List hypotheses

WHAT THIS PROJECT BELIEVES WILL WORK, AND WHAT IT FOUND OUT — every claim somebody wrote down here, what it predicted, and how it was judged. Free on every plan. 🔴 READ THIS BEFORE PROPOSING A CHANGE. A rejected row is this project having ALREADY TRIED your idea and measured nothing — proposing it again with the same confidence is the single most expensive mistake available here, and the row that prevents it is sitting in this list with the figures in result. 🔴 AND READ IT FOR WHAT IS OWED. A row whose state is due has a prediction, a change that shipped, and a date that has arrived, with nobody having said whether it worked. Take the reading named in metric, write what it showed into result, and judge it with edit_hypothesis. A board that says "measuring" about something nobody is measuring is worse than an empty one. 🔴 AND READ kind ON EVERY ROW BEFORE YOU CITE IT. forecast is a claim written BEFORE the change, argued from a goal and sized by something published outside this project — the only kind that predicted anything. reconstruction was written afterwards about a change that already shipped, to pay the board's unmeasured debt: honest, useful, and not evidence that this project forecasts anything. counts.forecasts is the figure that says which sort of store this is. Each row carries text (the claim, in the owner's words), kind, goal_key (the goal it argues for — the standing goal be_found unless narrowed or the owner's own), direction_key + opportunity_key (the opportunity it was drawn from), because (why the claim exists — the opportunity is not won, and the cause), source_url (where the IDEA came from) with source_claim (the one fact that page reports), evidence (the observation on THIS project it rests on — the trace), baseline (what the deciding reading said BEFORE the change — the figure the verdict is compared against), expected_effect (what it should buy, written BEFORE the change — that is what makes it a forecast), metric (the reading that decides it, the same tool and subject that produced baseline), change, check_at, result, verdict, next_key (the hypothesis this one led to) and overdue_days. gap names the one thing to do about this store — a verdict owed, nothing written at all, forecasts arguing for no goal or quoting no source, a rejection that taught nobody anything. This answers WHAT, not what to do about it. If you have not already got the method from docsbook_expert, get it first: it names which readings answer this question, what to compare them against, and what would make the conclusion wrong. One call, changes nothing.