06Articles
Field notes on building production AI.
Practical writing on what separates a demo from a system that survives contact with real data — evaluation, agents, and the judgment work in between. First pieces publishing soon.
01Why most AI pilots never reach productionThe gap between "it worked in the demo" and a system you can trust every day — and how to close it.
02Evaluation is the real moatAnyone can wire up a model. Measuring it honestly, before and after launch, is the discipline that compounds.
03Designing executable boundaries from fuzzy requirementsTurning "make us more efficient" into a scoped system with clear, testable limits.
04When to use an agent (and when not to)A practical test for whether a task actually needs autonomy, or just a well-built automation.
◆ Notes from real builds, not press releases. Subscribe via an intro call for now — a proper feed is coming.
→Let's Talk
Have a problem worth writing about?
Start with an intro call. From there, an AI Blueprint turns your highest-ROI opportunity into a costed, buildable plan — yours to keep whether or not we build it.