Field notes · Updated weekly

Writing for builders, not browsers.

Career strategy, technical deep-dives, hiring patterns, and honest takes from working engineers. No SEO filler. No course-mill copy.

Articles

SHOWING 4 OF 4
ai·9 MIN

RAG isn't a system, it's a 12-stage failure surface

Where retrieval pipelines actually break in production, and the eval harness that catches it before users do.

ai·13 MIN

An eval harness that survives contact with a real user base

How to build LLM evaluation infrastructure that catches regressions before users do — and stays maintained.

ai·10 MIN

Why your agent demo doesn't survive a real workflow

The five failure modes that kill agent systems after the demo — and the architecture patterns that prevent them.

ai·9 MIN

When fine-tuning is worth it (and the cheaper alternatives that usually win)

A decision framework for choosing between prompting, RAG, fine-tuning, and full training — with real cost numbers.

The trail dispatch · weekly

One letter every Sunday. Nothing else.

The week's best writing on AI, careers, and the craft — curated by working engineers. No drip campaigns, no sponsored anything.

NO DRIP CAMPAIGNS · UNSUBSCRIBE ANYTIME