▶ YouTube Summaries

• https://www.youtube.com/watch?v=HuDIPKKMCnk
• Original title: Embracing Failing

ThePrimeagen reacts to PewDiePie's unlikely arc: from installing Arch Linux, to building a home "supercomputer" that runs a council of competing AIs, to training his own coding model from scratch until it beat some of OpenAI's older benchmarks — and Prime uses PewDiePie's closing reflections on failure as a springboard to argue that today's developers, leaning on agents that "just fix it," are losing the ability to struggle, read error messages, and actually learn.

The PewDiePie arc that surprised Prime

Prime says nothing in the last few years surprised him more than watching PewDiePie — someone he'd written off as "just a goofball" — turn into a genuine, self-taught engineer. The journey: install Arch Linux, get into hardware and 3D printing to build a home supercomputer, put a "council" of competing AIs on it, then spend days generating and cleaning his own training data to train a coding bot from scratch, eventually beating some of OpenAI's older models on a coding benchmark. He did this while starting from zero programming knowledge, in part via boot.dev.

PewDiePie's own message: embrace failing

Prime plays PewDiePie's closing reflection: he describes going through "the whole alphabet of failures," generating "graveyards" of garbage data, and nearly giving up many times. His takeaway, inspired partly by watching Linus Torvalds talk about learning through failure: some people treat failure as bad, but it's actually how you learn. Expecting things to go smoothly sets you up to give up; expecting to fail and iterating is what lets you keep going. His message: "expect to fail, embrace failing."

Why this resonates with Prime

Prime connects it to his own first computer science lesson: "the most important thing I can tell you is that you will fail — when you see failures, read the error message." Once you've seen enough errors, they stop being scary and become an expected part of turning an idea into a product.

The generational worry: agents removing the struggle

Prime's core concern is that many newer developers only ever invest in "the agent coding lifestyle," so they hit a wall much later than someone who was getting syntax errors on their very first if-statement — and when they do hit an error, they don't have the muscle to push through it, because they've never had to. Instead of debugging, the pattern becomes "hey, fix it, don't forget to make it secure" and the agent just does it. He notes studies claiming AI increases failure rates or workload in real codebases, but says he finds the positive argument more convincing than the doom argument: struggling through problems makes you more patient, resilient, and capable of tackling bigger things.

Prime's ask

Acknowledging PewDiePie has more free time than most people, Prime argues that doesn't excuse skipping the struggle entirely — everyone can spend 30 minutes a day learning instead of just prompting, turn off the hype machine on Twitter, and over a year make a real difference. His closing line: make 2026 the year you go deeper instead of just faster.