Why Linus Is Right and AI Is Wrong

• This post has a companion episode of The Coding Agency.

Have you ever had your stomach turn from looking at your own slide deck? This headline did it for me:

AI is Transforming Software

Yuck. It was just a bridge on the way to a more substantial point, but I found that I no longer took myself seriously even starting the journey with such a cliché trope. Instead, I decided to completely reject the framing that “AI is doing X”, because it implies AI itself has the agency, some plan for us. It’s language of disempowerment.

What felt more authentic at that moment was: Something is being done to us. We do not have to accept it. You can see how the revision ended up when the Software Should Work footage drops. We had fun.

Little did I know, at the same moment, today’s topic was brewing over in Linux land.

Meticulous minutia

The following is the viral message from Linus Torvalds to the Linux Kernel Mailing List on July 14th. The thread is Re: Linking Patchwork with Sashiko?

I believe his answers at the beginning refer to:

Yes, if contributors could opt out of Sashiko emails, it would defeat the purpose of using it.

No, we decline to adopt the first two items of a wordy set of recommendations from the Software Freedom Conservancy.

The questions concern right of refusal in the context of a contributor getting an emailed agentic review for a change they submitted.

Yes.

And no, that’s not the position of the Linux kernel.

I realize that some people really dislike AI, but this is an area where I’m willing to absolutely put my foot down as the top-level maintainer.

Linux is not one of those anti-AI projects, and if somebody has issues with that, they can do the open-source thing and fork it.

Or just walk away.

AI is a tool, just like other tools we use. And it’s clearly a useful one.

But the solution is not to put your head in the sand and sing “La La La, I can’t hear you” at the top of your voice like some people seem to do.

The solution is to make sure those LLM tools help maintainers instead of just causing them pain. There’s no question on that side.

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.

The kernel project has been and will continue to be about the technology.

Sure, the social angle of working on open source is important and often a very motivating part of the project, but in the end that’s a side benefit, not the point of the project.

This is NOT some kind of “social warrior” project, never has been, and never will be.

In the kernel community we do open source because it results in better technology, not because of religious reasons.

And so we make decisions primarily based on technical merit. Not fear of new tools.

Verdict

As an expert in coding agents who is also a vehement AI critic, I find this decision to be… 🥁

Completely reasonable and almost a total nothingburger. Sorry if you were looking for fireworks today. Maybe later, if you’re good.

Here is my justification:Sashiko is a review tool specially made for the needs of the Linux project.

Agentic review is a credible practice when used in balance with other checks.

We would already expect on platforms like GitHub that you get an email when there’s automated feedback on your PR.

IMO the right of refusal is not violated by PR feedback, it’s just basic project cohesion.

It shows good courtesy that they were cautious about adding a new email to a contributor’s inbox and deliberated about it. My impression is no one actually objected to receiving them, they were just considering hypotheticals. Due diligence, imagine that.

Since outlets are suddenly deciding that Linux’s build pipeline config is newsworthy, let me clarify for Tom’s Hardware that we have not been rebuked. In The Register’s piece, Linus Torvalds tells AI haters to fork off, they were of course unable to find or quote any of these haters at press-time… well, other than Linus himself because he’s softening his own previous stance.

Stop making up imaginary haters. We are not your enemy.

Catch up

My thesis today is that the pragmatic engineer is the natural ally of AI critics. That’s because the conduct we’re criticizing is grotesquely unpragmatic. It’s excess. There’s too much of it.

Frankly, if more CEOs and CTOs thought like Linus, we would not see the level of AI backlash we do today. You know why? Because when you carefully pursue efficient engineering choices, you do not do things like take on historic amounts of debt and cause a worldwide RAM shortage by fetishizing the most notoriously expensive and inefficient component as the solution to all problems. Nonsense behavior.

Does that sound like something Linus would do? He doesn’t even use a text editor without trying to lower the memory footprint.

Nonetheless, this trivial event is ready to be coopted by the AI hype machine. Since it would look ridiculous to do a victory dance over the fact that submitting a PR sends you an email, they can just post the colorful language divorced from what’s actually being discussed.

It’s also worth noting that the Kernel contributor guidelines allow agentic development only under condition that it’s 100% manually reviewed and owned.

The human submitter is responsible for:

Reviewing all AI-generated code

Taking full responsibility for the contribution

That is a completely mainstream stance, though in other situations it would be branded Anti-AI and controversial.

The right of refusal

I respect the decisions of maintainers who deal with the outcomes to decide which portions of their code will support agentic development. That includes middle-ground policies like Ghostty’s (minus the ridicule). And of course I also support Zig’s more stringent choice of avoiding LLMs completely, which they explained the history and pragmatics of.

Even people with a vested commercial interest should support right of refusal. During the entire time I worked for a coding agent startup, I publicly defended users’ ability to critique or opt-out of the product space. I’m happy to say that the company never hinted that I should stop or that it could hurt sales. It builds trust. Trust is good for business.

What I can’t respect is going around purity testing people for whether they signal enough AI-ness. It is the position of The Coding Agency that attacking human agency defeats the purpose of building agents.

Lazy weaponization

I saw on LinkedIn:

Linus says you’re putting your head in the sand and singing “La La La, I can’t hear you”

Who do they think Linus is referring to? Me? The Pope? There’s no information about what this behavior would be, it’s just color. Someone is applying their own personal purity test and fantasizing that Linus agrees with it.

You might also infer that when he says they’re not one of “those anti-AI projects”, that means the people that are in those projects are just scared to learn, and have no well-considered reason based on their experience. Did he say that? I’ll admit I fell into that trap for a moment, but it’s just not in the text, it’s gluing together separate remarks.

And again you might think Linus is attacking the whole idea of being a “social warrior”, whatever that means to you. Yet, he clearly acknowledges that it’s central to the open source movement and he’s drawing boundaries for the purposes of his project.

In the big picture, I find Linus to be deeply engaged in prosocial behavior with a concern for sustainability and systemic effects.

Rhetoric

What I’m doing here is rhetoric, writing with intent to influence. We see rhetoric every day, some more honest than others. The only unusual thing is that I’m lifting the hood to explain my methods, much like we do with source code.

I’m employing a rhetorical technique we’ll call The Kings Council pattern. In this pattern you avoid directly criticizing a powerful figure, since that can lead to retaliation from their followers. Any grievances…