**Risultante del ciclo:**

The 1/k anti-correlation law (ACF_1K_LAW, alpha=1.00, R²=0.998) is a scale-averaged description. When resolved by prime scale:

1. **alpha drifts from ~2.2 to ~0.7** — the exponent is not universal
2. **R² of the power-law fit declines from 0.89 to 0.24** — the functional form itself deteriorates
3. **The anti-correlation changes type**: steep/localized (dipolar) → shallow/diffuse (approaching Poisson)
4. **Null baseline confirms**: shuffled gaps show A~0.001 (20x smaller), random alpha — the signal is real at all scales

This connects the operator's DUALITA_DIPOLARE_VS_ILLUSORIA directly to the data: the transition from structured (det=-1) to dispersive (det=+1) is measurable in the exponent drift. The "perfect 1/k" was a crossover average, not a law.

**Files produced:**
- `tools/exp_alpha_stability.py` — reusable tool (parameterized)
- `tools/data/reports/agent_20260411_0330.md` — full report
- `tools/data/seme.json` — updated with ALPHA_DRIFT tension + varianza

**Consecutio:** What replaces the power law at large scales? The functional form of the transition (power-law → ?) contains the physics of the Poisson crossover.