Locked In
1:1 AI Training
Four weeks. One app, shipped.
You have the idea. You have the tools.
What you probably do not have is four weeks where shipping the thing becomes the priority.
Locked In is a four-week, one-on-one build sprint.
We meet twice a week for 45 minutes. You get a private Slack channel with me between sessions. We take one idea, scope it properly, build it with AI, and get it into the real world.
Not a tutorial.
Not a prototype you abandon in a folder.
Not another app that looks like every other app built with AI.
A real product that someone other than you can use.
You do not need to know how to code. You do need to be willing to build.
And before we touch the code, we get painfully clear on what we are building.
Because after building a lot of products with AI, I have learned something the hard way:
The quality of what AI builds is mostly decided before you ask it to build anything.
Only a few spots run at a time. When they are full, new applications join the waitlist for the next opening.
- 4 Weeks
- 1-on-1
- Application Only
Most apps built with AI look like it.
You know the look.
The same gradient. The same three feature cards. The same empty dashboard nobody opens twice.
AI is very good at producing something shaped like an app. It is much worse at producing something a person actually wants to come back to.
Working is the low bar. The distance between working and worth using is design, and it is most of what separates a demo from a product.
I did product design and UX long before I did any of this with AI. Clean Pro, my HubSpot theme, has been the number one theme in that ecosystem, and that does not happen because of a feature list. It happens because of a thousand small decisions about how something looks, reads and behaves.
So the specification we write in week one is not only about what your app does.
It covers what it looks like. What it says. What happens on an empty screen, a slow connection, a wrong password. Which of the forty features you imagined actually earn a place, and which ones quietly make the product worse.
That is the part an agent will not decide well on its own. It has no taste, and no opinion about your user.
You should want to send someone the link.
Starting is easy now. Shipping is still hard.
AI has made it ridiculously easy to start building software.
That is not the same thing as making it easy to ship good software.
Open Claude Code or Codex with an idea in your head and start prompting. It feels incredible for the first few hours.
Then the app gets bigger.
Something that worked Tuesday breaks Thursday. You solve the same problem three different ways. The agent starts making decisions you never asked it to make. You fix one thing and quietly break another.
By week three, you are spending more time fighting your own app than building it.
The problem usually is not the agent.
The problem is that the agent never had a clear definition of what it was supposed to build.
You asked eleven times, slightly differently, and never wrote any of it down.
I have become obsessive about this.
Before I build, I write.
What does this product actually do? Who is it for? What is the data model? What are the screens? What happens when they are empty, loading, successful or broken? What stack are we using? What should the agent build first, second and third?
Then I break the work into pieces small enough that an agent can complete each one without having to invent the missing details.
It feels slower.
It is much faster.
Locked In is the process I use to build my own products, applied to yours, with me beside you while you do it.
The four weeks
We meet twice a week for 45 minutes.
Why twice?
Because when you only have four weeks, getting stuck for seven days is expensive. Two shorter sessions means you are never more than a few days away from getting unstuck, making a decision or correcting course.
You should plan on another 8 to 10 hours each week to actually build.
This is not something you watch.
You are going to make the thing.
-
Week One. Write the spec, not the code.
Get painfully clear before the agent starts building.
We start by deciding exactly what you are shipping.
Not the app you could build someday. The version we can actually get into someone's hands in four weeks.
Then we create the documents the agent will build from:
- What the product does
- Who it is for
- The core user journey
- The data model
- The screens
- Empty, loading, success and error states
- The technical stack and why we chose it
- The rules the agent should not break
- A build plan broken into small, clear tasks
If your scope is too big, I am going to tell you.
If a feature does not need to exist yet, we cut it.
If there is a simpler way to get the same result, we take it.
You end the week with a build-ready spec and a clear path from empty repo to shipped product.
-
Week Two. Build one path all the way through.
This is where the product becomes real.
Now the agent starts building.
But we do not ask it to build everything at once.
We build one complete vertical slice.
A real user can do the main thing your product exists to do. It talks to the real database. The data persists. The important pieces are connected.
No fake demo flow that only works when nobody touches it wrong.
This is also where you learn the build loop I use every day:
- Give the agent a clear task.
- Let it work.
- Check what changed.
- Test it.
- Catch bad assumptions before they spread.
- Correct course without blowing everything up.
- Then move to the next task.
You end the week with a working product you can actually click through.
-
Week Three. Build the unglamorous 80%.
Turn the build into a product.
This is the week that separates “I built an app with AI” from something you can actually give another person.
Depending on what you are building, that means things like:
- Authentication
- Real user data
- Permissions
- Payments
- Empty states
- Error handling
- Security rules
- Analytics
- Deployment
- TestFlight
- A real domain
- The thousand small things a screenshot never shows
This is also where solo builds tend to stall.
The fun part is over. The app mostly works. There are fifty boring things left.
Good.
We finish them.
You end the week with something you would be willing to hand to a stranger.
-
Week Four. Ship it.
Get it out of your hands and into someone else's.
This week is about the line between building and shipping.
We finish what matters. We cut what does not. We get the landing page up. We connect analytics. We deploy the web app or prepare and submit the iOS app for review.
Then we get another human being into it.
Not you.
Not your agent.
A real person.
They will find things you missed. We fix the important ones.
Our final session is partly retrospective and partly about what happens next, because I do not want the project to die the moment our four weeks are over.
You end the sprint with a shipped web product or an iOS app submitted for release, plus a clear plan for what comes next.
What’s included
- Two 45-minute working sessions every week
-
Eight private sessions across the four-week sprint.
These are working sessions.
We screen share. We look at the product. We make decisions. We solve whatever is stopping you from moving.
