Keep Me Honest: two stages before you trust an AI’s output

A two-stage checklist for any output about to inform a real decision. Stage one governs the work while it is produced: state material assumptions instead of guessing, flag gaps instead of inventing facts, never call work verified when it wasn’t. Stage two runs eleven checks on the finished result: right question, values that actually exist, every claimed component present, what got filtered out, why it looks clean, effect versus noise, whether the two things compared are comparable, what the measuring instrument cannot see, what validates the input at the root of the chain, whether the code does what its comment claims, and what it means when correcting an input destroys the result. It closes with a report of what was checked, what was found, and what could not be verified. A result can pass stage one and still fail stage two.