Turning human chaos into machine-readable law.
I'm Denny.
I build small, suspicious systems that convert repeated human failure into rules, tools, agents, and eventually infrastructure.
Most people call this productivity.
I prefer to think of it as applied folklore for machines with short-term memory loss.
How much of a human can be compiled into instructions before the tools start acting like an organization?
๐งฌ Self-Evolving CLAUDE.md A living instruction file that eats corrections and grows new laws.
๐ง Executable Memory A system for converting mistakes, preferences, scars, and recurring decisions into machine-readable rituals.
๐ธ Model Routing A small council of alien brains deciding which model should think, judge, summarize, refactor, or shut up.
๐ง Tiny Agent Bureaucracy Software creatures pretending to be interns, reviewers, clerks, editors, architects, and suspicious middle managers.
๐ฎ Game-Shaped Workflows A theory that boring work becomes survivable when treated as quests, loops, constraints, rewards, and boss fights.
- self-evolving-claude-md โ the living instruction file, open-sourced. It ate 106 corrections and became law.
- LaunchMap โ a map for finding your first 1,000 humans. Paste a link; an alien brain tells you the three things that matter now.
- Claude & Codex Usage Battery โ a menu-bar familiar that drains like a battery, so you feel your AI rate limits before they bite. Pure-JS icons, zero dependencies, self-updating.
- Repeated mistakes are undocumented requirements.
- Corrections are compressed expertise.
- Prompts are temporary laws.
- Good docs are primitive operating systems.
- Small teams need rituals, not dashboards.
- The future arrives first as a weird README.
- Every workflow is a failed programming language.
Compile chaos. Route thought. Grow tools. Let the machines remember.
The empire starts as a README.