the non-bloat collective

AI threatens the computer

Written by MESYETI on the 1st of June 2026

This is the first of two articles for the non-bloat collective about generative AI and its consequences, the next one will be "AI threatens humanity".

This collective primarily exists to fight bloat, which is an issue that has been attacking the computer for a long while now. In this decade, the rise of generative AI has started to also attack the computer. This article will focus solely on Large Language Models (LLMs). Other forms of generative AI (image, audio, and video generation) will be discussed in the next article.

Generative AI is this new thing that has become interesting to some people and is supposedly improving all the time. It seems like a lot of people (older people I think?) have an obsession with it and/or are fascinated by it. They're willing to look past what is so evil about it and how it really isn't as good as they think it is. They become reckless, and are willing to use it dangerously.

The rest of this article will be cases of this, and more will be added in the future.

Case one: rsync

At the time of writing this, the LLM "Claude" is 2nd in the list of contributors in the past 3 months, ranked by number of commits. It is the 3rd highest for all time. The rsync repository started in 1996, and rsync is one of the most popular programs used on Linux. Claude was able to become the 3rd highest contributor in a 30 year old repository because the maintainer decided to start letting Claude go wild on the codebase, with 15,035 additions in the past 3 months.

This, of course, worsened the stability of rsync. It led to rsync failing to compile on older versions of Linux (without openat2, according to #905, #924), and older Darwin (according to #896). It also led to features being broken (#910, #915, #897).