The Silent Revolution

What happened over Christmas

I've been working with AI since 2019. I've seen every wave. But this one felt different.

Over the Christmas break I decided to put a few of my Magic of Me automations into Claude Code. Just to try it. Everyone was talking about it and I wanted to see what the fuss was.

I fell in love. Same way I fell in love with Lovable before it. And Daydream before that.

After about a week I was sold. I went all in.

Since then I've been using Claude Code 8-12 hours a day. I've written 350,000 lines of code. Last week I burned through 80% of my tokens on the Max plan.

The proof

Let me give you a few concrete examples.

FitGen

Last summer we built a fitness app for a client of mine. He's a personal trainer, actually a good friend by now. The app has free trials, paid subscriptions, Stripe payments, personalized workouts, the ability to save past workouts and weights. It's still used daily by multiple people.

We built it in about 1-2 months. That was already insanely fast. Normally it's a minimum 4-month project.

If I built that same app today? 1-2 days. Maybe less.

That's not an exaggeration. That's just where we are now.

Magic of Me

Magic of Me is an AI-driven children's book platform. Super personalized: the child is the hero of the story. We generate all 32 pages with AI, images and text.

Over the Christmas break I rebuilt a bunch of our infrastructure in Claude Code. And I realized we didn't need half the tools we were paying for anymore.

Here's what we cancelled:

• BannerBear (~€140/month): we used it to put text on images. Built it ourselves.
• CloudConvert: for file conversions. No longer needed.
• Airtable: replaced with Supabase.
• Cloudinary: for image storage. No longer needed.
• n8n: for automation flows. No longer needed.

All replaced with custom code. Built in Claude Code. By someone who doesn't write code.

Why this one is different

This doesn't feel like a 0-to-1 moment. GPT-3.5 was a 0-to-1 moment. Everyone felt it.

This feels more like a 1-to-2 moment. Quieter. But for anyone paying attention, it's actually more impactful.

The build vs. buy equation has flipped. Buying used to be the default answer. Custom software was expensive. You needed teams. It took months.

Now? Building is the answer. For a lot of things, anyway.

I genuinely believe every business can have its own super-customized tools. Software that fits their business perfectly. Not 20 different SaaS licenses duct-taped together. Real custom software.

The critical threshold

Something broke through. I can't describe it perfectly, but the intelligence level of these models has crossed some kind of threshold.

Vibe coding used to get you 70% of the way. Then you'd hit a wall. You needed a real developer to go further.

That wall is gone. Or at least, it's much further out than it used to be.

I haven't hit it yet. And I've been shipping products for weeks: real products, for real customers.

What this means

The paradigm has shifted. A lot of business decisions that used to make sense don't anymore. The formulas have changed.

And most people don't see it yet. They're still thinking like it's 2023.

For SMBs this is huge. If you're a small or mid-sized business, you can move fast. The CEO decides, you just do it. Your systems aren't set in concrete.

Big companies? They have different countries, different legal departments, legacy systems everywhere. Adapting is much harder for them.

Which means SMBs just got a lot more competitive.

The silent part

The weird thing is how quiet this all is.

When GPT-3.5 dropped, it was everywhere. You couldn't escape it.

This? Barely a whisper. A few people on Twitter. Some developers who see it.

But the impact is bigger. Much bigger.

For those who see it, the world has quietly shifted. The rules are different now.

And that's exactly why it's so powerful.

This article was written with Claude Code. Obviously.