OPNI: when the Vibe Coder forgets the Stop button

The AI-augmented developer is not always a super-developer. Without a framework or a Stop button, they mostly become a developer without brakes.

Since the Masterclass on Vibe Coding that I gave a few weeks ago for Scanderia, I have been receiving more and more messages that all say roughly the same thing:

"I built something solo in 3 weeks with AI. It is clean, it is full-stack, it is beautiful. But I need help for what comes next."

I even built myself a platform (using Vibe Coding, to measure the phenomenon) to let them book a meeting with me.

The OPNI phenomenon

And with every project I look at, I observe the same phenomenon, almost a pathology: when you can code without friction, code for the vibe, code because it has become too easy to keep going, AI has removed the cost of the line of code.

As a result, no one realizes that they can and must know how to say "Stop."

The result? Magnificent projects, over-architected, ultra-modern... and completely unsellable.

I call them OPNI: Objets Programmes Non Identifies (Unidentified Programmed Objects).

An OPNI is:

• 14 dependencies where 3 would suffice
• 53 screens to do everything
• 400 commits and zero paying customers

And above all: a solo founder who no longer understands their own code after 4 months. Even with AI.

The Vibe Coder's dopamine loop

The Vibe Coder lives in a pure dopamine loop:

1. They have an idea on Friday evening
2. AI spits out 80% of the boilerplate in 2 hours
3. They think "while I am at it, I might as well do it right"
4. They spend 300 hours polishing the architecture
5. They ship a V2 before even talking to a single customer
6. They burn out or give up when they realize they now have to sell, maintain, and support

Classic conclusion:

• The market does not understand anything
• Users do not see the potential

Coding is not creating value

So I have to go back to the essentials and play the killjoy for a moment, always with kindness, but with firmness: you simply confused coding with creating value. AI gave you a Ferrari with no driver's license and no road.

I have infinite respect for the work put in. Whether you are developers or aspiring Vibe Coders, to me the effort is magnificent. But the only judge is customer adoption. And it is far simpler to sell a single, simple idea than an OPNI that tries to do everything.

All hyper-competent. All alone. All at zero euros of revenue.

The worst part? Even Claude, Cursor, and Codex no longer understand their code after a while.

The logical continuation

This is the logical continuation of the crisis I described in my previous article. We broke the transmission. We killed product vision. We replaced "understand, design, maintain" with "deliver, deliver, recycle."

The result: a generation that knows how to build cathedrals... but no longer knows how to build a cabin that brings in 500 euros a month.

AI did not kill the developer's job. It killed the "Stop" button. And with it, any notion of healthy constraint.

So yes, you can do everything alone now. But if you do too much, you sell nothing.

The decisive skill

The real skill for 2026-2030 will no longer be knowing how to code faster. It will be knowing how to stop. Knowing how to say: "That is enough. It works. I am now going to talk to 100 customers instead of adding one more layer."

This is not a recent phenomenon: I have paid the price for it multiple times with partners who always wanted more without knowing how to say no. But it turns out that among the new Vibe Coders, this is a tendency that appears immediately, whereas I thought they would be more likely to have that perspective.