From reviewing Tech to building companies with AI

If you scroll far enough back on this blog, you’ll find a teenager reviewing AirPods Max and arguing about iPadOS multitasking. I’ve decided to leave those posts up. Partly out of nostalgia, partly because they explain more about what I do today than my CV does.

Writing tech reviews taught me two habits that turned out to matter. First: test the thing, don’t trust the keynote. Every Apple event promised magic; the review unit told the truth. That’s now roughly my job — at Downstream Intelligence we build AI forecasting models and then test them in public, scoreboard and all. The second habit: explain complicated things to people who have better things to do. A camera review and a geopolitical scenario brief are, structurally, the same assignment.

So, in the spirit of the old blog, the AI tools that actually survive contact with my daily work: Claude for drafting and thinking through arguments — the first tool I stopped having to babysit. Our own model, which I get to watch evolve from the inside — because watching probability curves move with the news rewires how you read headlines. And a boring but real one: transcription plus a good notes system, because half of building a startup is remembering what was said in the last twelve calls.

What’s overhyped: anything promising to remove judgment from work that is fundamentally judgment. AI compresses the distance between a question and a well-structured answer. It does not tell you which question matters. That part is still on you — which, if you’re the kind of person who used to read 2,000-word iPad reviews, is the fun part anyway.