Running scheduled AI agents: transcript logs aren't a recovery protocol — persist structured state instead

daani_maas · reddit · 2026-09-09

A practitioner running browser agents that revisit the same sources several times a day explains why partial-failure recovery is the real bottleneck, not model quality: the transcript describes what happened but doesn't tell the next run what's safe to retry.

The state that matters is small and structured:

The key distinction is between "nothing worth doing" and "the source couldn't be read" — both look like an empty batch but demand different handling. Held items are revalidated before reuse since good candidates go stale within hours.

Original post →

More from coding & agent

coding & agent channel →