Six Apps in Six Weeks, Zero Euros: The Honest Retrospective

There's a number I kept leaving out of every version of this story: €0.

That's the combined all-time revenue of the six apps I built in six weeks. Zero euros. Zero paying customers. Every draft of this retrospective I wrote before this one found a way to bury that number under lessons and frameworks. This version starts with it, because it's the only number I have — and because it turned out to be the most useful thing the whole experiment produced.

What I set out to do

Earlier this year I built six exam-prep apps for the French market: citizenship exam prep (CitoyenApp), electrical safety certification, AIPR, food hygiene (HACCP), first aid (SST/PSC1), and the TCF/TEF French test for immigration candidates. Plus a seventh, driving theory, for good measure.

Real niches. People take these exams every day, they stress about them, they search for prep material. The plan was simple: AI has made building software absurdly fast, these niches are underserved, so build for all of them at once and see what sticks.

Main tool: Claude Code. Stack: Next.js, Vercel. Everything deployed live, everything functional. Quiz engines, spaced repetition, progress tracking, payment links ready to charge.

What actually worked

I want to be precise here, because this part is real and I don't want the €0 to swallow it.

The building worked. Six functional products, solo, in weeks. Five years ago this would have needed a small team and most of a year — not just for the code, but for the UX, the content structure, the quiz logic, the state management. AI absorbed almost all of that repetitive technical load. I made product decisions; the machine made components.

When people ask me whether AI coding tools are overhyped, I point at this: a hotel guy who teaches marketing shipped seven working apps in his spare weeks. The build revolution is not hype. I lived it.

That's the good news. Now the uncomfortable part.

What €0 means

It would be comfortable to blame the products. They have rough edges, sure. But that's not what the zero is telling me.

The zero is telling me that I solved a problem nobody was blocking on. Building was never my bottleneck — I just hadn't noticed, because building used to be everyone's bottleneck, and I inherited that assumption without checking it.

Shipping software is not the same as getting customers. I knew this the way you know statistics about other people. Six live apps and zero sales later, I know it the way you know a burn.

Nobody is going to find an exam-prep app because it exists. The person preparing their naturalisation interview is in a Facebook group asking strangers what to revise. The person who needs an HACCP certificate is asking their employer or their training centre. My apps are not in either conversation. They're just... online. Being online is not distribution. It's the absence of distribution with good uptime.

And there's a second layer under that one: trust. Even if those people found my apps, why would they pay a site they've never heard of to prepare an exam that decides their citizenship or their job? Code doesn't answer that question. A person does. A recommendation does. A track record does. I have none of those yet in these niches — and no amount of shipping adds them.

The actual lesson

AI collapsed the cost of building. It did not touch the cost of being trusted by a stranger.

Which means the moat moved. When building was expensive, building was the filter. Now anyone can have six apps — I'm the proof. What's scarce is everything I skipped: being in the rooms where the customers already are, having someone vouch for you, earning the first ten users one conversation at a time.

I built six apps to avoid doing that, if I'm honest. Building is comfortable. The feedback loop is instant and the machine never says no. Distribution means talking to people who might ignore you. I picked the comfortable work and called it a strategy.

What I'm doing about it

Not building app number eight. That's the whole point.

The plan now is embarrassingly analog. Go where the users already talk: the naturalisation Facebook groups, where people ask the exact questions CitoyenApp answers. Contact driving schools directly instead of waiting for their students to stumble in. Get one real person to use one app, watch what they do, and earn the first paying customer as a milestone in itself — not as a rounding error on the way to scale.

One app, one channel, real conversations. The other five stay live; they cost nothing to keep. But they only get attention again once one of them has proven that someone will pay.

Six apps in six weeks proved I can ship. It proved nothing about whether anyone wants what I shipped. Those are different questions, and I spent six weeks answering the wrong one very efficiently.

If you're a builder staring at these AI tools right now: take the speed, it's real. Just know that the part of the game AI made easy is no longer the part that decides whether you win.