Oversight & Governance","purpose":"center","tier":"core","attributes":{"actor":"human","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER"],"prov":["Resilience Ledger (author framework)"],"status":"reviewed","attributes_method":"curated","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"context-engineering","function_statement":"Designs what enters a model's context: retrieval, memory, and prompts assembled so the output stays grounded.","names":{"plain":["Context Engineering"],"technical":[{"name":"Context Engineering","source":"RESILIENCE-LEDGER"}]},"cluster":"Data & Provenance","purpose":"check","tier":"core","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER"],"prov":["Author synthesis (agentic AI governance framing)"],"status":"reviewed","attributes_method":"curated","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"solution-architecture","function_statement":"The overall design of how an agentic system's models, tools, data, controls, and people fit together to meet a goal safely.","names":{"plain":["Solution Architecture"],"technical":[{"name":"Solution Architecture","source":"RESILIENCE-LEDGER"}]},"cluster":"Oversight & Governance","purpose":"capability","tier":"core","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER"],"prov":["Author synthesis (agentic AI governance framing)"],"status":"reviewed","attributes_method":"curated","sysdyn":"Input","sysdyn_method":"derived"},{"id":"systems-thinking","function_statement":"Reasoning about the whole agentic system and its feedback loops, not isolated parts, so controls and risks are seen in relation.","names":{"plain":["Systems Thinking"],"technical":[{"name":"Systems Thinking","source":"RESILIENCE-LEDGER"}]},"cluster":"Oversight & Governance","purpose":"check","tier":"core","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER"],"prov":["Author synthesis (agentic AI governance framing)"],"status":"reviewed","attributes_method":"curated","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"scaffold","function_statement":"The supporting structure of prompts, tools, memory, and checks around a model that both enables and bounds agentic behavior.","names":{"plain":["Scaffold"],"technical":[{"name":"Scaffold","source":"RESILIENCE-LEDGER"},{"name":"Scaffolding","source":"RESILIENCE-LEDGER"}]},"cluster":"Agentic Mechanics","purpose":"capability","tier":"core","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER"],"prov":["Author synthesis (agentic AI governance framing)"],"status":"reviewed","attributes_method":"curated","sysdyn":"Input","sysdyn_method":"derived"},{"id":"frontier-ai","function_statement":"The most capable, newest models at the leading edge of ability and risk, the focus of heightened governance.","names":{"plain":["Frontier AI"],"technical":[{"name":"Frontier AI","source":"RESILIENCE-LEDGER"}]},"cluster":"Foundations","purpose":"capability","tier":"core","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER"],"prov":["Author synthesis (agentic AI governance framing)"],"status":"reviewed","attributes_method":"curated","sysdyn":"Input","sysdyn_method":"derived"},{"id":"go-to-market","function_statement":"The plan for releasing an AI capability to users, governed for staged, responsible rollout rather than a single launch.","names":{"plain":["Go-to-Market"],"technical":[{"name":"Go-to-Market","source":"RESILIENCE-LEDGER"},{"name":"GTM","source":"RESILIENCE-LEDGER"}]},"cluster":"Oversight & Governance","purpose":"capability","tier":"core","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER"],"prov":["Author synthesis (agentic AI governance framing)"],"status":"reviewed","attributes_method":"curated","sysdyn":"Input","sysdyn_method":"derived"},{"id":"loops","function_statement":"The feedback and control cycles an agent runs in: acting, being checked, and correcting, including human-in and human-on-the-loop.","names":{"plain":["Loops","Loop"],"technical":[{"name":"Loops","source":"RESILIENCE-LEDGER"},{"name":"Control Loop","source":"RESILIENCE-LEDGER"},{"name":"Feedback Loop","source":"RESILIENCE-LEDGER"},{"name":"Loop","source":"OTHER"}]},"cluster":"Agentic Mechanics","purpose":"check","tier":"core","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER","OTHER"],"prov":["Author synthesis (agentic AI governance framing)","OODA Loop / Agentic Patterns"],"status":"reviewed","attributes_method":"curated","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"skills","function_statement":"Discrete, scoped capabilities an agent can call: modular tools it composes, each permissioned and bounded.","names":{"plain":["Skills"],"technical":[{"name":"Skills","source":"RESILIENCE-LEDGER"}]},"cluster":"Agentic Mechanics","purpose":"capability","tier":"core","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER"],"prov":["Author synthesis (agentic AI governance framing)"],"status":"reviewed","attributes_method":"curated","sysdyn":"Input","sysdyn_method":"derived"},{"id":"query-concept","function_statement":"A structured request that asks the system by pattern and function, not by keyword, so the answer reflects the structure.","names":{"plain":["Query"],"technical":[{"name":"Query","source":"RESILIENCE-LEDGER"}]},"cluster":"Oversight & Governance","purpose":"check","tier":"core","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER"],"prov":["Author synthesis (agentic AI governance framing)"],"status":"reviewed","attributes_method":"curated","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"resilience","function_statement":"A system's capacity to absorb load, hold its chain of custody, stay anchored to ground truth, and recover, while respecting the human cost.","names":{"plain":["Resilience"],"technical":[{"name":"Resilience","source":"RESILIENCE-LEDGER"},{"name":"System Resilience","source":"RESILIENCE-LEDGER"},{"name":"Data Resilience","source":"RESILIENCE-LEDGER"}]},"cluster":"Oversight & Governance","purpose":"check","tier":"core","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER"],"prov":["Author synthesis (agentic AI governance framing)"],"status":"reviewed","attributes_method":"curated","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"chain-of-custody","function_statement":"The unbroken, traceable record of where data and decisions came from and how they changed, preserved end to end.","names":{"plain":["Chain of Custody"],"technical":[{"name":"Chain of Custody","source":"RESILIENCE-LEDGER"}]},"cluster":"Data & Provenance","purpose":"check","tier":"core","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":["RESILIENCE-LEDGER"],"prov":["Author synthesis (agentic AI governance framing)"],"status":"reviewed","attributes_method":"curated","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"state-delta-architecture","function_statement":"A way of building agentic systems where the AI proposes changes but a fixed, local rule-layer must verify each one against current reality before it can take effect. The model reasons; the structure decides what commits.","names":{"plain":["State-Delta Architecture","State-Delta"],"technical":[{"name":"State-Delta Architecture","source":"STATE-DELTA"}]},"cluster":"Foundations","purpose":"center","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA","NSA-CISA-AGENTIC"],"status":"candidate","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"friction-principle","function_statement":"The design stance that a calibrated amount of resistance, validation that can refuse, is what gives a system traction under load. Too little and it drifts; too much and it consumes itself.","names":{"plain":["Friction Principle"],"technical":[{"name":"Friction Principle","source":"STATE-DELTA"}]},"cluster":"Foundations","purpose":"center","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA"],"status":"candidate","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"system-state","function_statement":"What a system currently treats as true about its environment: the single current model it acts on, not the pile of everything it has ever seen. The map the GPS is using now, not every map it has held.","names":{"plain":["State"],"technical":[{"name":"State","source":"STATE-DELTA"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA"],"status":"candidate","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"state-delta","function_statement":"A small structured record of one change: what changed, where, and when, carrying only the minimum needed to update a single variable. Parts of the system talk in these instead of long histories. A receipt for one transaction, not the whole bank statement.","names":{"plain":["State-Delta (the unit)","Delta"],"technical":[{"name":"State-Delta (the unit)","source":"STATE-DELTA"}]},"cluster":"Data & Provenance","purpose":"capability","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA"],"status":"candidate","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"truth-ledger","function_statement":"An append-only local record of every change that passed validation; once written, entries cannot be edited or erased, so the audit trail is the system itself rather than a report bolted on later.","names":{"plain":["Truth Ledger","Grounding Ledger"],"technical":[{"name":"Truth Ledger","source":"STATE-DELTA"}]},"cluster":"Data & Provenance","purpose":"check","attributes":{"actor":"system","accountability":"org","cost_substrate":"non-resetting","reversibility":"irreversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA","NSA-CISA-AGENTIC"],"status":"candidate","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"representation-locality","function_statement":"The rule that components communicate only in compact state-deltas, never raw narrative or full histories, so contradictions cannot pile up in transit.","names":{"plain":["Representation Locality"],"technical":[{"name":"Representation Locality","source":"STATE-DELTA"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA"],"status":"candidate","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"context-reset","function_statement":"Flushing an agent's working memory after each committed change so the next task starts from verified current state, not accumulated