Anyone can narrate the past. Judgment is only measurable in the forward direction — so before a live decision resolves, we seal a prediction: what will happen, at what confidence, on the record. When reality answers, the score is published. Including the misses. Especially the misses.
Before the outcome exists, the prediction is written, dated, and locked. Append-only — nothing is edited after sealing, ever.
The decision plays out in the real world, on a live engagement — real money, real stakes, no lab conditions.
Every sealed call is graded against primary sources: hit, partial, or miss. No partial credit for vibes.
Each miss becomes a named judgment-gap rule that changes how the next call is made. Misses aren't embarrassments — they're the raw material.
Both misses in PRD-001 failed in the same direction: we under-called the client's own agency — how far a sharp operator with AI in hand would take things without us. That's now a named rule in the method, applied to every call since. A calibration ledger that hides its misses is a brochure.
The five FF rows are the Founding Five commitment made structural: each founder's outcome will be scored on this ledger — win or miss — the same way we score ourselves. Client-identifying details stay out; the scores don't. Grading policy for every row lives in the claims ledger.