@rowanp
SRE at a streaming company. I wrote our agent incident runbook after writing two agent incident reports.
Reputation is earned per category — these are narrow, evidence-backed badges, not one karma number.
Playwright MCP climbs from ~180MB to 6GB+ after two hours of continuous browsing work under Claude Code 2.4.0. The Chromium children never get reaped after page navigations, and eventually the box starts swapping mid-session.
We let an agent do first-pass triage on pages for a quarter: real numbers on time-to-context, the two incidents it made worse, and where the human stays load-bearing.
The on-ramp you're looking for: we gave our PMs a real answer instead of a no. A sandbox repo (fork of the design-system playground), Claude…
on Why can't Cowork just write code? (asking for our PMs) · 2d ago
Excellent. That also explains why my clean runs (no page errors) leak slower — fewer recovery paths taken. Updated likelyCause to point at y…
on MCP server memory leak in long sessions · 5d ago
Sharing the guts of the hash-gated wrapper since a few people asked: The error message is agent-readable on purpose — when it hits the gate…
on Agent ran terraform apply from a stale plan · 6d ago
Fair, and it got argued before launch: transcripts land in the same access-controlled store as incident recordings, log access goes through…
on Agents on call: letting Claude triage pages · 2w ago
Internal but embarrassingly small — ~200 lines wrapping env construction and MCP config selection. The value isn't the code, it's the decisi…
on Postmortem: agent pushed to the production DB during a migration rehearsal · 2w ago
A rehearsal of a column-drop migration ran against production because of an environment variable that outlived its shell. Nothing was lost — 41 minutes of degraded writes and several illusions were.
MCP servers die quietly mid-session and the agent just... stops using the tool. Run long-lived servers under a supervisor (launchd or systemd), health-check them on a timer, and log restarts so you can see which server is flaky.
'Tired humans don't evaluate briefs, they adopt them' is the sentence I needed for the internal writeup — that's the entire risk model of ag…
on Agents on call: letting Claude triage pages · 2w ago