reasoning. Drift cannot compound when memory resets against truth.","names":{"plain":["Context Reset After Commit","Context Reset"],"technical":[{"name":"Context Reset After Commit","source":"STATE-DELTA"}]},"cluster":"Agentic Mechanics","purpose":"reset","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA"],"status":"candidate","tier":"core","sysdyn":"Termination State","sysdyn_method":"derived"},{"id":"dynamic-agent-layer","function_statement":"The flexible, creative reasoning zone of a system, run in a temporary read-only sandbox with no direct authority over real systems. It proposes; it does not commit.","names":{"plain":["Dynamic Agent Layer"],"technical":[{"name":"Dynamic Agent Layer","source":"STATE-DELTA"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA"],"status":"candidate","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"structural-floor","function_statement":"The commitment that safety boundaries are enforced by fixed, non-AI components rather than by the model's own judgment. A fence at the cliff edge, not a sign asking you not to fall. Closely tied to the Harness.","names":{"plain":["Structural Floor"],"technical":[{"name":"Structural Floor","source":"STATE-DELTA"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA","NSA-CISA-AGENTIC"],"status":"candidate","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"deterministic-air-gapping","function_statement":"The requirement that safety-critical checks run entirely on local hardware with no network dependency, so they still work when the connection fails. The lock works whether or not the security company's servers are online.","names":{"plain":["Deterministic Air-Gapping"],"technical":[{"name":"Deterministic Air-Gapping","source":"STATE-DELTA"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA","NSA-CISA-AGENTIC"],"status":"candidate","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"firewall-of-cognition","function_statement":"Keeping the reasoning model in a read-only context that never directly touches infrastructure, with actions run in disposable sandboxes. The boundary does the work, so the reasoning engine is never trusted with the controls.","names":{"plain":["Firewall of Cognition"],"technical":[{"name":"Firewall of Cognition","source":"STATE-DELTA"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA","NSA-CISA-AGENTIC"],"status":"candidate","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"dual-pass-validation","function_statement":"Requiring every proposed change to clear two local checks before commit: does it fit the rules (ontology), and does it make sense in reality now (semantics). One check is not enough.","names":{"plain":["Dual-Pass Local Validation","Dual-Pass Validation"],"technical":[{"name":"Dual-Pass Local Validation","source":"STATE-DELTA"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA","LEAN-AGENT"],"status":"candidate","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"semantics-gate","function_statement":"The second validation pass: even if a change is technically allowed, does it cohere with current conditions and safety thresholds? Catches the move that is legal but catastrophic.","names":{"plain":["Semantics Gate"],"technical":[{"name":"Semantics Gate","source":"STATE-DELTA"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA"],"status":"candidate","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"refusal-logic","function_statement":"The deterministic, non-AI hard stop that blocks a change when validation fails, independent of the model's reasoning. The model does not get a vote; the breaker trips on its own, and the failure is logged.","names":{"plain":["Refusal Logic","Refusal Tray"],"technical":[{"name":"Refusal Logic","source":"STATE-DELTA"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA","NSA-CISA-AGENTIC"],"status":"candidate","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"mandate-gate","function_statement":"A check that routes high-consequence actions to a human instead of letting the agent commit them. Autonomy is scoped by how much is at stake.","names":{"plain":["Mandate Gate"],"technical":[{"name":"Mandate Gate","source":"STATE-DELTA"}]},"cluster":"Control & Safety","purpose":"center","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA","NSA-CISA-AGENTIC"],"status":"candidate","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"purpose-gate","function_statement":"A check that re-verifies an agent is still pursuing its authorized intent during a task, not only at startup, since static permissions do not survive dynamic workflows.","names":{"plain":["Purpose Gate"],"technical":[{"name":"Purpose Gate","source":"STATE-DELTA"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA","NSA-CISA-AGENTIC"],"status":"candidate","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"substrate-gate","function_statement":"A validation pass that checks a proposed change stays coupled to ground truth, real-world signal, rather than only being internally consistent.","names":{"plain":["Substrate Gate"],"technical":[{"name":"Substrate Gate","source":"STATE-DELTA"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA"],"status":"candidate","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"monotonicity-gate","function_statement":"A validation pass enforcing signed, append-only time: history moves forward only, so committed records cannot be silently rewritten.","names":{"plain":["Monotonicity Gate"],"technical":[{"name":"Monotonicity Gate","source":"STATE-DELTA"}]},"cluster":"Data & Provenance","purpose":"check","attributes":{"actor":"system","accountability":"org","cost_substrate":"non-resetting","reversibility":"irreversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA"],"status":"candidate","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"semantic-drift","function_statement":"When a system's internal picture slowly stops matching reality because it keeps absorbing data without checking it against current truth. Distinct from model drift: the world may be unchanged while the representation rots.","names":{"plain":["Semantic Drift"],"technical":[{"name":"Semantic Drift","source":"STATE-DELTA"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"latent"},"attributes_method":"curated","sources":["STATE-DELTA"],"status":"candidate","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"unbounded-accumulation","function_statement":"The failure of piling up inputs without ever deciding what is still true, until contradictions make a clean decision impossible. The closet you never clean out.","names":{"plain":["Unbounded Accumulation"],"technical":[{"name":"Unbounded Accumulation","source":"STATE-DELTA"}]},"cluster":"Failure Modes & Risks","purpose":"risk","attributes":{"actor":"ai","accountability":"org","cost_substrate":"non-resetting","reversibility":"reversible","observability":"latent"},"attributes_method":"curated","sources":["STATE-DELTA"],"status":"candidate","tier":"core","sysdyn":"Feedback Loop","sysdyn_method":"derived"},{"id":"elimination-earned-confidence","function_statement":"A proposed mechanism where a system earns the right to act in proportion to what it has ruled out, carried as an explicit margin, rather than from self-reported certainty. The right to commit scales with confidence and shrinks as consequence rises. Author's candidate contribution, pending review.","names":{"plain":["Elimination-Earned Confidence"],"technical":[{"name":"Elimination-Earned Confidence","source":"STATE-DELTA"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA"],"status":"candidate","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"metaphor-as-transmission","function_statement":"Treating a plain-language bridge image as a required layer of a control, not optional documentation, so engineers, operators, and affected people can all audit the same boundary. A control no outsider can read is one no outsider can challenge. Author's candidate contribution.","names":{"plain":["Metaphor as Transmission Layer"],"technical":[{"name":"Metaphor as Transmission Layer","source":"STATE-DELTA"}]},"cluster":"Oversight & Governance","purpose":"center","attributes":{"actor":"human","accountability":"human","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA"],"status":"candidate","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"impedance","function_statement":"Borrowed from electronics: how much a structure resists a signal passing through it. In governance, the deliberate friction a control layer adds so unverified change meets resistance and only a clean, in-tune signal (a grounded, coherent change) passes without distortion. Too little and drift flows straight through; too much and the system chokes: the tuning that keeps a structure's integrity under load.","names":{"plain":["Impedance"],"technical":[{"name":"Impedance","source":"STATE-DELTA"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"system","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["STATE-DELTA"],"status":"candidate","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"outcomes","function_statement":"The real-world results a system produces downstream, as opposed to its immediate outputs. Governance judges an AI by its outcomes, the impact on people and systems, not just what it prints on the screen.","names":{"plain":["Outcomes","Outcome"],"technical":[{"name":"Outcomes","source":"NIST-AI-RMF"}]},"cluster":"Oversight & Governance","purpose":"center","attributes":{"actor":"system","accountability":"org","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["NIST-AI-RMF","STANFORD-HAI"],"status":"candidate","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"yield","function_statement":"The share of a process's output that is actually usable or correct. Low yield means a lot of waste or error for each good result; used to judge how efficient or reliable a pipeline is.","names":{"plain":["Yield"],"technical":[{"name":"Yield","source":"FOUND"}]},"cluster":"Agentic Mechanics","purpose":"capability","attributes":{"actor":"system","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["FOUND","ARXIV"],"status":"candidate","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"depth","function_statement":"How