Instrumenting Doorways
Every tool call is a doorway. We log the doorway, not the room, and the doorway is where the signal is.
6 min readThe working parts.
What we learn from building.
Rollout Checklist
Fleet rollouts fail at the weakest checkpoint. A rollout checklist from sixteen days in a production fleet.
6 min readDetection Latency
By the time an incident is triaged, the agent has finished its task. Detection has to move inside the loop.
6 min read
Interruption Over Reconstruction
Interruption beats reconstruction. Wiring detection into the loop so incidents stop before they compound.
6 min read
Fleet Checkpoints
A fleet is only as safe as its weakest checkpoint. Fleet checkpoints need fleet-grade verification.
6 min read
Threshold Ranking
Thresholds are ranked, not guessed. The engineering method behind doorway instrumentation budgets.
6 min read
Doorway Thresholds
Engineering notes from instrumenting the loop: the doorways worth logging, and the ones that are noise.
6 min read
Production Fleet Diary
A deployment diary: what we logged, what we missed, and the one behavior no evaluation predicted.
6 min read