Build a Specialized AI Inference Startup

Recordings go up the same day, and your pair partner will have your notes, so you won't fall behind on the material. The pair feedback is the one part you genuinely can't get from a recording, so if you have to miss one session, try to make it one of the first two.

Three seventy-five-minute Fridays at noon Eastern, plus two to three hours of build work between sessions. Most of that work is on the job you brought with you, so in my experience it tends to feel less like homework and more like finally getting to the thing you've been meaning to do.

You don't, and I'd rather you didn't buy one for this. Most of what we do runs on a recent laptop, and I run the heavier training jobs on my GB10 for anyone who needs them, because the point is to learn what the hardware actually buys you before you spend money on it.

Then you'll know, with your own benchmark as the evidence, and you'll know what the gap costs in dollars. That saves you from building the wrong thing, which is a better outcome than most people get from a course.

It goes wherever you decide it goes, and that decision is the first thing we make. In week one you choose whether you're solving a cost problem or a privacy problem, because those have different buyers, and then we test what actually leaves your machine rather than taking a vendor's word for it.