many layers or reasoning steps a process runs through. Deeper models and chains capture more abstract structure but are harder to inspect and explain.","names":{"plain":["Depth"],"technical":[{"name":"Depth","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["FOUND","STANFORD-HAI"],"status":"candidate","tier":"core","sysdyn":"Input","sysdyn_method":"derived"},{"id":"shallow","function_statement":"Few layers or steps, handling surface patterns rather than deeper structure. Shallow methods are simpler and easier to inspect, but miss what depth would catch.","names":{"plain":["Shallow"],"technical":[{"name":"Shallow","source":"FOUND"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["FOUND"],"status":"candidate","tier":"technical","sysdyn":"Input","sysdyn_method":"derived"},{"id":"polarity","function_statement":"The positive-versus-negative orientation of a text or signal, as measured in sentiment analysis. When data or outputs cluster at opposing extremes they are polarized, which can skew what a model learns or produces.","names":{"plain":["Polarity","Polarized"],"technical":[{"name":"Polarity","source":"ARXIV"}]},"cluster":"Data & Provenance","purpose":"check","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"attributes_method":"curated","sources":["ARXIV","STANFORD-HAI"],"status":"candidate","tier":"core","sysdyn":"Constraint","sysdyn_method":"derived"},{"id":"deterministic","function_statement":"Produces the same output for the same input, every time, so a check built from it holds still while everything around it moves. The reason the floor is a rule-layer and not a model.","names":{"plain":["Deterministic"],"technical":[{"name":"Deterministic","source":"DELTA-ATLAS"}]},"cluster":"Control & Safety","purpose":"check","attributes":{"actor":"hybrid","accountability":"human","cost_substrate":"non-resetting","reversibility":"reversible","observability":"detectable"},"sources":[],"status":"candidate","attributes_method":"ai-draft (inherited from deterministic-constraint; cold-read pending)","tier":"core"},{"id":"probabilistic","function_statement":"Produces outputs drawn from likelihoods rather than certainties; powerful on open-ended problems, and the reason model output needs an external, non-probabilistic check before it commits.","names":{"plain":["Probabilistic"],"technical":[{"name":"Probabilistic","source":"DELTA-ATLAS"}]},"cluster":"Foundations","purpose":"capability","attributes":{"actor":"ai","accountability":"org","cost_substrate":"resetting","reversibility":"reversible","observability":"detectable"},"sources":[],"status":"candidate","attributes_method":"ai-draft (inherited from model; cold-read pending)","tier":"core"}],"relations":[{"source":"machine-learning","target":"artificial-intelligence","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"generative-ai","target":"artificial-intelligence","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"traditional-ai","target":"artificial-intelligence","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"traditional-ai","target":"machine-learning","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"deep-learning","target":"machine-learning","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"machine-learning","target":"training-data","type":"requires","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"machine-learning","target":"model","type":"enables","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"deep-learning","target":"neural-network","type":"requires","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"deep-learning","target":"foundation-model","type":"enables","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"training-data","target":"parameter-weights","type":"enables","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"fine-tuning","target":"parameter-weights","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"model","target":"parameter-weights","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"foundation-model","target":"model","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"inference","target":"model","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"large-language-model","target":"foundation-model","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"fine-tuning","target":"foundation-model","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"transformer","target":"foundation-model","type":"enables","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"large-language-model","target":"generative-ai","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"generative-ai","target":"hallucination","type":"causes","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"context-window","target":"token","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"large-language-model","target":"token","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"context-window","target":"memory","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"retrieval-augmented-generation","target":"context-window","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"vector-database","target":"embeddings","type":"requires","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"inference","target":"training-data","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"training-data","target":"data-poisoning","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"synthetic-data","target":"training-data","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"training-data","target":"bias","type":"causes","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"prompt","target":"prompt-engineering","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"prompt","target":"prompt-injection","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"prompt-engineering","target":"context-window","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"retrieval-augmented-generation","target":"grounding","type":"enables","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"retrieval-augmented-generation","target":"vector-database","type":"requires","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"retrieval-augmented-generation","target":"hallucination","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"reasoning-model","target":"agentic-ai","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"reasoning-model","target":"inference","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"multimodal-ai","target":"generative-ai","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"agent","target":"agentic-ai","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"autonomy","target":"agent","type":"enables","directed":true,"signals":{"function":false,"evidence":["NIST-AI-RMF"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"agentic-ai","target":"autonomy","type":"requires","directed":true,"signals":{"function":false,"evidence":["NIST-AI-RMF"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"multi-agent-system","target":"agentic-ai","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["NSA-CISA-AGENTIC"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"human-in-the-loop","target":"autonomy","type":"checks","directed":true,"signals":{"function":false,"evidence":["NSA-CISA-AGENTIC"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"autonomy","target":"excessive-agency","type":"enables","directed":true,"signals":{"function":false,"evidence":["NIST-AI-RMF"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"scoping","target":"autonomy","type":"checks","directed":true,"signals":{"function":false,"evidence":["OWASP-AGENTIC"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"tool-use-function-calling","target":"agent","type":"enables","directed":true,"signals":{"function":false,"evidence":["OWASP-AGENTIC"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"least-privilege","target":"tool-use-function-calling","type":"checks","directed":true,"signals":{"function":false,"evidence":["OWASP-AGENTIC"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"tool-use-function-calling","target":"tool-poisoning","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"orchestration","target":"workflow","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"orchestration","target":"planning","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"orchestration","target":"multi-agent-system","type":"enables","directed":true,"signals":{"function":false,"evidence":["NSA-CISA-AGENTIC"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"planning","target":"reasoning-model","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"goal-objective","target":"reward-hacking","type":"causes","directed":true,"signals":{"function":false,"evidence":["ARXIV"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"goal-objective","target":"alignment","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"workflow","target":"scoping","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"multi-agent-system","target":"cascading-failure","type":"causes","directed":true,"signals":{"function":false,"evidence":["NSA-CISA-AGENTIC"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"multi-agent-system","target":"sybil-resistance","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"model-context-protocol","target":"supply-chain","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"model-context-protocol","target":"least-privilege","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"retrieval-augmented-generation","target":"memory","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"memory","target":"data-poisoning","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"human-in-the-loop","target":"human-on-the-loop","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"human-in-the-loop","target":"accountability","type":"enables","directed":true,"signals":{"function":false,"evidence":["NSA-CISA-AGENTIC"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"human-on-the-loop","target":"autonomy","type":"handoff","directed":false,"signals":{"function":false,"evidence":["NSA-CISA-AGENTIC"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"human-on-the-loop","target":"automation-bias","type":"causes","directed":true,"signals":{"function":false,"evidence":["NIST-AI-RMF"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"least-privilege","target":"excessive-agency","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["OWASP-AGENTIC","NSA-CISA-AGENTIC"],"citation":[],"usage":0,"typed":true},"strength":"strong","status":"candidate","inferred":false},{"source":"scoping","target":"excessive-agency","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["OWASP-AGENTIC"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"ontology-policy-gate","target":"guardrail","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"validation-gate","target":"guardrail","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"guardrail","target":"alignment","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"ontology-policy-gate","target":"validation-gate","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"ontology-policy-gate","target":"grounding","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"validation-gate","target":"deterministic-constraint","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"deterministic-constraint","target":"guardrail","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"alignment","target":"ai-safety","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"alignment","target":"deceptive-alignment","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"red-teaming","target":"ai-safety","type":"enables","directed":true,"signals":{"function":false,"evidence":["OWASP-LLM"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"ai-safety","target":"responsible-ai","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"sandboxing","target":"defense-in-depth","type":"enables","directed":true,"signals":{"function":false,"evidence":["NIST-AI-RMF"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"sandboxing","target":"least-privilege","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"least-privilege","target":"zero-trust","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"zero-trust","target":"defense-in-depth","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"red-teaming","target":"prompt-injection","type":"detects","directed":true,"signals":{"function":false,"evidence":["OWASP-LLM"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"red-teaming","target":"jailbreak","type":"detects","directed":true,"signals":{"function":false,"evidence":["OWASP-LLM"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"evaluation-benchmark","target":"red-teaming","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"evaluation-benchmark","target":"model-drift","type":"detects","directed":true,"signals":{"function":false,"evidence":["NIST-AI-RMF"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"evaluation-benchmark","target":"validation-gate","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"kill-switch-circuit-breaker","target":"rate-limiting","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"kill-switch-circuit-breaker","target":"human-in-the-loop","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"rate-limiting","target":"defense-in-depth","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"scoping","target":"least-privilege","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"graceful-degradation","target":"defense-in-depth","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"graceful-degradation","target":"rate-limiting","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"ai-governance","target":"accountability","type":"enables","directed":true,"signals":{"function":false,"evidence":["NIST-AI-RMF"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"ai-governance","target":"compliance","type":"enables","directed":true,"signals":{"function":false,"evidence":["NIST-AI-RMF"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"nist-ai-risk-management-framework","target":"ai-governance","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["NIST-AI-RMF"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"nist-ai-risk-management-framework","target":"risk-assessment","type":"governs","directed":true,"signals":{"function":false,"evidence":["NIST-AI-RMF"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"nist-ai-risk-management-framework","target":"responsible-ai","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"responsible-ai","target":"ethical-ai","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"responsible-ai","target":"ai-governance","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"responsible-ai","target":"human-centered-ai","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"ethical-ai","target":"bias","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"accountability","target":"audit-trail","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"audit-trail","target":"provenance","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"model-card","target":"system-card","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"model-card","target":"transparency","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"system-card","target":"transparency","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"explainability","target":"transparency","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"explainability","target":"black-box","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["NIST-AI-RMF"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"compliance","target":"risk-assessment","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"red-teaming","target":"risk-assessment","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"ai-governance","target":"ai-lifecycle","type":"governs","directed":true,"signals":{"function":false,"evidence":["NIST-AI-RMF"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"ai-governance","target":"shadow-ai","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["NIST-AI-RMF"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"compliance","target":"shadow-ai","type":"mitigates","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"human-centered-ai","target":"human-in-the-loop","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"grounding","target":"ground-truth","type":"requires","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"validation-gate","target":"ground-truth","type":"requires","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"ground-truth","target":"hallucination","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"grounding","target":"hallucination","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"provenance","target":"attestation","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"provenance","target":"supply-chain","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"attestation","target":"sybil-resistance","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"attestation","target":"zero-trust","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"data-lineage","target":"provenance","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"data-lineage","target":"audit-trail","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"synthetic-data","target":"bias","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"vector-database","target":"memory","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"knowledge-graph","target":"ontology-policy-gate","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"knowledge-graph","target":"grounding","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"watermarking","target":"provenance","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"watermarking","target":"synthetic-data","type":"detects","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"sybil-resistance","target":"provenance","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"supply-chain","target":"tool-poisoning","type":"causes","directed":true,"signals":{"function":false,"evidence":["OWASP-AGENTIC"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"prompt-injection","target":"jailbreak","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"prompt-injection","target":"agent-goal-hijack","type":"causes","directed":true,"signals":{"function":false,"evidence":["OWASP-AGENTIC"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"guardrail","target":"jailbreak","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["OWASP-LLM"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"excessive-agency","target":"agent-goal-hijack","type":"enables","directed":true,"signals":{"function":false,"evidence":["OWASP-AGENTIC"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"agent-goal-hijack","target":"multi-agent-system","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"goal-misalignment-goal-misgeneralization","target":"alignment","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"goal-misalignment-goal-misgeneralization","target":"reward-hacking","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"goal-misalignment-goal-misgeneralization","target":"deceptive-alignment","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"reward-hacking","target":"alignment","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"red-teaming","target":"deceptive-alignment","type":"detects","directed":true,"signals":{"function":false,"evidence":["OWASP-LLM"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"model-drift","target":"training-data","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"defense-in-depth","target":"cascading-failure","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["NIST-AI-RMF"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"tool-poisoning","target":"data-poisoning","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"provenance","target":"data-poisoning","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["NSA-CISA-DATA"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"data-poisoning","target":"supply-chain","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"least-privilege","target":"sensitive-information-disclosure","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["OWASP-LLM"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"sensitive-information