=== KKLT-CERT W0 checks ===
certified |W0|(s*, tau*) balls:
  route R dps  60: mid=0.00000002036636685067410293293610713062482635492  rad=1.55e-76  rel_rad=7.61e-69
  route R dps 150: mid=0.00000002036636685067410293293610713062482635492  rad=4.62e-166  rel_rad=2.27e-158
  route C dps  60: mid=0.00000002036636685067410293293610713062482635492  rad=5.95e-67  rel_rad=2.92e-59
  route C dps 150: mid=0.00000002036636685067410293293610713062482635492  rad=9.16e-157  rel_rad=4.5e-149
[c1] dps-60 vs dps-150 matched digits (worst of R,C): 59   -> PASS (bar: 55, dps-60 radius-limited)
[c4] route R vs C: balls overlap = True; matched digits (dps150) = 148   -> PASS
[c2] same-point model-vs-certified: 4.489e-19 rel (model value read from out_model_at_point.txt; the same model-truncation offset is recomputed at the vacuum by ../../kklt-evaluate.py [v1-vs-ball] and ../cert_w0_vac/run_vac.py checks [gc4])
[c2] pilot VACUUM value vs certified curve-point value: rel diff = 0.000230482  == measured transverse-offset penalty (not an error: the curve point is not the vacuum; see README.md)
[c3] honest radii quoted above; headline rel radius (R,150): 2.27e-158
[st] tau_hat from curve stationarity = [+/- 1.60e-160] + [-6.69283972452 +/- 1.36e-12]j (vs vacuum 6.85546i): curve-restricted D_tauW=0 is offset-corrupted; tau is fixed externally (see README.md)
wrote checks_cert.json
