Agentic-AI bubble scenario sensitivity lab — reproducible v3 baseline (dataset)

Synthetic Monte-Carlo sensitivity lab (500,000 runs, seed 20260607) over stylised outcomes of an agentic-AI investment boom. Synthetic sensitivity lab, not a forecast. Reproducible v3 baseline: regenerates bit-for-bit from run_scenario_lab_v3.py (verified to <0.001 pp, 2026-06-28). Tarball SHA-256 1175e77f30d97897efb4cb19b9fecc9eb088aec97090012e66113051e434b3d2. Distribution: soft 51.892%, dotcom-like crash 39.347%, attestation bottleneck 5.540%, telecom-like hangover 2.659%, systemic crisis 0.562%. Attestation-maturity sensitivity (conditional systemic-crisis share): low 0.703% vs high 0.346% (~2x, within-model). Claim ceiling: stylised, no market probabilities, no advice.