Useful artifacts, not content confetti
Essays, white papers, and reusable artifacts from my own product work. What breaks, what holds, and what I would do differently. New pieces are published here and on Substack first.
AI Product Work Has a Failure Rate
The real product question is not whether the model can be perfect. It is what happens when it is wrong, late, overconfident, under-sourced, or asked to make a decision it should not make.
RAG Is Not a Product Strategy
Retrieval can be powerful, but users do not wake up wanting RAG. They need trusted knowledge at the moment a decision has to be made.
Human in the Loop Is Not Safety by Itself
A human reviewer is not a meaningful control if the product makes the important errors hard to see, too expensive to catch, or impossible to correct.
Enterprise AI Is Mostly Coordination
The model matters. The architecture matters. But enterprise AI usually stalls when teams do not share definitions for safe, useful, accurate, ready, and owned.
The Best AI Portfolio Is Proof of Judgment
Screenshots are not enough when the most valuable AI work is deciding what to trust, test, ship, reject, and improve.
Use Figma as the Napkin, Not the Finish Line
AI can generate a screen, but production-minded design starts when the work moves into coded systems, scenario data, review paths, and pull requests.