the team of one.

movement 1

the team changed. the standard didn't.

Until recently, shipping a real product meant coordinating multiple people: someone to decide what to build, someone to design it, and someone to engineer it. Three crafts had to point in the same direction before anything could ship.

AI changed that handoff. One person can now move across all three jobs.

I know because I did it. I stopped waiting for a team and built this site the way we teach: from spec to production, including payments, error handling, tests, and the checks that keep it working. The build story is public because the work should be visible.

The opportunity is obvious. The responsibility is less obvious.

AI can produce code and screens quickly. You still decide what is worth building, what good looks like, and whether the thing works for someone who has never seen it before.

exp.club is for people who want to take that responsibility on. The three jobs did not disappear. They became one person's work.

movement 2

from demo to product

Vibe coding is useful. You describe what you want, accept what the AI produces, and get a demo in front of users fast.

That is how we use it in the first week of the program. It helps you learn what people want before you spend weeks building the wrong thing.

Then the work changes.

A real product has to survive a stranger pressing every button. It needs to handle empty states, bad inputs, a dropped network, a half-completed payment, and the parts nobody includes in a polished demo.

That is agentic engineering — Andrej Karpathy's name for it, a year after he coined “vibe coding” . Agentic, because you are not writing the code directly; you are orchestrating agents who do, and acting as oversight. Engineering, because there is an art and a science to it, and you can get better at it. The goal in his words : take the leverage of agents “without any compromise on the quality of the software.”

You write the spec. Agents implement. You review what came back, test the result, and own the decision to ship it.

On this site, I wrote the spec before any code. Agents did the implementation. I did not read every line. I owned the spec going in, a review pass that attacks what came out, and a gate that checks every route for accessibility, rendering, and phone-width failures. None of that is airtight. It's just better than pretending you read everything.

vibe coding raises the floor;
agentic engineering holds the bar.

The point is not to get faster at prompting. The point is to turn AI output into software people can use.

movement 3

the skill that lasts

The tools will keep changing. Claude Code, Cursor, v0, and Lovable will all look dated soon enough. The work underneath them will not.

Say precisely what you want. Check that you got it. Answer for what ships.

That is the skill we are teaching.

It matters because information has never been the real bottleneck. Finishing is. A six-year study of Harvard and MIT open courses  found that, by its final year, about three percent of enrolled learners earned a certificate. The courses were not the problem. Most people had no deadline, no one waiting for them, and no cost to disappearing.

A cohort changes that, and the same research measured it. Learners who paid for a verified certificate — money down, intent declared — finished at fifty-nine percent, against five percent for everyone else. Same platform, same courses. What changed was the stake. That is what a seat buys: a deadline, a real product to ship, and people who notice when you go quiet.

Across five cohorts, twenty-eight builders started and twenty-eight finished. Every team launched by Demo Day. Every product they built is on /shipped, and most are still live and clickable.

We called it exp.club because experience was always the point: real problems, real users, and something you actually shipped. AI changed how much of that experience one person can compress into six weeks. The point never moved.

The only thing that reliably gets anyone across the line is a line, with people standing at it.