Everpresent Lambda adjudication

Public code, machine-readable results, and reproduction instructions for:

Supernova success and sound-horizon failure of Everpresent Lambda: an independent test of stochastic dark energy

The paper independently reproduces the rare-supernova-tail selection claim of Everpresent Lambda Model 1, then tests the selected histories against cosmic chronometers and DESI DR2 baryon acoustic oscillations.

Result in one paragraph

Two independent implementations recover the same six Pantheon+SH0ES winners from 90,000 seeded histories under both tested redshift conventions. Under the declared calibrated background sound-horizon prescription, all six are disfavored by the joint supernova, chronometer, and DESI DR2 BAO likelihood, with nominal penalties from +128 to +1974. A deliberately favorable ruler stress test still leaves a minimum penalty near +83.

This is a conditional result for those six selected histories. It is not a universal no-go theorem for Model 1 and does not replace a self-consistent recombination and perturbation calculation.

Public artifacts

• Paper, source package, and archived release
• Detailed reproduction protocol
• Paper source
• Primary implementation
• Independent verification implementation
• Machine-readable results

The PDF is served from Zenodo so that the cited manuscript and its immutable release files remain tied to the DOI.

Repository scope

This repository is a deliberately narrow public release. It was built from the allowlisted reproducibility archive already published on Zenodo, not from the author's private research workspace.

It contains only the code, small redistributable inputs, derived outputs, manuscript source, and reports needed to inspect this paper's evidence chain. Private notes, unrelated research, credentials, correspondence, local paths, and the private knowledge base are not included.

See PROVENANCE.md for the source archive hash and the small documentation-only cleanup applied to this public repository.

Quick verification

Python 3.14.3 was used for the frozen release. Create an isolated environment:

python3 -m venv .venv
source .venv/bin/activate
python -m pip install -r requirements.txt

Download the external inputs from pinned upstream commits:

./output/papers/everpresent-adjudication/download_data.sh
shasum -a 256 -c output/papers/everpresent-adjudication/checksums.sha256

Run the focused test suite:

python -m pytest \
  tools/inhomogeneous-cosmology/test_e1_everpresent_model1.py \
  tools/inhomogeneous-cosmology-codex-verify/test_verify.py \
  tools/inhomogeneous-cosmology-codex-verify/test_verify_bao.py \
  tools/inhomogeneous-cosmology-codex-verify/test_verify_full_90k.py

Run the official-stack one-seed ruler stress test:

python tools/inhomogeneous-cosmology/e5_official_refit.py \
  --pantheon-dir output/computations/holographic-de-desi/pantheon_plus_data \
  --seeds 5009 \
  --output /tmp/e5-seed-5009.json

The expected envelope-refit penalty is approximately +83.3. The full 90,000-seed scans are computationally heavier; exact commands and expected artifacts are documented in REPRODUCIBILITY.md.

Directory map

tools/
  inhomogeneous-cosmology/                primary analysis
  inhomogeneous-cosmology-codex-verify/   independent verification
raw/datasets/                              small included inputs and data notes
output/computations/                       load-bearing JSON artifacts
output/papers/everpresent-adjudication/    manuscript and reproduction tooling
output/*.md                                stage reports and audit trail

Data and licensing

• Original code is MIT licensed; see LICENSE.
• Original manuscript text, reports, and compatible derived research content are CC BY 4.0; see LICENSE.text-data.md.
• Third-party datasets retain their upstream terms. Large Pantheon+ and DESI likelihood files are downloaded from pinned upstream commits and verified by SHA-256 rather than committed here. See THIRD-PARTY-NOTICES.md.

Automated-assistance disclosure

Anthropic Claude and OpenAI Codex assisted with literature work, analysis code, cross-checking, and drafting. The manuscript states their roles and limitations. Numerical claims are tied to inspectable artifacts, two implementations where stated, and explicit reproduction commands. This public release is a non-peer-reviewed preprint.

Citation

Use the metadata in CITATION.cff, or cite:

Simon, R. (2026). Supernova success and sound-horizon failure of Everpresent Lambda: an independent test of stochastic dark energy (Version 1.0.0). Zenodo. https://doi.org/10.5281/zenodo.21721755