Conversation

Notices

1. Embed this notice
   Albert ARIBAUD ✎ (aaribaud@mastodon.art)'s status on Friday, 29-May-2026 22:37:54 JST Albert ARIBAUD ✎

   in reply to

   • Thom, a distro-agnostic tool
   • ARGVMI~1.PIF

   @argv_minus_one @thomholwerda Not only is the equipment still fine after the LLM faithfully executed the instructions which the LLM operator neglected to check *before* running them... but even the test deletion is purely symbolic, because the LLM operator would just need to fetch that code from the source connfiguration management too that they use, because of course, they use one, right? Right?

   • Embed this notice
     ARGVMI~1.PIF (argv_minus_one@mastodon.sdf.org)'s status on Friday, 29-May-2026 22:37:55 JST ARGVMI~1.PIF

     in reply to

     • Thom, a distro-agnostic tool

     @thomholwerda

     And it certainly doesn't qualify as “causing harm to computing equipment”. The equipment is fine after that, and so is everything stored on it, with the sole exception of this person's software, which has deleted itself.

     Again, this doesn't even disrupt production operations, because this code is only interpreted by AI coding agents when running tests.

     Unfortunately, I'm not confident that government prosecutors will understand this nuance and stay their hands…

   • Embed this notice
     ARGVMI~1.PIF (argv_minus_one@mastodon.sdf.org)'s status on Friday, 29-May-2026 22:37:57 JST ARGVMI~1.PIF

     in reply to

     • Thom, a distro-agnostic tool

     @thomholwerda

     anttibrax-2025 threatens criminal prosecution over this. Curious.

     The code in question causes this software to delete itself when exposed to an LLM coding agent. In development environments only, not in production.

     Does that qualify as “clear intent to cause damage”? I don't know about that.

     It arguably demonstrates clear intent merely for one's work to not be used with an LLM coding agent, and a minimally-disruptive method of enforcing this intent. Nothing criminal about that.

   • Embed this notice
     Thom, a distro-agnostic tool (thomholwerda@exquisite.social)'s status on Friday, 29-May-2026 22:38:00 JST Thom, a distro-agnostic tool

     Wait, did someone add a secret instruction to their code so that any developer using "AI" with that code would be in for a bad time?

     That's a type of chaotic good I can get behind.

     https://github.com/jqwik-team/jqwik/issues/708