Torvalds' 'debug session from hell': AI kept saying the bug was impossible to fix

Full text

Linux creator Linus Torvalds has famously been fairly hot and cold on generative AI in coding work, seeing a good purpose for it whilst also acknowledging a lack of quality control has led to a flood of unhelpful reports . Well, it seems like AI won't be fully replacing him just yet, as he's recently undergone a 'debug session from hell' where his AI assistant recommended he give up several times.

As spotted by International Cyber Digest , Linus took to a Linux Kernel commit to state all it took was "fixing a bogus 'round_up()' to a 'round_down()', but there were 24 patches adding more and more debug information to this, and 18 kernel boot to finally narrow it down to this."

This is a good example of where an AI might struggle to fix code, because it's not a syntax error or a bug, necessarily. The function is technically 'correct', it's just not doing what Torvalds wants it to . "The AI several times stated flat out that this was impossible and unsolvable," says Torvalds, "and that we should just write a report about it."

It seems like grit, determination, and stubbornness were what got the AI to behave in the end. Torvalds says, "I suspect those things have been trained by people who may not be quite as stubborn as I am."

Torvalds does praise the AI he's using, noting that, "while the AI was ready to give up several times, it did keep adding debug code and analyzing it faithfully when I pushed. So credit where credit is due and I let the AI write the commit message above."

Linus Torvalds says tracking down a bug in Linux's Intel GFX Driver was a "debug session from hell." The AI helping him repeatedly called it "impossible and unsolvable" and suggested they just write a report instead.He kept pushing. 24 debug patches and 18 kernel boots later he… pic.twitter.com/lUcMnECd9S August 26, 2026

Generative AI in coding has proved to be somewhat contentious. Last year, Collins Dictionary awarded 'vibe coding' (coding with AI) the word of the year , and a paper at the beginning of 2026 argued that " vibe coding kills open source " due to the lack of human attention and the ability for anyone to contribute huge amounts of data.

AI vibe coding ports of old console games have even become a bit of an ethical battleground . The team behind DK64 Randomizer (as you might guess, it randomises Donkey Kong 64) has been developing an unofficial PC port, specifically saying "Fuck AI vibe coding bullshit."

That's because, ignoring environmental concerns of AI plus the ethically dubious sourcing of data for training , teams like DK64 Randomizer argue that reliance on vibe coding results "in a poor quality project that will make it progressively harder to manage as time goes on."

Some developers clearly argue the benefit of AI code right now, but even so, it seems a human touch is often needed.