LLMs Are Too Big: Aronchick on Right-Sizing Models for Log Routing

aronchick · x · 2026-09-22

David Aronchick argues we over-ask of AI: general-purpose LLMs are routinely tasked with narrow jobs like deciding whether a log line deserves human attention, then constrained to stay on task. Using Jev for log routing at Expanso convinced him tasks are often smaller than models — capabilities for Navier–Stokes aren't needed to route an INFO-labeled config reload. His thesis: pick models sized to the job, not the biggest one available.

Original post →

More from coding & agent

coding & agent channel →