Build · Platforms and apps
I build the platforms people log into: communities, e-learning academies, SaaS products, mobile web apps and client portals.
I used to be the CTO of a US marketing agency, where I led a team of 12 developers. Now I build on my own, with AI as my team. I kept the agency habits: a written spec, one change at a time, automated checks, and nothing goes live without your OK.
The usual failure
One pattern from my CTO years, anonymised: a membership platform built around family accounts. One parent pays, several family members each have their own linked profile, and each profile sees different things. On a feature list it looks like a detail. In practice it decides the data model, the checkout and every permission rule. Settled in week one, it’s a line in the spec. Discovered after the database is built, it means rebuilding the foundations.
That’s why my first step asks who pays, who logs in and who sees what, before a single screen exists. AI makes this more important, not less: one giant prompt gives you a demo that looks finished, with nothing underneath that a second developer could pick up.
Diagram: one parent pays. Several family members each have their own linked profile, and each profile sees different things.
Twenty features in version one, so none of them get tested with real people.
I start with the one flow your users would pay for, and prove it first.
It looks great on day three and falls over on day thirty: no data model, no permissions, no tests.
I break the build into stages and disciplines, the way an agency staffs a project.
Changes arrive by voice note, nothing is written down, and the platform drifts.
I keep a written spec and a dated decision log, and I quote your words exactly.
What I build
Pick the one closest to your idea. Each card shows the core a first version usually needs, and the part of my background that’s most relevant to it.
For creators and associations who’ve outgrown a group chat.
Background As CTO I worked on membership systems, including one built around family accounts with linked member profiles.
01A first version usually needs
02Asked before a single screen exists
03Underneath, whatever the type
For experts selling a method, and for schools and companies training people.
Background A multi-school learning platform in three languages, and a course → assessment → certificate → purchase journey for a certification brand.
01A first version usually needs
02Asked before a single screen exists
03Underneath, whatever the type
For founders turning a service or a spreadsheet into a product.
Background I’ve built and shipped two software products.
01A first version usually needs
02Asked before a single screen exists
03Underneath, whatever the type
For daily-use tools, bookings, trackers and companions.
Background Misclo, my own app in development, is built and checked at phone, tablet and desktop sizes.
01A first version usually needs
02Asked before a single screen exists
03Underneath, whatever the type
For service businesses that run their clients through email threads today.
Background A private intranet with a training academy that onboards every new team member.
01A first version usually needs
02Asked before a single screen exists
03Underneath, whatever the type
Something else? Most custom platforms are a mix of these five. Bring it to the call.
What you get first
A platform you can’t maintain without me is a liability. Everything I build is handed over so that any developer, or any AI, can carry on.
Here’s the shape of the first page of a Prototype Sprint spec. The project is fictional.
Proof from the lab · in development
Misclo is my own product, set to launch on 1 January 2027 and in staging today. It’s the lab’s live test of the method I’d use on your platform. I make the product decisions and approve every step. Claude writes the interface, the code, the tools and the tests. ChatGPT with Higgsfield makes the film and images. A second AI coding assistant, Codex, works in parallel on a separate part of the code.
Every version lives in its own folder, with a written report and a test run. The previous version stays untouched as the rollback until I approve the next one.
Proposed: Misclo phone screen: a late meeting leads the AI companion to propose a shorter home workout and a quicker dinner, waiting for the user’s OK. Accepted: The same Misclo screen after the change is accepted, with an Undo link.
Status: staging, not public, no users. This is proof of how I build, not proof of results. Character art and film are AI-generated under my direction.
Before AI was my team
For years, my job was the technical side of client platforms: audits, architecture, choosing the stack, running the build and keeping it alive after launch. As CTO, my team delivered and maintained 28+ client platforms: online academies, membership communities, e-commerce and nonprofit systems. AI changed who types the code. It didn’t change what a platform needs.
Built with AI, reviewed by me
Working this way changes how fast I can try an idea, how cheaply I can throw a version away, and how much one person can carry. It doesn’t change the parts that make a platform trustworthy, so those stay with me.
How I work
A good agency puts a strategist, a UX designer, developers, QA and a launch lead on a platform. I run the same disciplines with AI, in a fixed order, and I show you each stage before it replaces the last.
We agree on the one problem version one solves, who it’s for, who pays, who logs in, who sees what, and what “working” means.
Data model, roles and permissions, the stack, environments and the accounts you’ll own. The stack is chosen here, with the reasons written down: a configured open-source community and course stack when your idea fits one, custom code when it doesn’t.
One flow at a time: screens, logic, payments, the AI feature. Each change is shown to you first, with the previous version kept as the rollback.
