The Honest Frame - Built with AI Playbook

Start here

I built a real software product with paying customers. I am one person, and I am not an engineer. I did it by directing an AI coding agent, day by day, and I kept the receipts.

This course teaches how. Not the hype version. The version where I tell you what the AI is genuinely great at, what it still cannot do, and the specific ways it burned me so you do not have to learn those the same way I did.

The one idea

AI changes the cost of execution. It does not change the need for judgment.

For most of business history, adding a capability meant adding people. Now a small team, or one focused person, can do what used to take a department. That is the whole shift, and it has nothing to do with the AI being magic. The AI writes the code. I decide what to build, who it is for, what to cut, and which of four reasonable next moves ships first. The leverage is real. It is also work. Six to ten hours a day of work.

If you take one thing from this course: the bottleneck moved. The cost of writing dropped to near zero, so your leverage is now entirely in deciding, reviewing, and integrating. Everything I teach is downstream of that.

The receipts

I am going to make you a promise the rest of this course keeps: I will not give you a number I cannot back. So here is the product, exactly as the dashboard shows it.

18 days from create-next-app to live with paying users.
A real, revenue-generating SaaS running today: a handful of paying customers, real monthly recurring revenue, a 25 percent free-to-paid rate, zero cancellations so far.
Total AI cost to run the entire product, all-time: about $27. Roughly $12 a month now. Around 43 percent of the input tokens come from cache, because caching is not optional when you are the one paying the bill.

Read that cost number again, because it is the most important sentence in the course. A live product, with real customers, run by a non-technical founder, for the price of a couple of lunches in AI spend. That is not a growth-hack. It is what disciplined AI-assisted development actually costs, and learning that discipline is most of what you are here for.

The objection, and the real moat

You should be skeptical right now, and here is the skepticism I respect most: if a non-technical founder can ship a real product in 18 days with an AI assistant, then so can everyone else. "I built it with AI" is not a moat. By the time you read this, it is table stakes.

That is correct, and it is the most important thing in this course. The hard part was never the building. The hard part is being found. The durable advantage is distribution, and it has almost nothing to do with the AI that wrote the code.

So the strategy that actually matters here was decided before most of the features existed: treat distribution as a system to build from day one. Personas written down in a file. Every feature shipping with a help article and a search-intent blog post. Every persona getting a landing page that names their specific pain. And then, the part that makes it believable rather than aspirational, all of it enforced by the build itself, so you cannot ship a feature and forget the growth work, because the deploy will not go out.

That flywheel, not the AI, is the answer to "anyone could do what you did." We give it a whole module, because it is the part most AI-build stories skip and the only part that compounds.

What this course is, and is not

It is not a "build a to-do app with AI" tutorial. It is the operating model of running a real product, taught through one that exists and is fully documented: the same repository, the same commit history, the same forty session logs, the same forty production scars.

The scars are the curriculum. The most valuable thing I can hand you is not the day everything worked. It is the day I took production down three times in one evening with the same bug, and exactly what I changed so it could not happen again. Almost every safety rail this product has was written the day after the thing it prevents shipped. That pattern, convention to miss to gate, is the honest shape of building with AI, and you will see it over and over.

Who this is for

Two kinds of people, one shared spine.

If you are a founder or operator who is not a deep engineer: you are me. I will teach you to direct, govern, and verify the work so you can ship a real product without a team.
If you are an engineering leader or builder adopting AI: the spine is the same, and the depth is here too. Parallel agents, review at scale, the approval ladder, the cost model, the production incidents. The rigor is what your team will respect.

I narrate it as the operator who learned to build, because that is the honest seat I sit in, and because it is the part most courses cannot teach: an engineer can write the code, but deciding what is worth building is a different muscle.

The honest caveats, up front

The numbers are small. This is an early product, not a unicorn. If you want a course about scaling to millions of users, this is not it. This is a course about going from zero to a real, working, paying thing, alone, for almost no money. The believability is the asset.
The AI cannot do everything. It writes correct application logic and then cannot debug your deployment, cannot see your dashboard, and cannot tell you which feature matters. You will learn exactly where the line is, because I crossed it many times.
This is not passive. A green build is not a finished feature, and the day you forget that is the day you ship a page that says "Saved" while saving nothing. I shipped that page. We will talk about it.

What you will have at the end

Not a folder of tips. An operating model: your own version of the project brief that keeps the AI on the rails, a build loop that ships small and reversible, an approval ladder matched to how much a mistake costs, a cost discipline that keeps the bill in the lunch-money range, and the judgment to know what to build and when to stop.

Let us get into it.

Sources for every claim in this module: content/origin-story/build-log.md, docs/curriculum/timeline.md, docs/curriculum/phase2-decisions.md, and the live admin dashboard. No figure here exceeds what those show.