From Blank Folder to Running System

The blank folder is where a lot of this dies. You have watched the demos, you have probably installed something already, and then one Tuesday evening you open an empty folder and there is genuinely nothing to type, because the demo was solving somebody else's problem and yours does not look anything like that.

Six of the nine people in my founding cohort wrote almost exactly that on their application form, in their own words, none of them knowing anybody else had said it. So you are in fairly normal company here. There is a real gap between watching something work on a screen and making it work on your own machine, and these two sessions exist to close it.

Four hours in total, split into two sessions of two hours each, one week apart. That week in the middle does most of the work.

Session one starts on a blank folder and ends with something running. Your build, your files, your machine, and me on the call while it happens rather than sending you away with a recipe. We go from empty to working, and then I make you run it once on your own while I watch, because that is the moment you find out what you actually understood.

Then you live with it for a week. You use the thing. It breaks somewhere, or it does something slightly stupid, or you think of the one feature that would make it genuinely useful on a Monday.

Session two is that list. We fix what broke, harden the parts that were held together with hope, and add the next layer, which is usually memory, a second agent, or wiring it into a tool you already pay for.

I used to offer this as one four hour block and stopped, because hour four is where everybody stops absorbing anything. Two afternoons a week apart is also far easier to actually schedule, which matters more than it sounds if you are in Riyadh or Bengaluru and I am not.

1. A working system in your own folder, which you built, and can therefore change later without me

2. The files that make it work. The instructions file, a command or two, a skill, and whatever context files your particular build needed, in whichever tool we picked together

3. Both recordings

4. A short written note from me after session two saying what to build next and what to leave alone for now

This one has a real setup list and I am strict about it, for a fairly unromantic reason. In my founding cohort we lost live minutes to installs and logins more than once, and in a group that is annoying, but in a paid 1:1 it is your money going on a progress bar.

The list itself gets decided on the free 30 minute call, because it depends entirely on what we are building with. If that turns out to be Claude Code you will need a Claude Pro or Max subscription. If it is Codex, an open model running on your own machine, Kimi or n8n, the list looks different, and I would rather give you the right one than a generic one you half need.

What does not change.

1. A laptop you have admin rights on. Mac, Windows or Linux, no tablet-only setups

2. Whatever we agreed on the call, installed and opened at least once without throwing an error

3. A folder for the thing we are building, even if it is completely empty

4. One build in mind, with a real input and a real output

On that last one, "something with AI" is not yet a build. "Take the messy notes I write after every client call and turn them into a follow up email that sounds like me" is a build, and we can start on it immediately.

If that fourth one is the bit you are stuck on, buy Your AI Automation Roadmap instead. That session exists to answer exactly that question, it costs less, and you will get much more out of these four hours afterwards.

If you are not yet sure what to build, start with Your AI Automation Roadmap at $497. Two hours, and you leave with a ranked plan rather than a half-built thing you are not sure about.

If you would rather learn all of this over six weeks with other people around you, and you like having a room to ask questions in, the Claude Code Builder cohort runs at $1,597 and covers considerably more ground than four hours can.

$997 is founding pricing and it goes up once there are reviews sitting on this page. I run 2 of these a month, which is honestly all the calendar has room for alongside the cohort.

Every build includes a free 30 minute call with me before session one. You do not pay anything extra for it.

We use it to agree two things. What we are building, and what we are building it with. That might be Claude Code, it might be Codex, an open model running on your own machine, Kimi, n8n, or a mix of them. I am not loyal to any particular tool, I have been wrong about several, and I would rather pick the one that fits your work and your patience for maintaining it than the one I happen to use most.

Your setup list comes out of that call too, which means you install one thing rather than guessing at five.

Once we have had it I set up both sessions and send you the invites, roughly a week apart.

If you would rather talk before you spend anything at all, the same 30 minutes is listed separately on my profile as a free AI discovery call.