Methodology

Where tools come from

Every tool is a real “Show HN” launch on Hacker News from the last ~90 days. We review each one by hand and only list actual AI products, models and developer tools (no essays, benchmarks, games about AI, or “HN without AI” filters). Names, taglines and categories are checked against each tool’s own site. Launch date = the Show HN post date. New launches go into a review queue daily, and founders can submit tools.

The score (0–100)

• Still online, 20: the site or repo still responds (daily HTTP check).
• Hacker News mentions after launch, 20: HN stories and comments mentioning the tool’s domain or repo 14+ days after launch (log scale). Tools on shared hosts (github.io, Hugging Face, etc.) get a neutral 6.
• X posts, 15: number of posts on X linking to the tool’s site or repo in the last 7 days (X API recent counts, log scale).
• Recent activity, 25: for GitHub projects, how recently code was pushed (under 14 days = full marks; archived = 0). Closed products get neutral partial credit (12) while online.
• Adoption, 15: GitHub stars (log scale), or launch discussion depth for closed products.
• Launch-day hype, 5: Show HN points. Deliberately small.

Statuses: Graduated = 90+ days and 55+ · On track = 60+ days and 55+ · Fading = 30+ days and under 35 · Dead = offline or archived.

Rankings can’t be bought

Paid fast-track listings and sponsorships buy faster review and a clearly labeled placement. They never change a score or a rank. The scoring code runs the same way for every tool.

Sources & limits

HN Algolia API, GitHub REST API, direct HTTP checks (refreshed daily), and X API recent post counts (refreshed as our X quota allows; the date is shown on each tool page). Mentions are matched by domain or repo URL. We don’t use Reddit data: Reddit’s API terms don’t allow commercial use without a separate agreement. Planned: Tranco traffic ranks, npm/PyPI downloads.