Private AI Coding Workshop
Ten private sessions over two weeks, on your own requirements. Your people learn the loop we use on every project: write down what should change, let Claude Code build exactly that, have an independent reviewer check it, and decide what ships.
Generated code can look right, pass its own tests and still do something you never asked for. The workshop gives your team a habit that catches it before it ships, and a record of every decision along the way.
The hands-on form of The Architect™
You write what should change and how you will know it is right, and sign it before any code is written.
Claude Code builds exactly what you signed, nothing more. A guard stops code being committed before a signed definition exists.
An independent reviewer, running on a model from a different company, checks the work against what you signed and twelve engineering rules.
You approve, ask for a fix, or accept with a written reason. Every decision is recorded in your own words.
Developers and business owners learn the same loop from different seats. Each workshop runs one track.
Hands on Claude Code from session 2, on their own machines. They write the definitions, run the build and the review, and by session 9 run the whole loop alone, against the clock.
No coding background needed. Owners agree a clickable mockup of each screen, sign what gets built and make every decision. From session 7 they take the keyboard themselves.
Every session ends with something you can see. The hands-on work runs on three requirements you bring, not a practice project.
A full loop live: the guard blocks a commit and the review catches a flaw the screen never showed.
Choose the first three and rank them by risk. Developers set up the kit on their machines.
Make every criterion something you can check. Owners start from an agreed mockup.
Claude Code implements the definition; you read the independent review section by section.
Approve, fix, accept with a written reason, or start again, and record it.
A bigger requirement, split into definitions, with the review tier chosen by risk.
Your people at the keyboard end to end; we watch and take notes.
Your own stumbles so far, and the classic ones, rewritten so they would have passed.
Requirement three, start to finish, without us.
Walk your project record, see how updates arrive, and leave with your next requirements drafted.
You will need Claude Code and an OpenAI API key on your own accounts. Your code goes to those two services under your own terms with them, never through ours.
Per workshop
$12,500
Paid in full when you book. Sessions can each be moved once with 48 hours' notice.
After the 90 days
Optional support at $1,000 a month: kit updates and email questions answered within two working days. Cancel any time; without it you keep the kit you have.
We will talk through what you want to build, which track fits, and dates for the ten sessions.