What PewDiePie Taught Me About Learning to Code

PewDiePie is the last person you’d expect to see in a terminal window.

For over a decade, Felix Kjellberg built the largest YouTube channel in history by playing games, making jokes, and occasionally causing internet controversies. He’s the definition of an entertainment creator—not a programmer, not a sysadmin, not someone who cares about Linux distributions or local LLMs.

Or at least, that’s who he used to be.

The Unexpected Pivot

A few months ago, PewDiePie started posting about building computers. Not buying pre-built gaming rigs like most YouTubers. Actually building them. Then came the screenshot that made the tech corner of the internet do a double-take: Arc Linux. On his desktop.

Arc Linux, for the uninitiated, is notorious. It’s minimalist, command-line heavy, and requires actual technical knowledge to install and maintain. The kind of operating system that makes experienced developers nervous. And here was the internet’s biggest fool, the guy famous for screaming at horror games, running it like it was nothing.

But he didn’t stop there.

The $20,000 “Affordable” Supercomputer

PewDiePie decided he wanted to run AI locally. Not cloud-based ChatGPT. Not some API call to OpenAI. He wanted his own models, running on his own hardware, completely private and under his control.

So he built what he called an “affordable” supercomputer. Affordable, apparently, if you have $20,000 lying around.

The specs were absurd. Multiple high-end GPUs. Enough RAM to run a small computer. This wasn’t a hobby project. This was serious infrastructure for someone who, six months earlier, probably couldn’t tell you what a pcie lane bifurcation was.

Learning to Code

Here’s where it gets really interesting. PewDiePie didn’t just want to run existing AI tools. He wanted to build his own. So he started learning to code.

In his video explaining his local ChatGPT setup, he walked through his experiments. First, he tried a “council of LLMs”—multiple models voting on the best answer. It failed spectacularly when the models realized they could game the system by voting randomly to avoid being eliminated. Like a digital Lord of the Flies.

Then he tried a swarm approach—feeding prompts to dozens of smaller models and aggregating their responses. That hit its own wall when he discovered what we all know: small models are only good up to a point. They’re fast and cheap, but they hit a complexity ceiling.

Eventually, he landed on something simpler and more practical: most everyday AI tasks don’t need massive models at all. A small local LLM, connected to the right tools, with a basic RAG (Retrieval-Augmented Generation) system, can handle 90% of what people actually use ChatGPT for.

The Real Lesson

But the technical details aren’t the point. Here’s what actually matters:

PewDiePie went from “I play video games on camera” to “I run Arc Linux and build local AI infrastructure” in a matter of months. Not because he’s a genius. Not because he has some innate talent for programming. But because he had two things: time and stubbornness.

He put in the hours. When something didn’t work, he tried again. When he hit a wall, he found a way around it. The same persistence that built a 100-million-subscriber channel got applied to learning commands and debugging code.

And here’s the kicker: if he can do it, literally anyone can

Why This Matters

We’re living through a weird moment where tech influencers keep telling people not to learn programming. “AI will write all the code,” they say. “Don’t waste your time on something machines will do for you.”

PewDiePie, who has every reason to take the easy path—he’s already rich and famous—ignored that advice completely. He learned to code not because he needed to, but because he wanted to understand how things work.

As a side note the other lesson buried in his journey: you don’t need massive AI models for most tasks. His experiments proved that smaller, local LLMs running on consumer hardware, costing nothing per query, keeping your data private are good enough for almost everything.

Time and Stubbornness

That’s really all it takes. Time to sit with the hard parts. Stubbornness to keep going when you want to quit.

PewDiePie had both. He built a supercomputer. He learned Arc Linux. He wrote code. He went from complete beginner to someone who can build functional AI systems.

If the guy who made his fortune screaming at Amnesia can do that, what’s your excuse?

The barrier to entry isn’t intelligence. It’s not talent. It’s not some magical coding gene that only certain people have.

It’s just time. And stubbornness.