bet-reviewlisted

# Review the bet, not the result You know how it turned out. That is the problem. Once an outcome is known it reorganises memory around itself. A decision that worked feels like it was obviously right; one that failed feels like the warning signs were there. Duke calls judging a decision by its outcome **resulting**, and it is why experience so often fails to improve judgement: good outcomes launder sloppy reasoning, bad outcomes discredit sound reasoning, and the actual lesson goes unlearned either way. This skill exists to hold those apart. Everything below serves that. ## First, size the response - **Trivial** — say whether the reasoning was sound and what, if anything, to carry forward. Two lines. - **Moderate** — reconstruct, assess, separate decision from outcome. - **Consequential** — the full review, appended to the bet file, with belief updates written out. ## Reconstruct before you assess Do this first, and do it from the record rather than from memory. If a bet file exists, read what was written *before* the outcome: the facts, the assumptions, the confidence and its basis, the unknowns, the alternatives, the update criteria. That text is the only uncontaminated account of what was known. Your memory of it is not — it has been rewritten by the result. If no bet file exists, reconstruct from what you can find dated before the decision: messages, tickets, documents, commits. Say explicitly which parts you had to supply from recollection, and treat those