- Private Slack access for four weeks
-
When you get stuck between sessions, you do not have to sit there until our next call.
Send me the problem, the screenshot, the prompt, the error or the decision you are wrestling with.
Slack is async. I am not sitting in it all day, but I will keep you moving.
- The same spec system I use for my own builds
-
Not a generic “build an app with AI” PDF.
You get the actual structure I use to give coding agents enough context to do good work.
- Product and code review
-
I will look at what the agent is producing with you, help spot bad decisions early, and help you figure out what needs fixing before the problems compound.
- Session recordings
-
Every session is yours to keep.
- Your post-sprint plan
-
You leave knowing what to work on next instead of waking up Monday wondering what happened to all the momentum.
What you’ll need
There are three things I cannot give you.
-
1. Claude Code or Codex
Budget around $100/month.
This is what does the majority of the building.
Either can work. I currently spend most of my build time in Claude Code, so that is where I can usually get you unstuck fastest.
-
2. A small budget for third-party services
Usually $20 to $50/month to start.
Hosting. Database. APIs. Email. Analytics. Whatever your specific product needs.
A lot of these services have generous free tiers, so your actual cost may be lower.
If you are building for iOS, you will also need an Apple Developer account.
-
3. Eight to ten hours a week
For four weeks.
This is the big one.
Our calls are not where your app gets built.
You build it between the calls.
If you cannot protect that time for four weeks, do not apply yet.
What can you build?
-
AI agents
Products and systems that actually do work. Tool use, workflows, memory, evals and the plumbing that makes an agent useful more than once.
-
iOS apps
Native apps built in Swift, tested through TestFlight and prepared for App Store submission.
-
SaaS web apps
Authentication, databases, subscriptions, deployment and all the boring infrastructure that turns an idea into an actual product.
Not sure what you should build?
That may be the first thing we figure out.
Who Locked In is for
This is probably for you if…
-
You want to build the thing yourself instead of hiring someone to disappear with your requirements.
-
You do not know how to code, or you know just enough to be dangerous.
-
You have 8 to 10 hours a week that you can actually protect.
-
You have an idea, even if it is still rough.
-
You have started projects before and watched them stall.
-
You want to ship something, not spend four weeks studying how people ship things.
-
You are willing to cut features.
-
You are fine with me telling you when your scope is ridiculous.
-
You want to leave knowing how to build the next thing without me.
This is probably not for you if…
-
You want me to build the app for you.
-
You cannot make the time.
-
You want to learn “AI” as a general subject.
-
You want a course you can watch whenever you get around to it.
-
You are looking for a technical co-founder.
-
You want me to take equity instead of getting paid.
-
You mostly want someone to tell you your idea is great.
-
You are not actually ready to build anything.
$3,000. Four weeks.
That is the whole number.
No upsell waiting for you in week four.
No twelve-month mastermind.
No surprise pitch for the “real” program once you are inside.
You pay $3,000. We spend four weeks getting locked in on one product and pushing it into the real world.
For context, that is less than you can easily spend on a contract developer before they have really learned your project.
The bigger difference is what happens afterward.
A developer can leave you with an app.
I want to leave you knowing how to build the next one.
You will still need your coding-agent subscription and whatever third-party services your product requires. Those costs are separate.
Before you apply
Do I need to know how to code?
No. You need to be able to explain what you want, follow a process, test what the agent gives you and keep going when the first attempt is wrong. AI can write an enormous amount of the code. Your job is to learn how to direct it.
What if I do not have an idea yet?
If you have a few directions and need help choosing, that is workable. If you have absolutely no idea what you want to build, I would wait. Four weeks is short. I do not want to spend half of it searching for a reason to build something.
How much time does this actually take?
Plan on 8 to 10 hours a week outside our sessions. Some weeks may take more. If you only have an hour or two available, this is the wrong sprint.
What if my app is not finished in four weeks?
Then we made the scope too big, hit a genuine technical problem, or did not put in the required time. The goal is to prevent all three. For web products, the target is a live product that real users can access. For iOS products, the target is a release-ready build submitted to Apple. Apple's review timeline is outside either of our control.
What else do I need to pay for?
Usually a Claude Code or Codex subscription, hosting and database services, any paid APIs your product requires, and an Apple Developer account if you are shipping on iOS. For most early products, the software costs are small.
Will you write the code for me?
No. I may jump into a problem with you, review what is happening, help with prompts, inspect code and show you how I would approach something. But this is not done-for-you development. The entire point is that you become capable of building without me.
What tools do we use?
That depends on what you are building. I am not going to force your product into a predetermined stack because it makes the course easier to teach. We choose the stack in week one and document why.
How is this different from Future Proof?
Future Proof is strategic AI transformation for founders and companies. Locked In is much narrower. One person. One product. Four weeks. Ship it.
What happens after I apply?
I read every application personally. If I think Locked In is a good fit for what you are trying to build, I will reach out with next steps. If I do not think we can make meaningful progress in four weeks, I will tell you.
Four weeks from now, you have either shipped something or you haven’t.
The difference is rarely access to another AI tool.
And it is usually not the idea.
- It is scope.
- Clarity.
- Consistency.
- Having someone there when you get stuck.
- And having a date that does not move every time life gets busy.
If you want that date to be four weeks from your first session, tell me what you are building.
Only a few spots run at a time. When they are full, new applications join the waitlist for the next opening. Every application is read personally, and every session is one-on-one.
Tell me what you’re building.
11 questions, one at a time. Answer honestly and you’ll get a straight answer back, including if that answer is “not yet”.
1 of 11
Question 1 of 11