About

Most of what you'll read about AI is written by someone who gets paid when you build. Vendors, consultants, the platform selling seats — the incentives all lean the same direction. Mine don't. Nobody here gets paid when you build, which frees this site to care about the only question that actually matters: what's the right build? Sometimes it's smaller than the pitch. Sometimes it's later than the hype. It's always aimed at the version that pays — because the problem was never AI. It's bad AI, sold as strategy.

By day I'm the AI Tech Lead at The Third Floor, one of the entertainment industry's leading previsualization studios, where I develop the AI systems the studio runs on. I architected Coyote, the studio's proprietary AI platform — built so artists can run virtually any model or workflow inside a Tier 1 secure environment trusted with unreleased material from major studios, without the traditional tradeoff between creative capability and security. It's become the gateway through which the studio's artists use generative AI on active productions.

I mention it for one reason: film production is the least forgiving production environment there is. Dates that don't move, budgets burned by the day, material where a leak costs more than the project. "It worked in the demo" is not a sentence anyone accepts there. That standard — does it survive contact with a real Tuesday — is the one everything on this site gets rated against. (Everything here is my own view, not the studio's.)

The route in wasn't straight. I did a PhD in neuroscience at USC, studying how the brain perceives motion — how we build a confident model of a moving world out of incomplete signals. Strangely good training for this moment, because the biggest problem in AI right now isn't the technology. It's perception. Demos are engineered for exactly the shortcuts our brains like to take, and knowing where those shortcuts live is half of what this site does. Then I spent a decade producing and directing across film, television, and branded work in Los Angeles, learning what deadlines and budgets do to beautiful plans. Then years advising companies on applied AI — training more than two hundred executives to tell practical AI from hype — which is where the questions this whole framework runs on got their road-testing.

The Before You Build trilogy came out of all of it: where AI work should live, whether a thing should be built at all, and how to run it once you've said yes. This site is the framework working in public. The tools make the calls. The glossary keeps the vocabulary honest. And every week, The Greenlight runs one real decision through the questions and reports what happened.

If you've got one on your desk, bring it. The right build is usually in there somewhere — smaller, later, or different than the one you walked in with — and finding it is the whole job.