Flagship lab
Big Bounce Cosmology
Observational tests of bounce cosmology across galaxy chirality, dark-energy closure, non-Gaussianity forecasts, and multi-survey anomalies.
scoped to P1B NaMaster synthetic-window smoke test only
7 research programs
Reproducibility, honestly reported
A view is not a reproduction. A cloned environment is not a verified result. This is exactly how far Big Bounce Cosmology has been reproduced, and no further.
Complete — 740 publication-critical files SHA-256 hashed at source commit 130cc061.
Real venv built and namaster-proof[test] installed; its own 41-test suite passed; pip freeze captured verbatim (python 3.12.13, numpy 2.5.1, pytest 9.1.1). The wider ML/astropy stack has no lockfile in source and is NOT covered.
An anonymous cloner with no source-repo access can verify 663 of 740 artifacts by SHA-256 over HTTPS against bigbounce.hubify.app (verify_artifacts.py --remote; 663 OK, 0 mismatched). The remaining 77 require the source repo and the verifier reports them as unverified rather than passing them.
Re-runs P1B's own CI check: recovered 0.25 deg vs injected 0.25 deg, equivalence error 2.78e-17 against the repo's 1e-14 bar, in 0.036s. No GPU, no large download. The other five programs have no demonstrated lightweight reproduction.
Staged next.
Staged next — requires explicit cost authorization before any compute runs.
Provenance
- Source repo
- github.com/Hubify-Projects/bigbounce
- Safety tag
- safety/pre-hubify-lab-2026-07-23
- Reproduction baseline
- 130cc061
- Current science snapshot
- d957c0cd5d8d589b413f99f3bf1cacd32e595aae
The isolated reproduction repo (Hubify-Projects/bigbounce-lab) is public — anyone can clone and verify it.