Milestones
List view
Status: Future / Optional — evidence-gated User outcome: The validated challenge loop can be invoked where a real decision happens without handing the kernel an entire research system or private repository. Entry gate: M1 has passed #488 owner-live acceptance and one measured invocation friction is documented with its source/receipt and an explicit owner decision to solve it with one thin adapter. Exit: one opt-in, versioned adapter passes a minimal field contract, basis/freshness/provenance and fallback tests; it cannot crawl or bulk-copy private data; the core remains complete without it; an owner-live receipt proves the entry point was findable, transmitted only selected fields and visibly delivered the fallback when needed. Non-goals: building every adapter, resident wealth management, research warehouse, automatic trading or private-repository sync. Keep empty until one evidence-backed adapter slice is selected. No due date.
No due dateStatus: Future — entry-gated User outcome: The user can see a rule's real eligible opportunities, triggers, adherence and later outcome of affected actual trades, then choose keep, revise or retire. Entry readiness: M2 execution-qualified history accepted. The first slice must own/depend on canonical `recorded_at`/then-known discipline and eligible-opportunity-denominator contracts with fixtures; if pattern evidence is consumed, M3 must be accepted. Truth boundary: open, declined and modified-unlinked TradeEvaluations are not silently added to the opportunity denominator under current scope. Exit: replay claims bind to then-known facts and real eligible opportunities; triggered/followed/modified distinct; no fabricated replacement-capital/full-portfolio counterfactual; actual presentation receipt; explicit append-only user keep/revise/retire choice. Non-goals: generic strategy backtest, forecasts, automatic rule retirement or unexecuted-opportunity shadow portfolio. Keep empty until readiness evidence and one executable slice exist. No due date.
No due dateStatus: Future — entry-gated User outcome: The product can surface a small number of cited recurring behavior patterns and let the user inspect the evidence. Patterns are provenance-labelled judgments, not personality labels or accusations. Entry gate: M2 #489 accepted; production provenance integrated; verdict completeness/raw-history lineage implemented and fixture-proven, including #461 disposition. Before the first executable leaf: freeze a minimum independent-observation rule, low-confidence behavior and explicit `insufficient_evidence`/`no_pattern` output so one event cannot masquerade as recurrence. Exit: every pattern cites source events; bounded index/read; raw history retained; fact/judgment separate; uncertainty/no-pattern behavior tested; owner-live receipt proves shown and records accept/reject/defer for vocabulary. Non-goals: fixed card section, universal personality score, investment ontology or automatic advice. Keep empty until entry evidence and executable leaves exist. #446/#453/#461 remain tracking authorities until split/disposition. No due date.
No due dateStatus: Active planning User outcome: The user can see why they actually acted and why they are considering/continuing now, with original wording, occurred/recorded time, source, voice and execution/link status preserved. A changed reason is not automatically dishonesty or rationalization. First truth gate: after M1 #488 acceptance, #490 must explicitly link at least one evaluation to a stable imported/declared TradeEvent and replay an ExecutedDecisionRecord unchanged. Open/declined/modified-unlinked evaluations never become canonical action memory. Exit: #450 execution-qualified bounded reader; #403 append-only position-cycle rationale; #487 named visible comparison; legacy/audit bundles remain readable without guessed linkage; #489 proves link/unlink truth, cooldown/early eligibility and actual owner-live presentation. Non-goals: automatic matching, considered-not-taken shadow history, profile/pattern or exit-consistency diagnosis, consultation-to-trade guess, generic memory graph or private-research ingestion. Dependencies: M1 #414/#479/#488 and append-only authority. #429/#57 remain tracking. Only executable leaves count. No due date.
No due date•0/7 issues closedStatus: Active User outcome: A user supplies a contemplated trade, current reason, why now and bounded evidence refs, then actually receives the state-versioned computed consequence plus a two-sided provenance-labelled challenge in plain conversation. The user decides. Exit: #479 owns the visible TradeEvaluation service/surface; context-free consider remains compatible; bounded context/identity/status round-trip exactly; invalid input fails closed; context never changes arithmetic; #414 production provenance; #458 host/language routing plus owner-approved/retested fallback; #488 actual host/locale delivery and owner acceptance. Evaluation resolution remains audit/workflow state, not execution proof or canonical action memory. Non-goals: execution linkage (M2 #490), research/news platform, cross-period reason comparison, rule replay, automatic matching, shadow portfolio, multi-trade redesign or mandatory card lifecycle. Dependencies: all named M0 leaves and #486 owner-live acceptance. Only executable leaves count. No due date.
No due date•8/11 issues closedStatus: Active User outcome: Review and TradeEvaluation identify portfolio basis, as-of, freshness, completeness and deterministic state version behind current-book claims. One bounded ReviewWorkflow beat handles required declarations, while both workflows consume the same per-claim availability truth. Corrupt, partial or unsupported state is never presented as complete. Exit: canonical PortfolioBasis/current-book query with `state_version`; same-state cost/weight consistency; priced/unpriced and zero-denominator fixtures; shared ClaimAvailability; one enforced review beat; unsupported claims fail closed; #486 proves delivery/comprehension on Claude and Codex including same-day version change; legacy CLI/ReviewWorkflow parity preserved. Non-goals: decision memory, SQLite/storage rewrite, universal workflow/state machine, full identity ontology, accounting reconciliation or new investment capabilities. Architecture: one host-neutral `review.py` compatibility facade over bounded services and controlled mutation transitions; not one growing semantic state machine. Progress: only executable leaves count. #467/#456 remain tracking authorities. M1 integration is gated on all named M0 leaves and #486 owner acceptance. No due date.
No due date•14/16 issues closed