Where AI-built things become products.

We compare healthy public GitHub repositories of public projects. Only successful observations within 1 hour(s) before the 00:00, 08:00 and 16:00 KST cycles are used; the actual interval between observations must be 7–9 hours. Projects must have 50–1,999 stars this cycle and gain at least one.

Sorted by gain, then current stars, then project URL. The home growth top five and the full rising list use the same cycle, eligibility and order. The browse tabs below show a separate public project catalogue. Growth cards emphasize “8h +gain” with total stars secondary. Hovering the number, or using assistive technology, exposes both actual observation times.

We use successful observations from the latest scheduled cycle; the last observation must be within 10 hours. Failed or missing comparisons, repository link changes and missed deadlines are excluded. Newly discovered projects enter collection in the next cycle and become eligible after baseline data is available.

“Last 8 hours” is the net change between the latest closed cycle and the cycle 8 hours earlier, rather than a live count of all star events during a rolling eight-hour window. Missing comparisons show a status message; total-star rankings do not replace them. Blocked, deleted or newly private projects are hidden immediately. Other rankings remain fixed until the next cycle.

Each tier ends just below the next tier; 30k+ ends below 100k. Ties use listing ID order. Home shows the top three per tier; the full list shows 15 per page.

Stars indicate interest in a repository, not verified users or quality. “Personal accounts only” includes GitHub User accounts; organizations and unconfirmed accounts are excluded.

Star counts begin with stored GitHub data and are rechecked in turn after at least one day, currently usually every 2–3 days. Failures retain the last successful value and date; queues can delay checks further. Nonpublic, blocked, unreachable and 100k+ star projects are excluded. Production evidence is available on each project page.

We count public projects whose GitHub repository was first created during the last completed seven days. This is neither the listing date nor a release date.

Discovery and review take time, so projects created in recent days may be added later. Unavailable rates are never filled with 0%.

Home’s “Created in the last 7 days” and its full list use the same period but include only public projects with at least 50 stars and a confirmed healthy GitHub repository, newest repository first. Repositories with unknown creation dates are excluded.

Each project is counted once, even with multiple releases: 28,358 releases belong to 8,344 projects.

Description edits and automatic redeploys are excluded. These figures do not guarantee code quality or uptime.

New versions include GitHub releases and maker updates. Ties are ordered by name.

Projects that split work into smaller releases may rank higher. This measures activity, rather than quality.

“Other” is excluded from rankings. Personal profiles only appear when registered or confirmed by their owner, so they are omitted from category rankings.

Merged AI agent PR — A record of a merged PR opened by an AI agent. It describes that PR’s changes, not the entire product.

AI commit attribution — Commit messages attribute contributions to an AI tool. This is repository-provided attribution, not technical verification.

Tool-specific configuration files — Configuration files specific to an AI coding tool exist. They indicate tool use, not the share of generated code.

No evidence proves that AI made the entire product. Agent PRs only describe their merged changes, while commit signatures and configuration files are self-reported and can be fabricated. Repositories with only shared instruction files such as CLAUDE.md or AGENTS.md do not get the “AI” badge.

Repositories are checked again every three days. Projects without a repository cannot be scanned and have no “AI” badge.