About Chronicle

Each event starts from its primary evidence: the original paper, the model card, the court ruling, the official archive. Independent sources check what the primary source claims. Announcement, release and later impact are kept apart, and a provider's benchmark claims are attributed to the provider, never treated as independent results.

Every event has a full research report of at least 2,000 words drawing on at least 5 distinct sources, with each claim linked where it is made. Its significance is scored on a published rubric of five criteria, with a written reason for every rating, so you can disagree with a score and see exactly where.

The archive is curated, not exhaustive. New developments are weighed against the same standard as the old ones, and one joins only with its report: no announcement becomes an event automatically.

Chronicle is an independent project. Its research and writing are produced with AI models working to a published editorial standard, and nothing is published until it passes automated checks: a report that falls short of the standard, or cites a source it does not list, is rejected. The links are the guarantee. Every claim can be checked against its source in one click.

If you find an error, a missing source or a milestone we left out, write to us. Corrections go into the record itself, and each report shows the date it was last reviewed.

The event index (dates, names, categories, summaries, ratings, scores and primary sources) is open data under CC BY 4.0. Download it from the dataset page, or cite the archived release by its DOI, 10.5281/zenodo.23395380. The research reports are free to read, quote briefly and link to, and remain all rights reserved.