-disclosure","target":"privilege","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"shadow-ai","target":"sensitive-information-disclosure","type":"causes","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"weak","status":"candidate"},{"source":"least-privilege","target":"privilege","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["NSA-CISA-AGENTIC"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"privilege","target":"excessive-agency","type":"causes","directed":true,"signals":{"function":false,"evidence":["OWASP-AGENTIC"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"privilege","target":"scoping","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"automation-bias","target":"human-in-the-loop","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"overfitting","target":"training-data","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"overfitting","target":"model-drift","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate","needs_typing":true},{"source":"red-teaming","target":"emergent-behavior","type":"detects","directed":true,"signals":{"function":false,"evidence":["OWASP-LLM"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"harness","target":"guardrail","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["RESILIENCE-LEDGER"],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate"},{"source":"harness","target":"validation-gate","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["RESILIENCE-LEDGER"],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate"},{"source":"tether","target":"grounding","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["RESILIENCE-LEDGER"],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate"},{"source":"buffer","target":"graceful-degradation","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["RESILIENCE-LEDGER"],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate"},{"source":"buffer","target":"sandboxing","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["RESILIENCE-LEDGER"],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate"},{"source":"floor","target":"guardrail","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["RESILIENCE-LEDGER"],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate"},{"source":"return","target":"audit-trail","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["RESILIENCE-LEDGER"],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate"},{"source":"grace","target":"graceful-degradation","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["RESILIENCE-LEDGER"],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate"},{"source":"forgiveness","target":"audit-trail","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["RESILIENCE-LEDGER"],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate"},{"source":"rlhf","target":"human-in-the-loop","type":"handoff","directed":false,"signals":{"function":true,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate"},{"source":"rlhf","target":"alignment","type":"handoff","directed":false,"signals":{"function":true,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate"},{"source":"human-in-the-loop","target":"kill-switch-circuit-breaker","type":"handoff","directed":false,"signals":{"function":true,"evidence":[],"citation":[],"usage":0,"authored":true},"strength":"asserted","status":"candidate"},{"source":"guardrail","target":"prompt-injection","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["OWASP-LLM"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"sandboxing","target":"tool-poisoning","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["OWASP-AGENTIC"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"rate-limiting","target":"cascading-failure","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["NSA-CISA-AGENTIC"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"kill-switch-circuit-breaker","target":"cascading-failure","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["NSA-CISA-AGENTIC"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"human-in-the-loop","target":"agent","type":"checks","directed":true,"signals":{"function":false,"evidence":["NSA-CISA-AGENTIC"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"reinforcement-learning","target":"machine-learning","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"transformer","target":"large-language-model","type":"enables","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"agentic-ai","target":"large-language-model","type":"requires","directed":true,"signals":{"function":false,"evidence":["NSA-CISA-AGENTIC"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"drift","target":"model-drift","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["MLOPS"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"recurrent","target":"neural-network","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"lstm","target":"recurrent","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"gru","target":"recurrent","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"lora","target":"peft","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"qlora","target":"lora","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"qlora","target":"quantization","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"adapter","target":"peft","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"attention","target":"transformer","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"backpropagation","target":"deep-learning","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"chain-of-thought","target":"reasoning","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"tree-of-thoughts","target":"reasoning","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"react","target":"reasoning","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"monitoring","target":"drift","type":"detects","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"telemetry","target":"observability","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"pinecone","target":"vector-database","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["FOUND"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"chroma","target":"vector-database","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["FOUND"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"weaviate","target":"vector-database","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["FOUND"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"milvus","target":"vector-database","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["FOUND"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"lineage","target":"data-lineage","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["W3C"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"q-learning","target":"reinforcement-learning","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"actor-critic","target":"reinforcement-learning","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"ppo","target":"reinforcement-learning","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"dpo","target":"reinforcement-learning","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"dqn","target":"q-learning","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"pomdp","target":"mdp","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"retrieval","target":"vector-database","type":"requires","directed":true,"signals":{"function":false,"evidence":["FOUND"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"chunking","target":"retrieval","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"retrieval-augmented-generation","target":"retrieval","type":"requires","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"rlaif","target":"alignment","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"anonymization","target":"sensitive-information-disclosure","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["EU"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"redaction","target":"sensitive-information-disclosure","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["NIST"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"privacy","target":"sensitive-information-disclosure","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["EU"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"weightsbiases","target":"wandb","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["INDUSTRY"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"human-in-the-command","target":"hitc","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["NSA-CISA-AGENTIC"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"hotl","target":"human-on-the-loop","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["NSA-CISA-AGENTIC"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"kahneman-tversky","target":"kto","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["ARXIV"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"regression","target":"machine-learning","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"classification","target":"machine-learning","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"clustering","target":"unsupervised","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"forest","target":"ensemble","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"bagging","target":"ensemble","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"boosting","target":"ensemble","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"gguf","target":"quantization","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"gptq","target":"quantization","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"awq","target":"quantization","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"vllm","target":"inference","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"tensorrt","target":"inference","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"triton","target":"inference","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"fsdp","target":"parallelism","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"sharding","target":"parallelism","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"tokenization","target":"token","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