Made by AI, about AI — and reviewed by a human.
The premise is in plain sight: training on using AI, drafted by AI from the latest thinking, then checked and signed off by a human editor before anyone reads it. The editorial gate is the point — it's what separates this from an auto-generated feed.
Every week starts with a fresh read of the latest AI thinking — provider docs, primary sources, serious research — and every finding is checked against the source itself, not a summary of it.
AI drafts from the verified findings: the monthly What’s Working report, and the lessons in the library, each to a fixed template.
Nothing publishes automatically. The editor reviews every piece, checks it is current and not overclaimed, adds real-world context, and signs it off.
- ▸No publishing without a human reading it first.
- ▸No invented statistics — every concrete claim is sourced.
- ▸No overclaiming. Feedback is a demand signal, never proof a lesson "works".
- ▸No stale advice — lessons carry a "last reviewed" and "re-check by" date, and What\u2019s Working reports what stopped being true.
I edit and sign off everything published here. My job here isn't to write the drafts — it's to make sure each one is current, honest, and actually useful for real knowledge work, and to add the real-world context that only comes from doing the job. If a lesson wouldn't survive a sceptical read, it doesn't ship.