What’s a prototype in software development? Sometimes, to prove a point, showing is better than telling. You've an idea that would help users and the bottom line, but what's it’s impact?

We're a small software company and do prototypes on a regular basis. Our devs often take the initiative; they encounter an opportunity while fixing an issue and build a little proof of concept.

Or we do larger ones in preparation of major projects. Such as for the refactor of our inspection platform, where we built and demoed sync improvements.

Before AI and coding agents, such prototypes could easily take two to three weeks. For two to three developers. For us, that’s a substantial investment. And that’s not all, many prototypes never get to production. End up in the prototype graveyard.

We have a prototype graveyard, I guess many companies have. All in all it’s a substantial chunk of lost revenue. Wouldn’t it be great if we could build a prototype in hours rather than weeks?

Well, now we can. With AI, vibe coding, coding agents. But here’s the thing. Now that we build prototypes so much faster, how to decide which to move forward on? Or the more pertinent question: how to even find the time?

Conferences and platforms are being overwhelmed. NeuralIPS got nearly 30,000 papers to process, triple as many as earlier years.

Some for software. GitHub is flooded with prototypes: millions a year, many of which AI-created. At Web Summit 2025 the company Lovable showcased a product which aspiring startups use to create 100,000 prototypes daily.

Let me say that again. One hundred thousand. Daily.

Many are crap. Many are useless. With vibe coding, it’s easy to create a standalone software product, like an app, that looks cool and moves like the real thing. But that doesn’t make it useful.

Useful software products fit into a larger infrastructure that often consists of components that have been developed for decades. Think legacy, think banks still running software in COBOL.

Useful products make use of existing data flows, enhance and augment them. They make it possible to create new marketing products out of a marketing history system.

Useful products take existing regulations into account: here in Amsterdam we're beholden to EU guidelines regarding GDPR; when we build out our API prototype we must follow the documentation published by the government.

Of course, young upstarts are wanting to build the next big thing and don't worry. Sometimes it works: VibeSail is a simulator game that has found quite a following. I’m not into games, not even a little bit, but check it out, looks really cool, especially how the water is rendered.

But most, and now we’re talking millions, millions upon millions, are lying there, useless. Their light bright for the few short minutes or hours while being built and then bound for the ever growing graveyard of abandoned ideas.

Don't let it be yours.