Eclipse 2026 Benchmark Results

The sealed benchmark — April 8, 2024, verdict CALORIE

Stage 1 of the arc: the substrate ingested the 2024 total-eclipse archive raw and its invariant verified the measured ionospheric response against computed shadow geometry — at the default pre-registered constants, with zero calibration. Sealed to V234 as ECLIPSE_C4_PROJECTION_INVARIANT | CALORIE on 2026-07-16.

The verdict (generated live from the corpus)

The sealed 2024 verdict — all 28 pairs (click to expand — generated live from the ledger)

Pair	Kind	Verdict	Projected (UT)	Depletion	Obscuration
2024-04-08-carbondale-eclipse	positive	PROJECTED ✓	18:46:29 UT	274457 ppm (27.4%)	800534 ppm
2024-04-08-cleveland-eclipse	positive	PROJECTED ✓	19:01:31 UT	353317 ppm (35.3%)	802627 ppm
2024-04-08-dallas-eclipse	positive	PROJECTED ✓	18:45:00 UT	308350 ppm (30.8%)	997624 ppm
2024-04-08-indianapolis-eclipse	positive	PROJECTED ✓	18:53:22 UT	286986 ppm (28.7%)	800050 ppm
2024-04-08-mazatlan-eclipse	positive	PROJECTED ✓	18:05:00 UT	323319 ppm (32.3%)	970682 ppm
2024-04-08-millstone-eclipse	positive	PROJECTED ✓	19:17:05 UT	419091 ppm (41.9%)	801464 ppm
2024-04-08-norwich-eclipse	positive	PROJECTED ✓	19:17:06 UT	426439 ppm (42.6%)	801484 ppm
neg-2024-04-08-carbondale-ctrl2024-03-12	control-negative	dead cat ✓	—	—	—
neg-2024-04-08-carbondale-ctrl2024-04-01	control-negative	dead cat ✓	—	—	—
neg-2024-04-08-carbondale-ctrl2024-04-15	control-negative	dead cat ✓	—	—	—
neg-2024-04-08-cleveland-ctrl2024-03-12	control-negative	dead cat ✓	—	—	—
neg-2024-04-08-cleveland-ctrl2024-04-01	control-negative	dead cat ✓	—	—	—
neg-2024-04-08-cleveland-ctrl2024-04-15	control-negative	dead cat ✓	—	—	—
neg-2024-04-08-dallas-ctrl2024-03-12	control-negative	dead cat ✓	—	—	—
neg-2024-04-08-dallas-ctrl2024-04-01	control-negative	dead cat ✓	—	—	—
neg-2024-04-08-dallas-ctrl2024-04-15	control-negative	dead cat ✓	—	—	—
neg-2024-04-08-indianapolis-ctrl2024-03-12	control-negative	dead cat ✓	—	—	—
neg-2024-04-08-indianapolis-ctrl2024-04-01	control-negative	dead cat ✓	—	—	—
neg-2024-04-08-indianapolis-ctrl2024-04-15	control-negative	dead cat ✓	—	—	—
neg-2024-04-08-mazatlan-ctrl2024-03-12	control-negative	dead cat ✓	—	—	—
neg-2024-04-08-mazatlan-ctrl2024-04-01	control-negative	dead cat ✓	—	—	—
neg-2024-04-08-mazatlan-ctrl2024-04-15	control-negative	dead cat ✓	—	—	—
neg-2024-04-08-millstone-ctrl2024-03-12	control-negative	dead cat ✓	—	—	—
neg-2024-04-08-millstone-ctrl2024-04-01	control-negative	dead cat ✓	—	—	—
neg-2024-04-08-millstone-ctrl2024-04-15	control-negative	dead cat ✓	—	—	—
neg-2024-04-08-norwich-ctrl2024-03-12	control-negative	dead cat ✓	—	—	—
neg-2024-04-08-norwich-ctrl2024-04-01	control-negative	dead cat ✓	—	—	—
neg-2024-04-08-norwich-ctrl2024-04-15	control-negative	dead cat ✓	—	—	—

The sealed 2024 benchmark scope (eclipse day + quiet controls). Generated live from the corpus through the runner.

