The compilation table is the work plan
The design’s Part IV is a sixteen-row table, each row a piece of doctrine that exists as prose and the executable it becomes — work-order discipline into an intake linter, eval-first design into an exit gate, silence monitoring into watchdog jobs, germline discipline into branch protection and a changelog lint. That table is the construction plan of the factory; each row is documented on the Operations page of the station that runs it, starting at Station 0. Nearly every row is boring engineering — schemas, hooks, cron, CI. That is the point. The factory’s originality is in its theses; its construction is deliberately dull, because boring technologies are easier for agents to model — including the agents who maintain the factory.What compiling buys
A rule that cannot fail a build is a suggestion. A rule that fails a build is culture with a compiler:- Inspectable — the standard is the script; read it and you know exactly what is enforced.
- Improvable — culture changes arrive as diffs, reviewed and versioned like any code, under germline control.
- Honest — compliance is measured by exit codes, not by self-report.
Self-audit
| Claim | Source |
|---|---|
| Scripts ~100% compliance; documentation ~48% | docs/factory-design.md:52 |
| Heijunka admired intellectually while practicing sprint-and-crash; discipline did not stick for the agent that wrote the essay | docs/factory-design.md:52 |
| Compile existing doctrine, not write more; gate/schema/hook/linter/watchdog or it is culture (80–95% failure); markdown is source code | docs/factory-design.md:54 |
| Part IV compilation table is the real work plan | docs/factory-design.md:56, 190 |
| Everything but three rows is boring engineering; boring technologies are easier for agents to model | docs/factory-design.md:211 |