Writing

Notes from the field on applied AI.

Essays on the engineering, evaluation, and judgment that take AI systems from a demo that works once to something a team can depend on every time.

Engineering

The model is the easy part

Engineering

The LLM does the reading. Deterministic code owns the truth.

Engineering

Vercel to Cloudflare Workers: not as straightforward as I assumed

Strategy

Shipping AI you can trust

Evaluation

Evaluation is the real product

Engineering

RAG that survives production

Strategy

Where AI actually earns its keep

Evaluation

The cost model nobody builds until it's too late

Engineering

Agentic workflows without the hype