"few-shot","target":"prompting","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"one-shot","target":"prompting","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"zero-shot","target":"prompting","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"kto","target":"alignment","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"preference","target":"alignment","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"cuda","target":"gpu","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"whisper","target":"speech-to-text","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"imbalance","target":"bias","type":"causes","directed":true,"signals":{"function":false,"evidence":["STANFORD-HAI"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"smote","target":"imbalance","type":"mitigates","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"balancing","target":"imbalance","type":"mitigates","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"regularization","target":"overfitting","type":"mitigates","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"dropout","target":"overfitting","type":"mitigates","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"redundancy","target":"fault","type":"mitigates","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"tolerance","target":"fault","type":"mitigates","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"staleness","target":"model-drift","type":"causes","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"retraining","target":"model-drift","type":"mitigates","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"auditing","target":"bias","type":"detects","directed":true,"signals":{"function":false,"evidence":["NIST-AI-RMF"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"explainability","target":"interpretability","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"human-in-the-command","target":"autonomy","type":"checks","directed":true,"signals":{"function":false,"evidence":["NSA-CISA-AGENTIC"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"hacking","target":"reward","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["ARXIV"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"tool","target":"agentic","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"plan","target":"agentic","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"decomposition","target":"plan","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"proximal","target":"ppo","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["ARXIV"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"tuning","target":"fine-tuning","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["FOUND"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"unmasking","target":"masking","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["ARXIV"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"loops","target":"human-in-the-loop","type":"alias_of","directed":false,"signals":{"function":true,"evidence":["NSA-CISA-AGENTIC"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"tanh","target":"activation","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"resilience-ledger","target":"sovereign-zero","type":"requires","directed":true,"signals":{"function":false,"evidence":["RESILIENCE-LEDGER"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"coherence-ledger","target":"sovereign-zero","type":"requires","directed":true,"signals":{"function":false,"evidence":["RESILIENCE-LEDGER"],"citation":[],"usage":0,"typed":true},"strength":"asserted","status":"candidate","inferred":false},{"source":"coherence-ledger","target":"resilience-ledger","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"sovereign-zero","target":"ground-truth","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"resilience-ledger","target":"ai-governance","type":"checks","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"harness","target":"autonomy","type":"checks","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"harness","target":"scaffold","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"harness","target":"agent","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"tether","target":"ground-truth","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"tether","target":"autonomy","type":"checks","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"tether","target":"human-in-the-loop","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"context-engineering","target":"grounding","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"context-engineering","target":"context-window","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"context-engineering","target":"retrieval-augmented-generation","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"solution-architecture","target":"orchestration","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"solution-architecture","target":"ai-lifecycle","type":"governs","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"systems-thinking","target":"ai-governance","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"systems-thinking","target":"resilience","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"scaffold","target":"agent","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"scaffold","target":"tool-use-function-calling","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"frontier-ai","target":"foundation-model","type":"instance_of","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"frontier-ai","target":"ai-governance","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"go-to-market","target":"responsible-ai","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"loops","target":"human-in-the-loop","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"loops","target":"human-on-the-loop","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"skills","target":"tool-use-function-calling","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"skills","target":"agent","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"query-concept","target":"retrieval","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"query-concept","target":"knowledge-graph","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"resilience","target":"ground-truth","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"resilience","target":"chain-of-custody","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"resilience","target":"resilience-ledger","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"resilience","target":"model-drift","type":"checks","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"chain-of-custody","target":"provenance","type":"requires","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"chain-of-custody","target":"audit-trail","type":"enables","directed":true,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"chain-of-custody","target":"data-lineage","type":"related","directed":false,"signals":{"function":false,"evidence":[],"citation":[],"usage":0,"typed":true},"strength":"weak","status":"candidate","inferred":false},{"source":"structural-floor","target":"state-delta-architecture","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"harness","target":"state-delta-architecture","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"tether","target":"state-delta-architecture","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"firewall-of-cognition","target":"state-delta-architecture","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"deterministic-air-gapping","target":"state-delta-architecture","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"dual-pass-validation","target":"state-delta-architecture","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"truth-ledger","target":"state-delta-architecture","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"refusal-logic","target":"state-delta-architecture","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"representation-locality","target":"state-delta-architecture","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"context-reset","target":"state-delta-architecture","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"dynamic-agent-layer","target":"state-delta-architecture","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"state-delta","target":"state-delta-architecture","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"system-state","target":"state-delta-architecture","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"friction-principle","target":"state-delta-architecture","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"ontology-policy-gate","target":"dual-pass-validation","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"ontology-policy-gate","target":"state-delta","type":"checks","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"semantics-gate","target":"dual-pass-validation","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"semantics-gate","target":"state-delta","type":"checks","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"substrate-gate","target":"dual-pass-validation","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"substrate-gate","target":"state-delta","type":"checks","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"monotonicity-gate","target":"dual-pass-validation","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"monotonicity-gate","target":"state-delta","type":"checks","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"mandate-gate","target":"dual-pass-validation","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"mandate-gate","target":"state-delta","type":"checks","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"purpose-gate","target":"dual-pass-validation","type":"instance_of","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"purpose-gate","target":"state-delta","type":"checks","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"ontology-policy-gate","target":"semantics-gate","type":"handoff","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"structural-floor","target":"harness","type":"related","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA","NSA-CISA-AGENTIC"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