Writing

What I keep coming back to, and where I have written about it. Longer pieces are on the way. For now these are the ideas I would defend in a room.

Things I believe

Adoption is a confidence problem

Almost every failed AI rollout I have seen was treated as a licensing or training problem. People do not avoid new tools because they cannot use them. They avoid them because being visibly bad at something in front of colleagues is unpleasant.

Programs die, systems persist

A workshop changes behavior for about two weeks. If the new way of working is not supported by something automated, the old way wins, because the old way requires no effort to remember.

Build the thing before you have an opinion about it

I cannot help a team adopt tools I have only read about. Building two products solo taught me more about what AI is actually good at than any amount of analysis would have.

Elsewhere

LinkedIn2026-07

I built two AI products from scratch, and got beaten to market twice

Satori, then AIgentz. Both reached working MVPs with real testers. Cursor shipped one, OpenAI shipped the other. The biggest takeaway was not about the products. It was that one person with some technical knowledge and a willingness to learn can now build what used to require entire teams.

LinkedIn2026-06

Closing five years at Lightricks

I joined as a student, not knowing where it would go. What followed was more than a job: a place that challenged me and gave me room to grow. Companies are made special by their people.

LinkedIn2025-09

Are personalized apps the future?

Someone close to me was frustrated that no cycle-tracking app worked the way she wanted. I built one in under ten minutes, exactly to her spec. Software for one person is now worth writing.