Pirate Face is a community platform for sharing open AI models through peer-to-peer torrents, with an activity feed that records model listings, downloads, and revisions witnessed in a public provenance log. Its central idea is to make model availability and history visible: users can see who submitted a model, the stated license, whether a revision was recorded, and when a download was initiated. The feed presents these events as community activity rather than as a conventional model catalog, and offers filters for event types and time periods.
The displayed feed contains 100 of 200 records and ranges from small specialist models - such as French CamemBERT classifiers and TinyStories models - to larger language models and fine-tunes, including Qwen, Gemma, Granite, and Korean YeJin variants. Many entries are torrent listings submitted by one account, with MIT or Apache-2.0 licenses; others show revision identifiers witnessed in the public log, including a cluster of “heretic” model variants. The page also tracks swarm participation and download activity. Its counters show 29,778 magnets, 52,753 seeders, and 51,508 witnessed records, alongside a timeline and a chart of seeders by model family. Users can contribute by downloading models, which the site frames as helping keep them available through the network.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.