Linus Torvalds the guy who created Linux, the operating system that runs basically every server, every Android phone, and most of the internet just put his foot down. Linux is not an anti-AI project.
And if you have a problem with that, you can fork the code or walk away.
His exact words on the kernel mailing list this week:
"AI is a tool, just like other tools we use. And it's clearly a useful one. It may not have been that clearly even just a year ago, but it's no longer in question today."
"We're not forcing anybody to use it, but I will very loudly ignore people who try to argue against other people from using it."
Here's how fast this shifted:
October 2024: Torvalds called 90% of AI "marketing hype." Said he'd basically ignore it.
January 2026: He shipped AI-generated Python in his own hobby repo. Said he skipped googling and let the AI handle it directly.
July 2026: The Linux kernel now formally accepts AI-written code.
Mandatory disclosure. Human takes full responsibility. AI cannot sign off on contributions. But the code is welcome.
The thing that triggered this was Sashiko an AI tool built by Google engineers that automatically reviews kernel patches for bugs.
It caught 53.6% of real bugs that had already passed human review. Some developers pushed back against using it. Torvalds shut that down.
"This is NOT some kind of social warrior project. We do open source because it results in better technology, not because of religious reasons."
The person who built the operating system the internet runs on just told the world: use the tools or get out of the way.
Ever run MEMMAKER or edit CONFIG.SYS? Here's a story...
One of my very first projects at Microsoft was for MS-DOS 6.2, and it involved moving the Doublespace compression engine and parts of SmartDrv into the HMA, or High Memory Area.
Then you could run MEMMAKER, and it would find up to 25K more than you had before. 25K was a big deal in that place and time!
The only problem was that I was new to x86 and had no idea what the HMA even was. It turns out that on an 80286, when you get to the top of the 1M area, what happens if you point "past" it by combining your segment and offset registers to a value greater than 1M?
In other words, does your pointer wrap at 1M or go up into high memory that might be installed?
On an 8088, all it can do is wrap. But on an 80286, it COULD be resolved into the next higher pages of memory. But which behavior should the system do?
To solve this, IBM used the A20 line on the PC, which is actually a keyboard line. If set, your pointers wrapped, and if not, they went into high memory.
This gave you the ability to reach into "memory above 1M" even on a 1M system, which is kinda crazy, but that's how it worked.
So I put all the code I could stuff into the 64K right above the 1M boundary, then stubbed everything with jump tables in low memory. I forget why the net was only 25K and not 64K - but there were probably other HMA consumers as well, I imagine.
Anyway, that's how I spent the first month of the summer of 1993! And there was much rejoicing.