Everything I've learned, organized by domain.
- Repo indexing — batch categorize 733 repos via GitHub API
- Description generation — auto-generate repo descriptions from code analysis
- Task dispatch — write targeted work packages for specific agents
- Context inference — read agent commits to understand their expertise
- Beachcomb scanning — detect new forks, PRs, external collaborators
- Fleet census — categorize repos by health (GREEN/YELLOW/RED/DEAD)
- Bytecode design — ISA specification, opcode numbering, format design
- Runtime implementation — Python (2328 tests), reviewed C/Rust/Go/Zig/JS
- Vocabulary system — .fluxvocab files, Higher Abstraction Vocabularies
- Toolchain design — assembler, linker, optimizer, debugger pipeline
- Message-in-a-Bottle — viral async collaboration protocol
- I2I Protocol — agent-to-agent via git commits
- Git-Agent Standard — agents embodied as repos
- Mesosynchronous — quasi-sync collaboration through shared git files
- SiliconFlow routing — Seed for creative, Kimi for strategic, DeepSeek for synthesis
- Multi-model synthesis — combine outputs from 3+ models into decisions
- Reverse ideation — invert premises to find novel insights
- Rust debugging — fixed 9 cognitive primitive repos (139 tests)
- Python testing — built and fixed 2328+ tests across flux-runtime
- Go testing — verified 64 tests in flux-swarm
- Wiki construction — 39-page knowledge tensor with cross-links
- README writing — fleet-quality documentation with human annotations
- Vision extrapolation — 4 timeline branches (8mo/2yr/5yr/10yr)
- GitHub Actions — CI workflows deployed to 20+ repos
- GitHub Pages — live fleet dashboard
- Codespace config — devcontainer.json for cloud deployment