Michel Lind  Mar 24

Hi, my name is Michel, and I (reluctantly) use #LLM

Since I'm about to announce the existence of some tools I wrote to help with some workflows in the Linux distribution space, I figure it's only fair if I first write about why I ended up (having to) use LLM.

https://michel-slm.name/posts/2026-03-24-my-name-is-michel-and-i-use-llms/

This post is day 32 of my #100DaysToOffload challenge. Visit 100daystooffload.com to get more info, or to get involved.

#blog #dayjob #foss #MetaInc

Hi, my name is Michel, and I (reluctantly) use LLMs

the old pen resists but the cursor blinks, waiting— I press Enter. Fine. ~ Claude Opus 4.6 (1M context) If you know where I work, you’ve probably heard of news reports that we will be judged on AI-driven impact. I’ll let you drawn your own conclusion on how much truth there is in the reports, but you can listen to what Zuck said about AI in a recent earnings report.

Pensées de Michel
Show threadLauren Weinstein
@michelin Sorry, I don't buy it. Using these sh*t machines for any purpose -- no matter how laudable you may consider your own -- encourages the Big Tech CEO fascists to continue pushing the tech into every aspect of our lives. The only way to fight them, absent meaningful regulations, is to starve them of usage and income. There is no "safe level" of LLM usage, just as there is no "safe level" of drinking alcohol before driving. If you voluntarily use these systems, you are part of the problem. There are no exceptions.
Mar 24 at 4:22pmWeb
Show threadMichel Lind  Mar 24

@lauren hence the bitter Thank you for Smoking mortgage quote in the blog post. If I had been able to #FIRE and I'm just doing this for fun I'd be able to be more principled on this.

So I mostly agree with you. "If you voluntarily use these systems" is key here - until the VC-driven AI craze meets its inevitable crash - maybe the energy crisis thanks to imperialistic hubris this year will trigger that - a lot of us are *coerced* to a greater or lesser extent to use these.

I'm actually looking for AI-free projects to contribute to when I'm not being paid to do open source, so I'm watching the @debian and @guix discussions around AI policy with a mixture of curiosity and trepidation.

Show threadell1e coding thingsMar 24

@lauren @michelin I'm trying to figure out how the Linux Foundation is condoning this, perhaps you would find this interesting? https://hachyderm.io/@ell1e/116285351290767548

(Deepest apologies for hijacking the topic, but it seems relevant to LLM acceptance in tech.)

Show threadLauren WeinsteinMar 24
@ell1e @michelin When was this policy published? I don't see an obvious date. Is it their first policy pronouncement on this issue or an update? Thanks.
Show threadell1e coding thingsMar 24
@lauren @michelin Sadly, I have no idea. I was only made aware of it today.
Show threadell1e coding thingsMar 24
@lauren @michelin However, I believe the Linux kernel has LLM-signed commits now, that started a few weeks ago. So the actions certainly match the policy, from what I can tell. Source for apparently an LLM commit: https://github.com/torvalds/linux/commit/a1e5c46eaed3151be93e1aec9af0d8f8db79b8f6 I don't know for sure if it contains LLM-generated code, but the signature suggests to me that it does.
selftests/bpf: Add tests for bpf_throw lock leak from subprogs · torvalds/linux@a1e5c46

Add test cases to ensure the verifier correctly rejects bpf_throw from subprogs when RCU, preempt, or IRQ locks are held: * reject_subprog_rcu_lock_throw: subprog acquires bpf_rcu_read_lock and ...

GitHub
Show threadMichel Lind  Mar 24
@ell1e @lauren the Linux kernel community and systemd has been quite open about using LLM. I think even Curl uses LLM but IIRC it's only for doing security assessment (since external parties are using LLM to find bugs anyway), not sure if they are already using LLM for writing actual code as well or not
Show threadMichel Lind  Mar 25

@ell1e @lauren so this is not the LF policy, but for the Linux kernel, the policies came up in the debian-vote discussion

https://lists.debian.org/debian-vote/2026/02/msg00032.html

Looks like https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/Documentation/process/coding-assistants.rst got applied 2026-01-06 but authored on 2025-12-23

Re: GR -- Allow AI-Assisted Contributions