mitsuhiko: Claude Opus traces macOS BTLEServer CPU spin to an AirPods firmware check bug

mitsuhiko · x · 2026-10-18

Armin Ronacher (mitsuhiko) shared how an agentic Claude Opus session diagnosed his macOS BTLEServer pegging the CPU. The agent produced a verifiable hypothesis with evidence: a single thread (PID 821) burned 863 minutes of CPU since 19:17:59 the day before, exactly matching a UARP firmware-update check for his new AirPods Pro firmware (9A348), after which BTLEServer stopped logging and entered a tight loop—a bug in macOS 26.6.2, not real Bluetooth activity.

Fix: sudo killall BTLEServer (launchd restarts it); toggling Bluetooth or letting the AirPods finish the firmware update in their case prevents recurrence; sudo sample 821 5 captures stack proof. His takeaway: even when the fix is simple, an agent that hands you an investigable hypothesis is the real win.

Original post →

More from coding & agent

coding & agent channel →