"deterministic-constraint","target":"refusal-logic","type":"related","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"deterministic-air-gapping","target":"zero-trust","type":"related","directed":true,"signals":{"function":false,"evidence":["NSA-CISA-AGENTIC"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"firewall-of-cognition","target":"zero-trust","type":"related","directed":true,"signals":{"function":false,"evidence":["NSA-CISA-AGENTIC"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"truth-ledger","target":"audit-trail","type":"related","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"truth-ledger","target":"provenance","type":"related","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"truth-ledger","target":"chain-of-custody","type":"related","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"mandate-gate","target":"human-in-the-loop","type":"requires","directed":true,"signals":{"function":false,"evidence":["NSA-CISA-AGENTIC"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"dual-pass-validation","target":"semantic-drift","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"context-reset","target":"unbounded-accumulation","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"context-reset","target":"semantic-drift","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"refusal-logic","target":"state-delta","type":"checks","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"unbounded-accumulation","target":"semantic-drift","type":"causes","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"semantic-drift","target":"model-drift","type":"related","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"semantic-drift","target":"drift","type":"related","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"state-delta","target":"system-state","type":"enables","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"system-state","target":"ground-truth","type":"related","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"structural-floor","target":"ground-truth","type":"related","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"friction-principle","target":"coherence","type":"enables","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"elimination-earned-confidence","target":"mandate-gate","type":"related","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"metaphor-as-transmission","target":"transparency","type":"enables","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"impedance","target":"friction-principle","type":"related","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"impedance","target":"harness","type":"related","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"impedance","target":"deterministic-constraint","type":"related","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"impedance","target":"semantic-drift","type":"mitigates","directed":true,"signals":{"function":false,"evidence":["STATE-DELTA"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"shallow","target":"depth","type":"related","directed":true,"signals":{"function":false,"evidence":["FOUND"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"depth","target":"explainability","type":"related","directed":true,"signals":{"function":false,"evidence":["FOUND"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"depth","target":"model","type":"related","directed":true,"signals":{"function":false,"evidence":["FOUND"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"density","target":"embeddings","type":"related","directed":true,"signals":{"function":false,"evidence":["FOUND"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"polarity","target":"embeddings","type":"related","directed":true,"signals":{"function":false,"evidence":["ARXIV"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"outcomes","target":"evaluation-benchmark","type":"related","directed":true,"signals":{"function":false,"evidence":["NIST-AI-RMF"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"outcomes","target":"accountability","type":"related","directed":true,"signals":{"function":false,"evidence":["NIST-AI-RMF"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"yield","target":"evaluation-benchmark","type":"related","directed":true,"signals":{"function":false,"evidence":["FOUND"],"citation":[],"usage":0,"authored":true,"typed":true},"strength":"asserted","status":"candidate"},{"source":"deterministic-constraint","type":"instance_of","target":"deterministic","strength":"asserted"},{"source":"refusal-logic","type":"instance_of","target":"deterministic","strength":"asserted"},{"source":"structural-floor","type":"requires","target":"deterministic","strength":"asserted"},{"source":"model","type":"instance_of","target":"probabilistic","strength":"asserted"},{"source":"probabilistic","type":"causes","target":"hallucination","strength":"asserted"},{"source":"probabilistic","type":"related","target":"semantic-drift","strength":"asserted"},{"source":"deterministic","type":"related","target":"probabilistic","strength":"asserted"}]};
const T=DATA.terms, R=DATA.relations;
const byId={}; T.forEach(t=>byId[t.id]=t);
function nm(t){return t.names.plain[0]||(t.names.technical[0]&&t.names.technical[0].name)||t.id;}
function esc(s){return (s==null?'':String(s)).replace(/[&<>"']/g,c=>({'&':'&amp;','<':'&lt;','>':'&gt;','"':'&quot;',"'":'&#39;'}[c]));}
const PURP={capability:"capability",absorb:"absorb",check:"check",reset:"reset",center:"center",risk:"risk"};
const DIR=new Set(["enables","requires","mitigates","detects","causes","checks","governs","instance_of"]);

// adjacency + load-bearing signal (typed, non-'related' relations; weighted by strength)
const deg={}; const out={}, inc={};
T.forEach(t=>{deg[t.id]=0; out[t.id]=[]; inc[t.id]=[];});
const SW={strong:3,asserted:2,weak:1};
R.forEach(r=>{ if(!byId[r.source]||!byId[r.target])return;
  out[r.source].push(r); inc[r.target].push(r);
  if(r.type!=='related'){ const w=SW[r.strength]||1; deg[r.source]+=w; deg[r.target]+=w; } });

// facet definitions: key -> accessor
const FACETS=[
 {key:'cluster',label:'Area',get:t=>t.cluster},
 {key:'purpose',label:'What it does',get:t=>PURP[t.purpose]||t.purpose},
 {key:'sysdyn',label:'System role',get:t=>t.sysdyn||'undefined'},
 {key:'tier',label:'Level (core or detailed)',get:t=>t.tier},
 {key:'actor',label:'Who acts',get:t=>t.attributes.actor},
 {key:'accountability',label:'Who is accountable',get:t=>t.attributes.accountability},
 {key:'cost_substrate',label:'Lasting cost (resets or sticks)',get:t=>t.attributes.cost_substrate},
 {key:'reversibility',label:'Can it be undone?',get:t=>t.attributes.reversibility},
 {key:'observability',label:'Can you see it happen?',get:t=>t.attributes.observability},
];
const sel={}; FACETS.forEach(f=>sel[f.key]=new Set());
let relPred={type:'',target:''};

// build value sets with counts
const VALS={};
FACETS.forEach(f=>{ const c={}; T.forEach(t=>{const v=f.get(t)||'undefined'; c[v]=(c[v]||0)+1;});
  VALS[f.key]=Object.keys(c).sort().map(v=>[v,c[v]]); });

function buildSide(){
  const s=document.getElementById('side'); let h='';
  FACETS.forEach(f=>{ h+='<div class="facet"><h4>'+esc(f.label)+'</h4>';
    VALS[f.key].forEach(vc=>{ h+='<span class="chip" data-f="'+f.key+'" data-v="'+esc(vc[0])+'">'+esc(vc[0])+'<span class="ct">'+vc[1]+'</span></span>'; });
    h+='</div>'; });
  // advanced relation predicate
  h+='<div class="adv"><h4>Find connections</h4><div class="advhelp">Show every term that has a chosen relationship to a chosen term.</div><label class="advlab" for="relType">Relationship</label><select id="relType"><option value="">any relationship</option>';
  [['mitigates','defends against'],['detects','detects'],['enables','enables'],['requires','needs'],['causes','can cause'],['checks','checks'],['governs','governs'],['instance_of','is a kind of'],['alias_of','is another name for'],['handoff','hands off to']].forEach(rt=>h+='<option value="'+rt[0]+'">'+rt[1]+' &hellip;</option>');
  h+='</select><label class="advlab" for="relTarget">Term</label><select id="relTarget"><option value="">choose a term</option>';
  T.slice().sort((a,b)=>nm(a).localeCompare(nm(b))).forEach(t=>h+='<option value="'+esc(t.id)+'">'+esc(nm(t))+'</option>');
  h+='</select><span class="clear" id="clearAll">clear all filters</span></div>';
  s.innerHTML=h;
  s.querySelectorAll('.chip').forEach(c=>c.onclick=()=>{
    const f=c.dataset.f,v=c.dataset.v; if(sel[f].has(v))sel[f].delete(v); else sel[f].add(v);
    c.classList.toggle('on'); focusId=null; run(); });
  document.getElementById('relType').onchange=e=>{relPred.type=e.target.value; focusId=null; run();};
  document.getElementById('relTarget').onchange=e=>{relPred.target=e.target.value; focusId=null; run();};
  document.getElementById('clearAll').onclick=()=>{ FACETS.forEach(f=>sel[f.key].clear()); relPred={type:'',target:''};
    document.querySelectorAll('.chip').forEach(c=>c.classList.remove('on'));
    document.getElementById('relType').value=''; document.getElementById('relTarget').value=''; focusId=null; run(); };
}

function matchFacets(t){
  for(const f of FACETS){ if(sel[f.key].size===0) continue;
    const v=f.get(t)||'undefined'; if(!sel[f.key].has(v)) return false; }
  return true;
}
function matchRel(t){
  if(!relPred.type||!relPred.target) return true;
  return out[t.id].some(r=>r.type===relPred.type && r.target===relPred.target);
}
function queryString(){
  const parts=[]; FACETS.forEach(f=>{ if(sel[f.key].size) parts.push(esc(f.label)+': '+[...sel[f.key]].join(' or ')); });
  if(relPred.type&&relPred.target) parts.push(relPred.type+' &rarr; '+esc(nm(byId[relPred.target])));
  return parts.length? parts.join('  and  ') : 'Showing all terms (nothing ticked yet)';
}

let focusId=null;
function anyFilter(){ return FACETS.some(f=>sel[f.key].size>0) || (relPred.type&&relPred.target); }
const TW={mitigates:'defends against',checks:'checks',detects:'detects',enables:'enables',requires:'needs',instance_of:'is a kind of',governs:'governs',causes:'can cause',handoff:'hands off to',alias_of:'same as',related:'relates to'};
// Incoming phrasing must name BOTH ends or the sentence dangles ("X same as" ...what?).
