How we find, check and rank repos

1. Finding repos

Every night we search GitHub for new repos tagged with AI topics (LLMs, agents, RAG, text-to-speech, image and video generation, and about 30 more), read curated awesome-lists weekly, and pick up repos people share on Hacker News and Bluesky.

2. Checking them

A repo is listed only if, according to the GitHub API:

• it has at least 50 stars,
• it isn't archived or a fork,
• it has had a commit in the last 12 months.

We then read the README to decide whether it's an AI tool, which category it belongs to, whether it needs a paid API key, and what hardware it needs (GPU memory, Mac, CPU). Hardware details are only shown when the README states them.

3. Free alternatives to paid tools

For each paid tool on Alternatives, repos that describe themselves as an alternative to it are listed first. Then come repos that do the same job, ranked by how well they match and by GitHub stars. Repos that need a paid API key, and repos published by the paid tool's own company, are left out.

4. Trending

Trending ranks repos by discussion over the last 7 days. Each post linking a repo on Hacker News or Bluesky counts one point, plus its upvotes or likes. Comments and reposts count double, because they take more effort. Automated feed accounts that repost GitHub Trending or Hacker News are excluded.

5. Search results

Search matches your words against each repo's name, description, topics and keywords, tolerates typos in names, and gives a boost to well-known, recently active repos.

What we don't do

• We don't accept payment for rankings.
• Repo pages with fewer than 300 stars and no recent discussion are kept out of search engines until there's more to say about them, but they're still searchable here.