Thirty-five years ago today, on August 25, 1991, a Finnish student posted to comp.os.minix that he was writing a free operating system, adding that it was “just a hobby, won’t be big and professional like gnu.”

The first public release came the following month. The initial codebase was around 10,000 lines. It is now roughly 40 million.

The quote everyone gets slightly wrong

The famous line is usually deployed as an example of someone spectacularly underestimating their own work, which is fair enough. But the more interesting part of that post is what came after it: a request for feedback on what people would want, and a frank list of what the thing could not yet do.

Linux did not win because Torvalds had a grand vision. It won because he shipped early, took patches, and kept a release cadence that made contributing worthwhile. The technical decisions that mattered most were arguably the social ones.

Where it actually ended up

The scoreboard at 35 is difficult to overstate:

• Effectively all of the world’s top supercomputers
• The overwhelming majority of public cloud infrastructure and web servers
• Android, and therefore most of the phones on the planet
• Embedded systems from routers to cars to television sets
• The infrastructure underneath essentially every large internet service

Desktop remains the one place it never dominated, which is the thing Linux gets teased about and the least important category on the list.

What 35 years looks like in practice

The anniversary lands in a week that says a lot about the project’s current state. The 7.3 merge window is open, removing two filesystem drivers that predate most of its contributors while merging support for a controller Valve has not shipped yet. Torvalds himself just spent 18 reboots chasing a two-year-old one-line bug in a GPU driver.

That combination is the actual character of the thing. It is simultaneously carrying decades of accumulated history, adding hardware support faster than the hardware ships, and still occasionally requiring its creator to personally bisect something horrible.

The part that is genuinely remarkable

Not the technology. Plenty of good kernels have existed.

What is remarkable is that a project this large, this critical, and this contested has kept a functioning development process for 35 years without a company owning it, and without fragmenting into incompatible forks the way almost every other Unix did. The kernel has survived commercial pressure, licensing fights, its creator’s occasional communication failures, and now an argument about AI-generated contributions that is far from settled.

It is still shipping releases on a predictable schedule. That is the achievement.

Happy birthday to the hobby project.