Every test this instrument has run against itself. Question, statistic, threshold, and confound controls fixed in writing before the data. Verdicts published whether they flattered us or not.
Most systems in this space are built so they can never be wrong. Scalar Flower is built so it can be. This page is the proof — kept current, in public, permanently.
This page renders directly from ledger.json, the same file the record keeps. It cannot drift from the corpus, because it is the corpus.
Pre-registered means committed before computed. The question, the test statistic, the pass threshold, and the controls were written down before any data ran. Nothing on this page was decided after seeing the numbers.
The layers. Every entry names the layer it tested: REAL — exact mathematics and astronomy, reproducible by anyone; ARCH — the chosen architecture: planets treated as wave sources, an axiom we selected, not established physics (Test 15, below, is why we say that in every footer on this site); INTERP — the symbolic craft of readings, offered for reflection, never as prediction. Largely untestable by design — and wherever a test could reach it, we ran the test.
The verdicts. SURVIVED / REAL — the pre-registered threshold was met and held under every committed control. NULL — the effect was absent, or vanished under a control we bound ourselves to in advance. Nulls are kept, printed, and put to work: each one permanently narrows what a reading is allowed to say. FALSIFIED — tested and found false; the claim is dead. UNRESOLVED — the test could not decide, usually because a positive control failed. Not a survival: it obligates a better-powered successor. RETIRED — a claim we withdrew ourselves, because a later test superseded it or because we caught ourselves overstating it. SPLIT — part survived a fair control, part did not; both halves stated.
Reconciling the counts. The launch letter says "eleven tests pre-registered… four came out the other side alive," and this page holds more entries. Both are true: the eleven were the launch-era empirical series asking whether sky correlates with life, of which four survived; the entries below include those plus the structural results, mathematical characterizations, the census, and the retractions. The ledger records all of it, including the retractions. Especially the retractions.
| № | Test | Question asked | Layer | Verdict | Finding |
|---|
Each survivor is stated with its honest scope, because a survivor oversold is a null waiting to happen.
Every null on this page is converted into a permanent constraint on what this instrument may say. Current doctrine, forced by verdicts:
A system that publishes its nulls but doesn't obey them is theater. This section is where obedience is recorded.
Pre-registrations are posted on this page before any data runs. Currently scheduled:
Birth data: the open Gauquelin archive, 21,165 timed records. Ephemeris: Swiss Ephemeris. Statistics: seeded, cached, chart-independent null distributions. Verdicts live in ledger.json; this page renders from it. The full method, equations, and null model are on The Method and The Science. The complete argument — the frame-invariance theorems, the pre-registered null corpus, and the caught-artifact record this ledger tabulates — is set out in the Open-Core White Paper.