decision-memorylisted

# Persist Non-Obvious Decisions Across Sessions When you discover *why* something must be done a certain way — a decision, a trap, a preference — write it down at that moment as a small, dated, cross-linked note. A rationale that lives only in chat history is lost the instant the session ends, and the next session pays to re-derive it. ## When to use Reach for this the moment you learn something a future session would otherwise have to rediscover: - A non-obvious choice with a real reason ("we pinned X because Y breaks Z"). - A gotcha with an exact remediation (a flaky timing window, a silent deprecation warning, a model that loops on a certain prompt). - A stable user or project preference ("always do A before B"). - You are *resuming* work and can't remember where the last session left off. Red-flag thoughts — if you catch yourself thinking any of these, STOP and write the note: - "I'll remember this." (You won't. The session will end.) - "It's obvious why we did it this way." (It won't be obvious next week, to you or anyone.) - "I'll document it later when I clean up." (Later never comes; the rationale is freshest now.) - "The old note says X, so X." (Notes go stale — verify before you trust.) And the mirror-image red flags — these mean STOP and do **not** write (or delete what you wrote): - "I'll keep this in case it changes." (Speculation is not the trigger; a fact captured for a hypothetical future is noise today.) - "Better to have it than not." (Every note tax