The Complete
Story
What actually happened — and what it means
The article we published earlier got the facts right and the story wrong. This is the account of the redesign, the onboarding, and the way the work actually gets done.
The facts were right.
The story was incomplete.
"Non-technical founder uses AI to finish what developers couldn't" is a human-interest story, and an incomplete one. Across three years in Mississauga — a war, a cold basement, a prayer mat, and a production server in a conflict zone — the work was more specific than that framing.
This account sits next to Built From the Rubble and We Stopped Teaching. We Started Doing. Those essays hold the episodes. This one holds the judgment underneath them.
"The AI did not rescue the platform. The architectural redesign made the platform rescuable by AI. Those are different things, and the difference matters."
The rebuild was not a rescue.
It was a redesign.
When the development team went offline in late 2025, the usual move is to hire a replacement and finish what was started. That is not what happened.
The original architecture was a proper microservices stack: ASP.NET Core 9, PostgreSQL, RabbitMQ, Vue.js dashboards, thirty-six normalized tables, services split across boundaries. It was built for a human team that could hold the model, trace calls, and debug across those boundaries. When that team disappeared, the architecture became a liability — not because it was broken, but because it was the wrong structure for what came next.
The question was not how to patch the codebase. It was what kind of codebase can run without a permanent engineering team. The answer was fat controllers.
Fat controllers are an anti-pattern for human maintainers: too much logic in one place. The critique flips when the developer is an AI agent. Agents stay coherent when a whole workflow sits in one execution scope, and they lose the thread across service boundaries. A fat controller is that scope. The agent can read it, change it, test it, and ship it without losing the path.
The pivot was deliberate. Microservice boundaries were consolidated. Two Vue.js 2.6 dashboards were decommissioned for a single React 18 / TypeScript / Vite frontend. Fragmented API responses became one mandatory envelope — the same wrapper on every endpoint, every time. Permission lookup tables were flattened. What had been navigable for humans became operable for agents.
Then the rebuild: twenty-seven days, and zero critical defects in production.
"Seven hours at a time on a picnic chair in a cold basement in Mississauga, feet wrapped in a prayer mat against the concrete floor. No salary. No team. No guarantee it would work."The sessions where the redesign actually happened
The onboarding was not a feature.
It was a philosophy.
Once the platform was running, the next problem was partner activation. The usual fix is better documentation: cleaner guides, longer tutorials, more tooltips. That asks a busy partner to read, apply, and stay patient before they have seen the product work. Most will not. They need proof before they can sell it internally.
The question was not how to make the documentation clearer. It was what happens if a partner does not need documentation to start. The 60-Second Path is four steps from signup to a first successful API transaction, so they can prove the system works before they have to understand it.
The path is scored on four time-to-value metrics.
Three failures surface immediately, with a next step instead of a cryptic error. Fast diagnosis is part of the path, not only the happy path.
"The barrier to a signed partnership is rarely the commercial terms. It is the internal conviction that the thing actually works."Fifteen years of deals stalling — Sudan, Abu Dhabi, Dubai
That lesson came from fifteen years of deals stalling in Sudan, Abu Dhabi, and Dubai. The onboarding makes the partner a reviewer rather than a builder: do the first part of the work, and let them see the result.
Calling the platform "live and stress tested" is true, and it is not the point. The point is proof in the partner's hands — a first successful API transaction in under sixty seconds, with no documentation required to start.
Not a developer.
A precise description of the gap.
The redesign and the onboarding are how I actually work. I am not a developer. I cannot write the code. I can say what a system has to accomplish, name the structure that would allow it, and work backward from the outcome. That is the fat-controller pivot. That is the 60-Second Path. Neither required a line of code from me.
The pattern is the same in every role: find the gap between what exists and what should, describe the fix precisely enough to build or commission it, then move to the next gap.
Steller is the fullest version of that pattern. B2B companies had no clean infrastructure for digital gift card programs. Closing the gap meant choosing fat controllers over microservices, then directing AI agents through the rebuild without being able to read the code they were writing.
The result is a live platform, stress tested to zero critical defects, with an active API integration for Bede — the digital wallet of Zain Sudan, one of Africa's largest telecoms — and a first successful partner transaction in under sixty seconds.
That is the story the original article did not tell.