// Each phrase completes with "this term" so every relation line reads as a full statement.
const TWIN={mitigates:'defends against this term',checks:'checks this term',detects:'detects this term',enables:'enables this term',requires:'needs this term',governs:'governs this term',causes:'can cause this term',handoff:'hands off to this term',alias_of:'is another name for this term',related:'relates to this term'};
function jlink(id,label){ return '<span class="jump" data-id="'+id+'">'+esc(label)+'</span>'; }
function coupleLine(t){
  const outs=out[t.id].filter(r=>r.type!=='related').slice(0,8).map(r=>(TW[r.type]||r.type)+' '+jlink(r.target,nm(byId[r.target])));
  const insr=inc[t.id].filter(r=>r.type!=='related'&&r.type!=='instance_of').slice(0,6).map(r=>jlink(r.source,nm(byId[r.source]))+' '+(TWIN[r.type]||((TW[r.type]||r.type)+' this term')));
  return (outs.length?'<div class="rel"><b>couples with:</b> '+outs.join(', ')+'</div>':'')+
         (insr.length?'<div class="rel"><b>coupled from:</b> '+insr.join(', ')+'</div>':'');
}
function termCard(t){
  return '<div class="card"><h3>'+sdot(t.status)+jlink(t.id,nm(t))+'</h3>'+
    '<div class="fn">'+esc(t.function_statement||'(not yet written)')+'</div>'+
    '<span class="tag p">'+esc(PURP[t.purpose]||t.purpose)+'</span><span class="tag">'+esc(t.cluster)+'</span>'+
    coupleLine(t)+'</div>';
}
function bindJumps(scope){ scope.querySelectorAll('.jump').forEach(el=>el.onclick=()=>{ focusId=el.dataset.id; run(); var m=document.querySelector('main'); if(m)m.scrollTop=0; }); }
function run(){
  const rdiv=document.getElementById('results'), cnt=document.getElementById('count');
  if(focusId && byId[focusId]){
    const f=byId[focusId];
    const nb=[...new Set(out[focusId].concat(inc[focusId]).map(r=>r.source===focusId?r.target:r.source))].map(id=>byId[id]).filter(Boolean).sort((a,b)=>nm(a).localeCompare(nm(b)));
    cnt.innerHTML='Showing <b>'+esc(nm(f))+'</b> and what it couples with &middot; <span class="jclear" style="cursor:pointer;color:var(--accent)">show all terms</span>';
    rdiv.innerHTML=termCard(f)+(nb.length?'<div class="seclbl2">Connected terms</div>'+nb.map(termCard).join(''):'<div class="none">No connections recorded for this term yet.</div>');
    var jc=cnt.querySelector('.jclear'); if(jc)jc.onclick=()=>{ focusId=null; run(); };
    bindJumps(rdiv); return;
  }
  const res=T.filter(t=>matchFacets(t)&&matchRel(t)).sort((a,b)=>nm(a).localeCompare(nm(b)));
  cnt.innerHTML='<b>'+res.length+'</b> term'+(res.length===1?'':'s')+(anyFilter()?' match your filters':'')+' &middot; click any term to follow its connections';
  if(!res.length){ rdiv.innerHTML='<div class="none">Nothing matches. Untick something on the left.</div>'; return; }
  rdiv.innerHTML=res.map(termCard).join(''); bindJumps(rdiv);
}

// non-narrative preset queries
const PRESETS=[
 {label:'Hidden risks that never reset', set:{cost_substrate:['non-resetting'],observability:['silent'],purpose:['risk']}},
 {label:'Irreversible failures the AI causes', set:{actor:['ai'],reversibility:['irreversible'],purpose:['risk']}},
 {label:'Human-accountable checks', set:{accountability:['human'],purpose:['check']}},
 {label:'Costs that stick to people', set:{cost_substrate:['non-resetting'],tier:['core']}},
 {label:'Latent risks nobody owns', set:{observability:['latent'],accountability:['undefined'],purpose:['risk']}},
 {label:'Core controls (Control & Safety)', set:{cluster:['Control & Safety'],tier:['core']}},
 {label:'What mitigates Hallucination', rel:{type:'mitigates',targetName:'Hallucination'}},
];
function applyPreset(p){
  focusId=null; FACETS.forEach(f=>sel[f.key].clear()); relPred={type:'',target:''};
  if(p.set){ Object.keys(p.set).forEach(k=>p.set[k].forEach(v=>sel[k].add(v))); }
  if(p.rel){ const tgt=T.find(t=>nm(t)===p.rel.targetName); relPred={type:p.rel.type,target:tgt?tgt.id:''}; }
  // reflect into UI
  document.querySelectorAll('.chip').forEach(c=>c.classList.toggle('on', sel[c.dataset.f].has(c.dataset.v)));
  document.getElementById('relType').value=relPred.type||'';
  document.getElementById('relTarget').value=relPred.target||'';
  run();
}
function buildPresets(){
  const p=document.getElementById('presets');
  p.innerHTML='<span style="color:var(--dim);font-size:11.5px;align-self:center;margin-right:4px">Try:</span>'+
    PRESETS.map((x,i)=>'<span class="preset" data-i="'+i+'">'+esc(x.label)+'</span>').join('');
  p.querySelectorAll('.preset').forEach(el=>el.onclick=()=>applyPreset(PRESETS[+el.dataset.i]));
}
buildSide(); buildPresets(); run();
</script></body></html>