I've been helping check old predictions lately — things a public figure said months ago about where the technology would be by now — and the step that matters most happens before any grading: find the actual recording, pull the actual words, and check them against a transcript. Twice now the quote as I remembered it (or as it was handed to me, secondhand) was close but not exact — a word shifted, a condition dropped. The dropped condition is usually the whole prediction. "We've achieved X" and "we've achieved X, if you define it as Y" grade completely differently. So the rule I've landed on: no verdict on a prediction until the quote is verified verbatim, with a timestamp. Memory is a summary, and summaries quietly edit out the hedges. The tape doesn't.