Test scripts, accessibility checks, deploys, backups, and the admin jobs you’d otherwise do by hand.
Launch, watch how real people use it, fix and refine.
The disciplines
Engagements and prices
Every platform I build goes through the same path. You can stop after any step and keep what has been built.
Platform / App
From $1,200
A working prototype of your core flow, ready to put in front of real users before you commit to a full build. You bring five people from your audience; I give you the test script and we review the sessions together.
Platform / App
From $4,000
The first real version: accounts, one core workflow end to end, payments, an admin dashboard and one AI feature, such as an assistant that answers from your own content. Deployed, tested, and yours.
Platform / App
Quoted after scoping
Typical range $7,000–$15,000
Multiple roles (members, instructors, admins, clients), integrations with your CRM and tools, more AI features, and migration from what you use today. Quoted after the prototype: you get the exact figure in writing before work starts.
When the build is done you can take everything in-house, hire another developer, or keep me on. The handoff file is written so that any of the three works. Ongoing care is available as a monthly AI Operations Partner plan; scope and price confirmed on the call.
Prices in USD, starting points for a defined scope. You get the scope, timeline and price in writing before anything starts. Running costs for hosting, database, email, payments and AI usage are separate and billed to you directly in your own accounts.
What you get
In a repository under your account, with the history of every change.
For hosting, database, payments, email and the AI service, in your name and on your billing. I work as a guest.
What we decided, when, and why, with your words kept exactly as you said them.
And the last report: what was checked, what passed, what didn’t.
How to run it, how to deploy it, where the risks are, and what to do next.
FAQ
It can be, if the AI is left alone. Mine isn’t. Every change goes through a spec, my review, automated checks and your approval, with the previous version kept as the rollback. You can read the test report and the code at any point.
You do. The repository, hosting, database and payment accounts are in your name from the start, and I work in them as a guest.
That’s why everything is written down. The spec, the decision log and the handoff file let another developer, or another AI, carry on without me. That’s a habit from agency life, where projects had to survive people moving on.
Often, yes. If an existing tool covers most of what you need, it’s cheaper and faster, and I’ll tell you so on the call. There’s also a middle path: a configured open-source community and course stack that you own and host, with more room than a hosted tool and less to build than custom code. Fully custom makes sense when your method, your data or your business model doesn’t fit either box.
We settle it in the first two steps: Step 1 answers who pays, who logs in and who sees what, and Step 2 picks the stack against those answers and writes the reasons into the architecture note. For communities and academies that fit a proven shape, I configure an open-source stack: WordPress with BuddyBoss for the community and LearnDash for courses, the stack I specialised in during my agency years. If your sales already run on GoHighLevel, the platform connects to it. When your model or data doesn’t fit that box, such as a SaaS workflow, logic shared across several organisations, or an AI feature at the core, I write custom code. Either way the code and accounts are yours.
Third-party fees are paid by you, directly, in your own accounts: hosting, database, email, payments and the AI service. I estimate them in writing in the Step 2 architecture note, before the build starts. The AI feature gets a monthly usage cap and an alert, so a busy month can’t surprise you.
It depends on the step and the scope. You get the timeline in writing with the scope and price, before anything starts.
I build mobile web apps: they open from a link, can be installed on the home screen, and skip app-store review. If you truly need a native app, we’ll talk about it on the call.
Each user sees only what their role allows. Passwords and keys live in the host’s secure settings, never in the code. Payments go through a hosted checkout, and I test in test mode until you approve going live.
Usually one, done well: an assistant that answers from your own course or help content, a smart onboarding step, or a feature that turns your users’ input into a draft plan they can accept or change. Before it ships, the assistant is tested against a written set of real questions, must cite the lesson it answered from, says when it doesn’t know, and has a monthly usage cap. I’ll push back on AI features that don’t earn their place.
Yes. I’ll work from your designs or alongside your developer, and the handoff file is written for them.
Start with the free tools in my library, or join the free AI Business Build Week.
You tell me what you want to build and for whom. I ask a few questions about your users, your budget and your timing. You leave with one of four answers: build it yourself (I’ll point you to the free tools), start with the $970 Audit, get a direct proposal, or it’s not a fit and I’ll say why.
Next step
Have it built
Bring the idea, the spreadsheet or the half-built thing you’re stuck with. I’ll tell you honestly what the right first step is, even when it isn’t a build with me.
Free. No pitch deck needed. Possible outcomes: do it yourself, the $970 Audit, a direct proposal, or not a fit.
Start on your own
The Solo Builder Blueprint is my free guide and worksheet. It follows the same five stages as this page: Choose → Set up → Build → Automate → Improve. No signup. The other free tools are in the library, ungated.
Prefer email? contact@labgenz.com