CultivateOSField notes from cultivation.
Slow, considered essays on the methodology, the modules, and the counter-position. No release notes here; those live on the changelog.
Refuted evidence is the most expensive evidence to delete
When a team takes a refuted hypothesis off the wiki, it doesn't disappear — it goes into people's heads, where it can't be cited. This is the case for keeping the strike-through visible.
Why Bushel ships nine artefacts, not three
Most launch toolkits stop at the announcement and the email. The cost of a launch drifting between audiences is paid in support tickets six months later.
The brief is the bottleneck
Engineering is fast. Research is fast. The artefact in the middle — the brief — takes weeks and gets stale before it ships. Cultivate OS is the deletion of the brief as a separate artefact.
Three states, one trail
Validated. Hypothesis. Refuted. A short essay on why the three-state evidence model is doing more work in Winnow than its size suggests.
AI as a farm implement
The counter-position. Velocity is not the metric. The metric is whether the thing you shipped came from real evidence — and whether anyone can find that evidence six months from now.