What the table shows, in words:

• Seven positives, seven projections. Every 2024 station-day — five totality-path stations (Mazatlán, Dallas, Carbondale, Indianapolis, Cleveland) plus the two New-England deep partials (Millstone Hill 94.0%, Norwich CT 90.6%) — projected its whole child inside the ≥80% obscuration gate.
• The projected instants track the Moon's shadow west→east across the continent, each within minutes of local maximum — consistent with the published bottomside response lag.
• Depletions of 27–43% of quiet baseline — inside the published 20–60% band for this eclipse — with the deep-partial New-England stations at the top of the range, matching the regionally-strong 2024 northeast response in the literature.
• All 21 control negatives sealed dead cat: the same geometry against the same stations' quiet control days (2024-04-01, 2024-04-15, 2024-03-12; every one max Kp ≤ 3.7), leave-one-out lower-median baselines so no control day is ever judged against itself. Each negative died in ~1,400–1,900 wholesale patch voids with zero leaves sealed — the unimodal prune doing O(1) work.
• Zero false positives. Zero calibration. Gate 800,000 ppm and coupling 3/10 were the defaults written before the first crucible run; the physics closed on first presentation.

Trust chain

1. Geometry earns trust first: the Besselian calculator reproduces independently published circumstances — Dallas totality contacts ±3 s, Norwich deep partial ±3 s with obscuration 905,593 ppm vs the published ~90.6% — before anything persists (the fixture gate refuses otherwise).
2. Raw data with provenance: one Madrigal world-file scan per day, SHA-256 over the raw bytes, sliced per station locally; all quiet days verified Kp ≤ 4 against GFZ live.
3. Exact integers from ingest to seal: milliTECU / ppm / whole seconds; cross-multiplied Int64 law; no floats anywhere on the sealed path.
4. Determinism: running the setter twice yields the identical contentFingerprint; a fresh process re-verifies the Form from disk with no network and no DB.

The seals

Form	Kind	Class	Closure	Terminal	Fingerprint	Stamped
2024 benchmark	projection	shadow-projection	7/7	CALORIE	89afd825011f060e	2026-07-16T16:01:47Z
2026 prediction	prediction-2026-08-12	shadow-projection	PREREGISTERED	CALORIE	c3e6841a206b8058	2026-07-16T16:02:00Z
2026 prediction extensions	prediction-extensions-2026-08-12	shadow-projection	PREREGISTERED	CALORIE	b106d2926b1bf2cf	2026-07-16T18:54:39Z
historical universality	historical-universality	shadow-projection	16/20	CURE	cdf43e80f06f10a3	2026-08-18T16:42:21Z

Loaded and fingerprint-checked from ~/.gaiaftcl/franklin/invariants/ at render time.

Reproduction runbook (from a cold machine)

# 0. build, then geometry with the fixture gate:
swift build -c release --product GaiaFTCLCLI
gaiaftcl eclipse-geometry --eclipse 2024-04-08 --all-stations       # fixtures must print GREEN

# 1. raw ingest — eclipse day + three controls (one world-scan each), plus Kp:
for d in 2024-04-08 2024-04-01 2024-04-15 2024-03-12; do
  gaiaftcl helio-ingest kp --day $d
  gaiaftcl helio-ingest madrigal-tec --day $d \
    --stations mazatlan,dallas,carbondale,indianapolis,cleveland,millstone,norwich
done

# 2. corpus (quiet rule checked live) + the crucible + seal + Form:
gaiaftcl eclipse-geometry --eclipse 2024-04-08 --build-fossils
gaiaftcl eclipse-invariant --seal --save                            # exit 0 ⟺ CALORIE

# 3. the SAVE→RELOAD→RE-VERIFY loop from disk only:
gaiaftcl invariant-verify --game eclipse                            # exit 0

# 4. the ledger row:
sqlite3 "$HOME/Library/Application Support/GaiaFTCL/substrate.sqlite" \
  "SELECT game,terminal,stamped_at_iso FROM ruliology_memory_state_changes
   WHERE game LIKE 'ECLIPSE%' ORDER BY stamped_at_iso DESC LIMIT 5"

The benchmark scope is pinned: as the fossil corpus grows multi-eclipse (2017, 2023, shear), invariant-verify --game eclipse keeps re-proving over exactly the sealed 2024 scope, so the fingerprint above never drifts from corpus growth — the historical runs seal their own separate Form (Historical corpus).