Resources and baseline
Fresh isolated worktree, starting at origin/main 8feec794. Shared checkout and its uncommitted files were left untouched. Python is the repository venv (3.14); python-flint and mpmath.iv are both installed. This hunt's numerical probes use mpmath, not interval arithmetic.
Before mathematical implementation, the focused baseline passed 29 tests with 5 slow tests deselected in 6.95 seconds:
.venv/bin/python -m pytest -q -n0 -m 'not slow' tests/test_huntspec.py tests/test_docs_numbering.py tests/test_hunt_probe_discipline.py tests/test_doors.py .venv/bin/python scripts/make_context.py --check
The context check passed. This is not a full-suite green claim. The preceding central_moments/RUNS.md records an existing dossier staleness failure. No heavy local suite or Lean build is authorized by this hunt.
Pilot and estimate, before the sweep
A foreground pilot at 40 decimal digits evaluated H_1(200) and its first two derivatives by incomplete gamma functions in 0.098 seconds. It measured endpoint slope 0.039498765267749934683 and scaled curvature -1.7401769956.
Planned sweep: N=1,2,3 at t=200,400,800, at 50 and 80 digits, 18 cases. At average N=2, three times the pilot cost for higher precision gives an estimate of 18*2*3*0.098 < 11 seconds for that component. Independent quadratures and actual Xi derivatives get a separate 90-second foreground cap. No paid compute, detached work, or build is launched. Console progress, explicit case counts including zero failures, and nonzero exit on any failed check are required.
Terminal record
The probe completed in 14.344 seconds. It wrote results.json with 18 truncation evaluations (16 negative L_1, 2 positive), 8 actual-Xi evaluations (0 negative L_1), 6 independent quadrature/derivative comparisons, and 2 precision runs of the nonreal-zero control. These are floating-point checks. The all-N asymptotic sign and the exact rational control failure are separate ordinary derivations in RESULTS.md.
The initial 14 hunt tests passed in 3.04 seconds. The initial combined check passed 43 tests, with 5 slow tests deselected, in 11.06 seconds. Generated context was regenerated and remained unchanged. Its check and git whitespace checks passed. No full local suite, Lean build, paid resource, or detached research job was run. The repository's existing CI is not included in those local counts. No external mathematical review was performed.
A subsequent algebraic refinement strengthens the control: for each finite order cutoff K, C_K passes every order through K at every real argument while retaining the explicit nonreal zeros. This adds no scan of zeta and no heavy computation. Three additional test cases check the parameter bound and exact failures of C_2 at order 3 and C_3 at order 4. The revised combined suite passed 46 tests with 5 slow tests deselected in 11.34 seconds, including all 17 hunt cases. The numerical artifact above is unchanged; it records the original K=1 control. The generalization and its exact witnesses are reproduced by test_probe.py.
id: rh_theta_laguerre-2026-10-03-checkpoint1
hunt: rh_theta_laguerre
started: 2026-10-03T04:41:47Z baseline completed; exact session start not recorded
finished: 2026-10-03T05:05:33Z refined local verification recorded
ran:
- .venv/bin/python hunts/rh_theta_laguerre/probe.py
- .venv/bin/python -m pytest -q -n0 hunts/rh_theta_laguerre/test_probe.py
- .venv/bin/python -m pytest -q -n0 -m 'not slow' hunts/rh_theta_laguerre/test_probe.py tests/test_huntspec.py tests/test_docs_numbering.py tests/test_hunt_probe_discipline.py tests/test_doors.py
- .venv/bin/python scripts/make_context.py
- .venv/bin/python scripts/make_context.py --check
outcome: RH unresolved; finite-theta asymptotic failure and exact finite-order-positive nonreal-zero controls preserved with 46 passing checks
artifacts:
- hunts/rh_theta_laguerre/MISSION.md
- hunts/rh_theta_laguerre/RESULTS.md
- hunts/rh_theta_laguerre/RUNS.md
- hunts/rh_theta_laguerre/probe.py
- hunts/rh_theta_laguerre/results.json
- hunts/rh_theta_laguerre/test_probe.pyCI check and documentation correction
CI run 37098771010 (https://github.com/teal-sea/zeta-lab/actions/runs/37098771010) finished with 3179 passed, 1 failed, 3 skipped and 6 expected failures in 1041.72 seconds of pytest time. The single failure was this hunt's doors section: it covered the subjects but did not use the required labels "constraints" and "information class". The focused command had included test_doors.py (reader entry points), not the different test_hunt_doors.py. This was a missed relevant check, not a baseline defect or a numerical failure. Only those two prose labels were changed; no test was weakened.
Including the missing test file, the local check on 2026-10-03T05:26:48Z passed 64 tests, with 5 slow tests deselected and 3 existing expected failures, in 10.60 seconds. Reproduce with the combined command above plus tests/test_hunt_doors.py. The existing CI rerun is expected to take about 18 minutes, based on the first run's measured 18m7s job duration. It uses standard Actions runners and is watched to a terminal state. The earlier superseded run 37098415230 ended cancelled when the control refinement was pushed; it is not a passing run.