[jeremymanning]

The guide, API reference, gallery, and all 13 tutorials are self-contained first-release documentation. Experimental datasets and fictional tasks are introduced as illustrative examples, with their assumptions and interpretation limits. PyPI installation instructions and badges preview the release, and the Mitchell tutorial uses the current HyperTools Figure API.

The audit also corrects public behavior: permutation refits use a common label-independent factor reference; undefined statistics retain undefined p-values; spatial inputs reject mismatched registration and complex values; mixture optimization handles vanishing-component scaling and extreme step norms; and plotting keeps readable colorbar ticks and comparable group color scales. Regression tests cover these paths. The group tutorial includes a recomputed, checksum-verified 100-permutation asset.

Performance claims use 24 matched fits on an isolated runner with CPU-interference checks, recorded inputs/settings/source hashes, and downloadable raw monitoring. An independent reviewer verifies the measurements and summary arithmetic. Fresh reviewers also audit the documentation, scientific interpretation, API behavior, and rendered tutorial/gallery outputs.

Download the rendered HTML documentation preview.

Validation:

• Installed wheel: 1,067 passed, 17 skipped, 28 deselected.
• All 13 notebooks execute; outputs and figures reviewed.
• BrainIAK: 27 passed; source-distribution smoke/statistics checks: 15 passed.
• Network suite: 27 passed and one NITRC download timeout; the exact failed test passes on retry. The replacement group-null asset also passes a fresh download check.
• 341 ordinary-RBF fingerprint arrays remain identical across the numerical hardening.
• Ruff, formatting, mypy, package metadata checks, strict Sphinx build, and all local HTML file/image references pass. The wheel README is explicitly rendered with readme_renderer[md], and its image/document URLs are absolute for PyPI. The README-only packaging correction preserves every packaged library file byte-for-byte.
• Final commit CI: all workflows pass, including all 16 source-test jobs and all 16 installed-wheel/tutorial jobs. tests, installed-wheel/tutorial matrix, BrainIAK, docs.
• Isolated benchmark.

No package publication: PyPI/TestPyPI remain untouched and the GitHub release stays a draft. The Zenodo DOI badge is deferred until after review and the approved release. See notes/RELEASE_REVIEW.md for evidence and remaining publication gates.