Takara TLDR - Daily AI Papers
・When an expert corrects an LLM assistant's error, the correction usually dies with the session, and the error class returns.
・I argue this is an operations problem, not a tooling problem: mechanisms for persisting corrections exist and are shipping, but the discipline for governing them -- versioning with provenance, recurrence monitoring, counter-metrics, retirement of stale rules -- does not.
・Writing as a systems en