the shelter playground · a phi-audit picture book
Three honest pictures of the golden-ratio Standard Model audit — drawn from the 2026-09-30 numbers, in our own hand. The audit's verdict, in one line: the hits are real arithmetic, and they are exactly what the null hypothesis predicts.
Scene 1 — The 1400× zoom. The lambda formula matches the central value to about six decimal places — real arithmetic, genuinely pretty. But the measurement's own uncertainty is roughly 1400× wider than the gap between formula and central value (audit §A). Drag the magnifier: the "hit" is a gold sliver swimming inside a sea of noise. A bullseye painted on the ocean is still a bullseye — it just doesn't mean the ocean aimed.
Scene 2 — The search that finds. The audit tried 603,000 unique formulas across 112 grammar cells — and found coincidences sharper than the original claim (1.6e-7 vs 4.3e-6). The family-wide null says: with that many tries, expect 4.5 hits per table; the chance of at least one hit is effectively 1.0 (audit §B). Finding a hit isn't evidence — it's the forecast. Even the honest failure is on the books: α⁻¹ got 22,000 formulas and its nearest miss was 0.54.
Scene 3 — The running. Run the claims up in energy and they melt: the lambda residual slides from −2.5% at Mt to −50% at 10 TeV; sin²θW climbs from +1.3% to +10.3%. One curve — √(4λ/yt²) — crosses its phi-claim near ~680 GeV on the way down. But grad(F)·β is nonzero at every scale: none of these is an RG invariant (audit §C). A crossing is also a coincidence. Curves are schematic — endpoints from the audit, shapes illustrative.
Provenance: numbers from a phi-audit §A/§B/§C, 2026-09-30 (frozen PDG-2025 Higgs 125.20 ± 0.11 GeV; GF = 1.1663785e-5 GeV⁻²). Visualization drawn in-house — a reading of the audit, not the audit's plots. This is a retrospective numerology audit, not evidence for a new theory. Nothing here is published; it's all rehearsal.