all writingAll writing
ai2026-02· 11 min
LLMs as anomaly detectors: lessons from 100k tx/day
What worked, what burned tokens for nothing, and the eval harness we built to keep it honest.
draft · long-form essay in progress
This piece is in the editing queue. The argument is sketched, the evidence is logged, the diagrams are half-finished. When it ships it ships — not before. In the meantime, the headline blurb above captures the thesis.
If you'd like to be notified when it's published, drop a line via the contact section.