From the founder · BlueberryS

AI won't earn for you. But in half a year, this is what I built with it.

"AI will make more millionaires than anything before" — you read it everywhere. And most people picture AI earning for them. It won't. Here's the honest middle: what AI actually changes — and what I built in half a year, so I don't just have to claim it.

By Gabriel Meždej · founder of BlueberryS, maker of Darwin

Related: Money is a river — what it actually is ›

"AI will earn for you" is a myth

The picture is tempting: you sit down, say a prompt and wait for a miracle. Now and then someone gets lucky and their story circulates online — usually distorted — and millions of people hope for the same miracle. Reality is different, and always will be. AI won't sell your product. It won't earn you trust. That stays with you.

What actually changed: speed

I've spent over thirteen years on software for bridge engineers — EasyBridge — and I hand-coded it myself, for years before AI even existed. That's exactly why I have something to compare against. And today it moves at a completely different speed. Making things is genuinely easy now — and that's the real change: not that AI earns, but that it lifts the ceiling on what one person can build. The bottleneck moved from "make it" to "sell it and earn trust."

That's why I show the result, not promises

Plenty of people say "Claude Code does this and that." Few show the result. Over the last half year I've built several real applications with Claude Code — some free, some paid — each solving a specific problem. Don't take my word for it: look. They're at blueberrysdigital.com.

How much work it is — by AI's own estimate

I had the same work priced independently by three tools — Codex, GPT and Claude Code. The Darwin platform alone — over 100,000 lines of my own code, 1,251 commits, one main author — came out at 11 to 15 person-years of work and a value of around €2–3 million at a quality European team; by US pricing, up to ~$7 million. And that's just Darwin — across the whole portfolio (EasySolutions, OnlineHelp, PrivateChat, DriveLog and more) it adds up to roughly 20+ person-years and several million euros. Take it as an estimate, not an audit; but even as a ballpark it says plenty about what one person with AI can do today.

A different kind of development when you use what you build

And one more thing that's hard to price, yet may matter most: I build tools I use myself. I know exactly what I need from them — and that's a completely different kind of development than when a pure software house does it. There, a lot of energy is lost just moving requirements around: what the client means gets translated into a spec, the spec into code, and noise creeps in at every step. When something needs to change, my decision is a matter of hours or days. At a larger firm it's days to months before testing even surfaces it. Domain knowledge and your own skin in the game — that can't be replaced by an hourly rate.

But building ≠ selling

Having an idea and building something doesn't mean people know about it, trust it, or buy it. That's the real work — and it starts when the product is finished. That's why I monetise gradually, and I'm starting with Darwin. And it's why my products carry an affiliate program: if you have an audience and can present a product honestly, the door is open. More in Money is a river.

AI isn't a miracle that earns for you. It's the fastest way to build something real. The rest — selling, earning trust, sticking with it — is still on you. And that's exactly why it's worth something.

Tools & communities I use

If you want to learn to make videos, prompts and content with AI:

• Dreamfall Art Community (Skool) — learning video and prompt craft.
• Higgsfield — AI video generation.

These are affiliate links — signing up through them supports me, at no extra cost to you.