Your GitHub history as a seismograph. Lines of code per month are the ground motion; major AI model releases are the fault lines running through it. The trace lies near-flat for years, then ruptures — often right where the coding-model fault lines cluster.
gitquake is built to be driven by a coding agent, not installed by hand. Point any agent at this repo and say:
Build me a gitquake seismograph for GitHub user
<name>.
Your agent already has the runtime (Python) and GitHub creds (GITHUB_TOKEN / gh),
so it just runs python3 seismograph.py --user <name> --open and reads
AGENTS.md to customize. One stdlib-only script in, a self-contained
animated HTML file out. No package installs, no setup.
- Pulls your per-month additions and commits from GitHub's
stats/contributorsAPI across every repo you commit to (public + private if you own them). - Drops bulk-import repos (>10k additions per commit — vendored deps, datasets, generated files) so the trace reflects code you wrote, not code you pasted.
- Inlines the data and a curated AI-model-release timeline into a single animated HTML file. The seismograph pen draws itself on load; toggle Lines / Raw / Tremors.
- Caches each repo's stats by last-push, so the first run takes a few minutes (it is
almost entirely network wait) but every re-run finishes in seconds.
--no-cacheforces a full refresh.
- "Lines" is GitHub's additions count — a measure of activity, not artistry. The Raw channel keeps the bulk imports if you want the unfiltered truth.
- Model release dates are curated (
EVENTSinseismograph.py); the most recent ones are approximate. Edit them freely. - Correlation, not causation. The chart shows your output rose alongside the coding-model era. It can't prove one caused the other.
MIT — see LICENSE.