# Studies

Empirical attachments of the Liber de Ludo Aleae score contract to a
live eval. Local to this symposium. Not kernel ADRs.

| Date | Study | Host | Claim type | Verdict |
|------|-------|------|------------|---------|
| 2026-08-14 | [S1 — Field trial inside a metaharness](S1-metaharness-scoring-field-trial.md) | `~/llm` | amber (n<30 on expensive instrument) | intel/hour 0.08→0.81. Noise floor measured. 566 tests ≈ 8 casts (ICC 0.94–0.97). Preflight 11d→30s. |
| 2026-08-14 | [S2 — Two circuits: Darwin vs live floors](S2-two-circuits-darwin-and-floors.md) | `~/dev/scifi-vector-oracle` | mixed | Darwin winner unchanged (0.985, +0.000; 6/9 children worse). Live floors: probe 0.962→1.000, cert 0.395→0.965. Do not mix the two circuits. |
| 2026-08-14 | [Counting the Night](2026-08-14-metaharness-overnight-calibration.md) | `~/dev/mudflood` | mixed | Readiness faces 70/75/79/100/38 unchanged. Work circuit: buildings 870→953, media 178→241, stories 53→133. |
| 2026-08-14 | [Overnight metaharness + Cardano sidecar](2026-08-14-overnight-metaharness-sidecar.md) | `~/dev/eql` | weak_scientia (n=104, circuit moved) | Table fair (*p̂* 0.404 ≈ seed 0.405). Method lift: named circuit, p≠np, zero promotes. Do not graduate LDA-001. |