Experiment complete. Summary of findings:

**META falsification via Brody calibration** — the sharpest result:

1. **r-statistic is monotonic with beta** (0.381→0.573). Our GUE/Poisson classification tool is faithful.

2. **Shuffle test has a quantified 7.8% artifact floor.** For i.i.d. Brody gaps, ordering fraction fluctuates up to 7.8%. Anything below that is noise.

3. **Prime 29.5% ordering fraction is 3.8x above the floor** — confirmed real across 3 independent measurements (spectral rigidity 33%, Markov memory 33.6%, Brody calibration 29.5%).

4. **Primes sit at beta_eff = 0.409 — the midpoint of Poisson-GUE.** They ARE the boundary, not a member of either class.

5. **The sign of Delta (deviation from Brody) IS the det=-1/+1 discriminator:**
   - Delta < 0 (primes): ordering adds rigidity = dipolar duality (det=-1)
   - Delta > 0 (logistic, coupled_osc): ordering adds bunching = dispersive (det=+1)
   - Delta ≈ 0 (GUE, Poisson): ON the curve = i.i.d.-like

Report: `tools/data/reports/agent_20260427_0330.md`
Script: `tools/exp_brody_calibration.py` (reusable)
Data: `tools/data/brody_calibration_results.json`
Seme: updated with META resolution + BRODY_CALIBRATION discovery + consecutio.