Home · How we work
How we work
This site is a public record of beliefs about learning, tested on purpose. The story on the home page is what has held so far. This page is the machinery behind it.
The three beats
- We write down a beliefOne plain sentence about learning that could turn out to be wrong.
- Every night we attack itWith real studies that could prove it wrong.
- We keep what survivesWhat fails gets narrowed or dropped. Nothing is scored; we only count the attacks.
Explore the record
- MapAreas and how beliefs connect
- All claimsEvery belief under test, with its attack count
- AreasBrowse by domain of learning
- TrailsShort paths through related claims
- SearchFind a claim, area or question
- Open questionsWhat the attacks have not answered
- Where it is weakestBeliefs with the most open challenges
- Ideas that failedExclusions and killed claims
- Evidence librarySources, notes and the method
Recent work
Newly added, still under test: 6, 14, 16, 27, 29 October, 5 and 12 November 2026. See all 26 →
Where it is weakest right now: C2 Answer-giving AI in practice — 4 open challenges, 7 attacks. Full list →