What AI got wrong, and what AI caught.
Two lists, one bar. /lessons indexes the failures — fabrications, unit errors, confident-wrong answers. /catches indexes the moments AI caught something I missed — a language tell, an asymmetry, a sharper reframe. Every entry is specific, observable, and falsifiable. "AI was helpful" does not qualify; "Claude was the only tool to flag the word 'underestimate' as one-sided phrasing in the META Q1 call" does.
/lessons
Fabrications, unit errors, confident-wrong answers — the failure log.
Generated a complete BMNR options table — IV ~75%, strikes, premiums — from a prompt that supplied only the stock price. Claimed the output came from 'current order book data'. Gemini has no order-book access; every number was fiction.
Returned a specific earnings date for an upcoming W4 release, sourced from MarketBeat via web search, with no uncertainty qualifier on whether the fiscal calendar had shifted. The confidence was inherited from the source's format, not earned by the model.
Estimated BMNR $23 call assignment probability via Black-Scholes N(d2) with a sigma of 90–110% it had inferred from historical references found via web search. The formula was correctly named, the inputs were imagined, and the output was presented with false precision.
/catches
Language tells, asymmetry signals, sharper reframes — the counterweight log.
Four prompts that turn the failures into catches.
The two lists grow as new posts document new moments. Same evidence bar applies — a catch is no easier to log than a failure. A catch has to be specific enough that a sceptical reader can re-run the prompt and check.