More and more people write to me with the same story: "I built an app in Lovable / v0 / Bolt / ChatGPT. It looks exactly how I imagined it. But I can't launch it." Good news: it's not your fault, and your work isn't wasted. Even better news: the distance from prototype to product is shorter than you think — if you know exactly what's missing.
Why AI prototypes stall at 80%
AI tools are excellent at the visible part — screens, buttons, copy. They get stuck on the invisible part:
- Accounts and authentication — real users, passwords, resets, sessions that don't get lost.
- Real data — a demo holds "toy" data; a product needs a secure database with backups, hosted in the EU.
- Payments — integrating a payment processor, invoices, what happens when a payment fails.
- The weird errors — it works on your machine, breaks for your first customer, and the AI keeps giving you the same answer that fixes nothing.
- Hosting and a domain — so the app lives on the internet, reliably, on your own domain, not in a preview tab.
Your prototype isn't a failure — it's an excellent specification. It shows me exactly what you want, better than any document.
What "in production" actually means
To judge for yourself how close you are, this is the checklist I run before any launch:
- Your own domain — the app lives at your address, not on a preview subdomain.
- A real database — customer data is stored, protected, and backed up automatically.
- Accounts that work — sign-up, login, forgotten password, with nobody getting stuck along the way.
- Payments tested end to end — including the ugly cases: declined card, abandoned checkout.
- Errors that are watched — when something breaks for a customer, you find out before they do.
- Decent speed on a phone — that's where your customers use it, not on your laptop.
If you tick fewer than half, you're at the demo stage. That's normal — it's where almost every AI prototype stops.
Where each tool gets stuck
From the prototypes that have come through my hands, each tool has its typical weak spot:
- Lovable — closest to a real product, but on accounts, permissions and third-party integrations you quickly hit "same error, try again".
- v0 — makes gorgeous screens, but they're often just the facade: there's no real data behind them, only sample content.
- Bolt — great for fast prototypes, but the project gets fragile as it grows; one small change breaks three things.
- ChatGPT / Claude — writes good code in pieces, but the pieces don't hold together on their own: you're left responsible for assembly, hosting and debugging.
No tool is "wrong". They all get you to validation fast — exactly the job of an MVP. It's just that the last 20% is a different kind of work.
What NOT to do
- Don't start over. The design, the flows, the copy and the decisions you've made are valuable work, already done.
- Don't keep prompting in circles. If the same error comes back a fifth time, the problem isn't your prompt — you've hit the tool's limit.
- Don't accept "everything must be rewritten" at face value. Sometimes a part gets rewritten, rarely all of it. A big quote that starts that way deserves a second opinion.
What it costs and how long it takes
With me, the first working version — your prototype taken to production — is free. You see the result before any money. If you want to go further, most projects land between €500 and €2,500, agreed clearly before I start, and a typical first version is ready in 1–3 weeks. I've also written about what an app costs for a small business if you want the detailed numbers.
The questions everyone asks me
Do you keep what I built, or throw it all away? I keep everything that's healthy — often the design and flows stay, and I build the plumbing behind them.
Do I need to understand the technical side? No. You show me the prototype and tell me what it should do; translating that into technical terms is my job.
Whose app is it in the end? Yours. Once a project is paid, the app and its code belong to you, with no forced subscriptions.
Next step
Send me the prototype link (or a few screenshots) and I'll tell you honestly what's missing and how long it would take. I built FastOrder and Beez — products used by hundreds of thousands of people — and that's exactly what I do for others now: I take prototypes to production. The first version I build for free.