Small, dated, tested records from real AI-agent development work.
What this is
Each record documents one concrete technical problem I actually hit while building things with
AI coding agents — what the exact failure was, which tool and version, the date I tested it,
what worked (or didn't), and what my record adds beyond the original issue threads and docs.
How records are made
Every record comes from real work — no synthetic test farms, no rewrites of other people's issues.
Each page states the last tested date and the exact environment. Pages are not silently updated; if something changes, the page gets a dated revision note.
If the original issue or doc already covers a problem completely, no page is produced.
There is no monitoring, no mailing list, and no promise of coverage beyond what is written.