- 2026-10-05Sentinel Dev Diary: Auditing the WatchdogBuild step four was audited against the watchdog document. One sentence had two halves and a test that asserted the half that worked, three exit criteria passed without a running core anywhere near them, and a stale point's checkpoint never moved until a deployment showed it.13 min
- 2026-09-28Sentinel Dev Diary: Checks & BalancesFive processes act as checks and balances on the Sentinel build. Code and documents will drift apart on any long project; the aim is to notice quickly and correct it where the next person will see. Here is what each process watches, what keeps it honest, and where it stops.13 min
- 2026-09-21Sentinel Dev Diary: Where the Check LivesThe driver contract says a module emits and says a host reads, and only one of those was built. Three conformance checks placed where one of them can go, a register sentence that was false for three pull requests, and fifty-seven tests that cannot check where a test lives.12 min
- 2026-09-14Sentinel Dev Diary: What the Specification Left UnsaidBuild step one was called complete, then audited against the document that owns it. Six named rejections had become twenty-eight, two of them rules the specification never states, and one guard existed as a function with a fixture and no caller.9 min
- 2026-09-08Sentinel Dev Diary: Four Registers and a Drift CheckThe specification was finished before the build started, and then the build began producing statements it had no home for. Four numbered registers, one rule about where a sentence is allowed to live, and the distinction that stops a specification gap being closed by the commit that routed around it.13 min
- 2026-08-14Notes to Self: The Interview Between an Issue and a SpecI write issues as three-sentence notes to myself. That worked while I was the one writing the code, because I filled the gaps from memory without noticing. An agent can read every file in the repo and none of my recall — so the missing step is an interview, and its output is a spec.10 min
- 2026-08-07Worktrees: The Missing Primitive for Multi-Agent CodingGit worktrees are usually sold as a convenience for juggling two branches. Run several coding agents against the same repos and they become something else — a concurrency primitive, where the branch is the lock and the only hard part is cleaning up.9 min
- 2026-08-06The Advisor Tool: The Question Code Review Can't AskTests and code review ask whether a change is right. Almost nothing asks whether it's complete — whether the request, as written, described the whole job. That third question is what a stronger model reviewing the transcript is unusually good at, and the terminal shows you almost none of it.7 min