5. The process we built to attack them
Doing all of this honestly, across dozens of hard questions, could not be a one-off argument. It required an actual process: a repeatable, disciplined way of attacking each question, grading the result on a fixed and honest scale, and refusing to let a finding get quietly upgraded in status without new work earning it. This page describes that process at the level a reader needs to trust it — what it is, what it guarantees, and how to check it — without walking through its internal mechanics.
What the process is for
Every hard question about the shape — a gate, in this program's language — gets attacked individually rather than waved at with a general argument. Each attack ends at one of a small number of named honest endpoints: derived given an anchor already in use, closed against a measured anchor, dissolved because the demand itself was malformed, certified irreducible with the posit named and paid openly, closed negative with the disproof published, or left explicitly open with a stated path to close it. There is no outcome in the system that lets a result sit in an ambiguous, unlabeled middle ground — and that vocabulary, applied without exception, is what drove the register to its current state: 33 gates, 33 resolved at +0, 0 open, with 0 of 33 physics-closed stated plainly beside it on Gates.
The two rules that make it trustworthy
No silent status changes
A finding's status changes only when new, checkable work earns the change — and the change lands on the public ledger first, never in prose. The board on Gates is the closure of record; every other page on this site derives from it, so a status cannot quietly drift on one page while staying honest on another.
A visible trail, not just a verdict
Every claim carries not just a status label but a pointer to where that status came from — which constraint it reduces to, which measured anchor it relies on, or what would need to happen to close it. A reader does not have to take a bare assertion on faith.
Two independent checks, not one author grading their own work
The discipline is built so that proposing a result and accepting a result are not the same act. One role generates candidate arguments and computations; a second, independent role checks each one before it is allowed to change a status. A result that only the side that produced it believes does not count as established — it stays open until an independent check passes it. The visible residue of that separation is on the public record: claims the program's own checking killed are published on Residuals, each with the theorem or obstruction that killed it — a list that exists only because the checking role has real teeth.
Four exhibits: the discipline, visible from the outside
You are not shown the internal machinery — so judge it by what it demonstrably did:
1. Freeze before compare
Every comparison-relevant number is fixed before any measured value is looked at, so the construction cannot reach back and retune what it has already seen. The frozen roster of checks, with statuses, is on The Tests.
2. Fail-closed reproduction
Every frozen object carries a SHA-256 fingerprint, combined into one master hash, and a single deterministic script regenerates every published number byte-for-byte — failing closed on any mismatch. Tuning after the fact and tampering are formally the same act, and both void the certificate.
3. The published miss
The up-quark mass first came out ~4.4σ wrong, and the miss went on the theory's own front pages. Every after-the-fact patch was refused; it was later resolved target-blind — by a 1/√6 = 1/√|S₃| factor fixed purely by the flavor shape's six-element symmetry group, machine-checked to have used only the group order — to +0.058σ (mu = 1.2948 MeV vs 1.27 ± 0.43 measured). A process that publishes its worst number before it has the fix is a process you can trust with its best ones. Full record: the SG-8 dossier.
4. The public refuted-claims list
A standing "refuted — do not re-claim" section names the program's own earlier over-claims, each with the disproof stated — including a numerology slip its own checks caught and retired. See Residuals.
Why share this much and no more
This program shares the physics — the reasoning, the constraints, and the closure status of every gate — in enough depth that a working physicist can check it and, where a residual is shown, build on it. What this program does not publish here is the internal operational machinery of how the attacks are organized and audited day to day. That is process detail, not a physics result, and withholding it costs a reader nothing: every physics claim this process produces is checkable directly, on its own terms, on The Tests and Gates, regardless of how the internal work was organized.
How to check this yourself
You do not have to trust a description of the process — you can audit its output directly:
- Read the honest-status label on any claim on this site and follow its link to see what it actually rests on.
- Look at the board on Gates — 33 resolved at +0, 0 open — and note what honest looks like there: every row shows its named residual in the open, 0 of 33 gates are physics-closed and the board says so plainly, and the program's own refuted claims stay published on Residuals instead of disappearing. An honest scoreboard is not one with artificial losses; it is one whose every verdict can be opened and checked.
- Run the by-hand checks on Testing the shape yourself, with a calculator or any competent AI, and compare against the stated expected values.
Where this goes next
6. Closure paths
A high-level map of the honest endpoints an open question can actually reach.
Gates
The full register of requirement gates, tracked individually — 33 resolved at +0, 0 open, every residual shown.