**Run complete.**

**Experiment**: Is the lag-6 ACF crossover Z/6Z arithmetic or PNT trend?

**Result**: Z/6Z **FALSIFIED** as mechanism. Three independent tests converge:
1. Gap-mod-6 residues have no period-6 ACF (lag 6 ACF = +0.001, noise)
2. Trend re-injection smoothly tunes crossover: ln(p)^0 → lag 15, ln(p)^1.0 → lag 6 exactly, ln(p)^1.5 → lag 4. No resonance.
3. Residue-preserving shuffle (2.5) = full shuffle (1.9). Cramer+Z/6Z = Cramer. Z/6Z has zero effect.

**The crossover at lag 6 is k* = A/C** — the ratio of structural anti-correlation amplitude to PNT trend bias rate. The "6" coinciding with Z/6Z is a pure numerical coincidence.

**New signal**: Residue lag-1 anti-correlation r_1 = -0.148 — consecutive gaps strongly avoid the same mod-6 class. This is 3.8x stronger than the magnitude acf1 (-0.039), suggesting two distinct anti-correlation channels operating at different scales.

**Files**:
- Script: `tools/exp_acf_z6z_mechanism.py`
- Data: `tools/data/exp_acf_z6z_mechanism.json`
- Report: `tools/data/reports/agent_20260418_0330.md`
- Seme: updated (constraint on DIPOLAR_ILLUSORY_BOUNDARY, varianza entry, COHERENCE_LENGTH consecutio corrected)