About AFH Findings

Atoms for Humanity publishes results here as soon as we understand them, rather than waiting for a paper. A finding is small: one claim, the numbers behind it, where it holds, what was not tested, and the data and code to check it.

What to expect

Not peer-reviewed. We read each other's findings, but nothing here has been through formal review. Some are marked verified: a script published with the finding recomputes its numbers from its data, and it passed when the finding was published.
Versioned. When later work changes a finding, we publish a new version and say how it changed (corrected, expanded, strengthened, weakened or revised). Older versions stay readable at /F0001/v1/, with a note pointing to the latest.
Corrected in public. A finding that turns out to be wrong is retracted, not deleted: it stays up with the reason. Only content that should never have been public (for example personal or partner data) is withdrawn, and a notice stays at its address.
Written with AI assistance. We use AI tools for analysis code and drafting. The named authors checked each finding and are responsible for every claim in it.

Citing a finding

Every finding page has a Cite button. Cite the version you read, for example https://findings.afh.org/F0001/v2/, since the latest version can change. Findings that settle into a longer write-up are published on The Stacks, which gives them a DOI; the finding links to it.

Reusing our work

Unless a finding says otherwise: text and figures are CC-BY-4.0, data is CC0 (CC-BY-4.0 when it derives from CC-BY sources, which you should credit too), and code is MIT. Each version's Files & history page lists its files with SHA-256 checksums, and its zip holds everything.

For machines

Atom feed of new findings and versions · index.json · index.csv · sitemap. Each version's finding.json describes it.

Found a problem? Tell us. We correct findings in public.

Atoms for Humanity is part of the Astera Institute, which publishes openly under its open science policy.