Categories
Deep Dives
MLOps
Notes
2026
A Stale Vector Store Caused a Phantom Regression
A Quality Gate Must Be Reproducible Before It Can Gate Anything
Confident Matches on Vocabulary, Not Mechanism
Why Layer 2 Needed Its Own Threshold
The Test Suite That Was Too Slow To Run
The Layer 2 Baseline, and a Grounding Filter That Leaked
Chasing a Bug That Didn't Exist
BUG-001 — A Threshold Applied Where It Didn't Belong
The three layers where ML systems fail
Why this site exists
2024
The Non-Code Half of Engineering: Lessons from a Front-End Internship