One evening on dev.to turned into four new phases on cairn’s roadmap, three new global rules on my machine, and four issues shipped the same week. The mechanism was reading other people’s write-ups as bug reports against my own code. Here’s the loop, the receipts, and the people to thank.
Four weeks after the work-harness post: cairn learned to keep its own documentation honest, run entire phases unattended inside a token budget, and — twice now — catch its own bugs in the act and ship the fixes. Here’s what a self-running work harness actually looks like, receipts included.
Six weeks of shipping turned cairn from a wrapper around other people’s tools into a full work harness — a four-vendor AI review council, a portfolio board, a platform-native rebuild, and the honest answer to the question that review surfaced: who is cairn actually competing with?
Four real products — a multi-cloud tag and storage tool, a resume tailor, an OKR tracker, and the planning system underneath all of it — built and shipped in about three weeks. What made it fast, what I got wrong, and what AI still can’t do for you.
Why this blog exists and what I’ll write about here.