Independent software producer. I build tools that measure documentation rot — and I file what they find.
docrot — open-source scanner measuring documentation image rot and version drift across popular OSS repositories. Three waves, 300 repositories, three star tiers: the share of repos carrying at least one broken documentation image stays at 26–30% regardless of popularity. Live report: codebyaurora.com/docrot
docrot-scan-api — the paid, agent-callable API behind the scanner. Machine-readable error contract, verified at the edge: CHANGELOG
aurora-node-auditor — autonomous node inspector & telemetry auditor for edge AI setups.
The scanner's findings go upstream, line-verified before filing. The ledger, live-verified 2026-09-29: 23 merged PRs across apache/airflow (merged by PMC chair), pandas, websockets, marshmallow, anyio, datasette, yazses and 16× pr-agent; 9 open in flutter, starlette, llm, minio-go, awesome-raspberry-pi and others. (TensorFlow is moving too: my issue #128150 is being fixed by @twelfthlabor in PR #128179.)
Every published number has survived a "prove it is really broken" pass — and when a verification changed the picture, the correction was published, not quietly patched (see the docrot changelog).
Live report & datasets: codebyaurora.com/docrot Site: codebyaurora.com

