Ship one bounded AI build in four weeks, with review every week.
One build, scoped on day one. A live build session and a clinic every week. A deployed system with a README someone else could run. Application-based. Founding cohort $1,497 (list price $1,997).
Who it’s for
Technical operators, developers, consultants, and agency builders who have a real problem from their work or a client’s work, access to the system or data it involves, and a specific thing they want to automate or an agent they want to ship. Comfortable in a terminal or a no-code automation tool, or able to read and modify code.
What you get
- One build, bounded. A single trigger, a single output, a single user. Week one exists to cut your project down to that, because the most common failure in builder programs is a project that’s three projects.
- A live build session every Wednesday, 11:00 Central. Dan builds the week’s step on a reference project from a blank state, on camera, narrating every decision, including the ones he abandons and the things that break.
- A clinic every Friday, noon Central. You show your build running, or failing. Dan reads the code or the flow with you and names the one change that matters most.
- A failure list before demo day. Week three is about the ways it fails: bad input, tool errors, refusals, hallucinated fields, cost, and what you did about each.
- A deployed build, a README, and a demo. Week four ships it somewhere other than your laptop and hands it to someone who could run it.
The four weeks
| Week | Wednesday build | Friday clinic | By Friday you have |
|---|---|---|---|
| 1 | Scope and stand up | Scope review | A scope sentence, an environment, a first successful call |
| 2 | The core path | Debugging | The right output on three real inputs |
| 3 | Survive contact: evals, failures, guardrails, logging, review points, cost | Failure clinic | A failure list with fixes, ten real inputs run |
| 4 | Ship: deploy, README, handoff, what to monitor | Demo day | A deployed build and a certificate |
What it costs
$1,497 (list price $1,997) a seat, by card, after your application is accepted. The founding cohort is priced below the list price because it’s the first run and we’ll ask you for feedback every week. EdgeCore members and EdgeCorporate participants attend through their programs.
What it is not
It isn’t a course with a curriculum you consume. Every week is your build. It isn’t platform-locked: Claude Code and the Claude API are the default path, and builders on Make, n8n, or another agent framework are welcome. It isn’t for client data in personal accounts; business-tier accounts are required, and the terms say so.
Where this comes from
This cohort hasn’t run yet, which is why the first one is application-only and priced as a founding run. The outcome it’s built to produce we’ve watched happen once, slowly: an operations lead at a real estate company with no developer background who, over four months of weekly review, moved her project from a no-code tool to a real application, deploys through GitHub, runs the tests through Claude, and now ships every week. Four weeks is that loop, compressed.
Apply
Eight questions, ten minutes. Dan reads every application and replies within three business days with one of three answers: you’re in, not this run and here’s the specific thing to sort out first, or this isn’t the right program and here’s the one that is.