Conversation

Notices

1. Embed this notice
   SuperDicq (superdicq@minidisc.tokyo)'s status on Saturday, 09-May-2026 02:52:01 JST SuperDicq

   in reply to

   • PuercoPop
   • tusharhero

   @tusharhero@mathstodon.xyz @PuercoPop@mastodon.social Yeah that is the right move. The legal precedence around LLM generated code is currently still unclear.
   
   Code generated using an LLM that was trained on proprietary code might be considered a derivative work which means it might be not usable. It is safe to wait until the dust settles on this.

   • Embed this notice
     PuercoPop (puercopop@mastodon.social)'s status on Saturday, 09-May-2026 02:52:03 JST PuercoPop

     in reply to

     I really hope that Emacs takes a stance against LLM generated code sooner rather than later.

   • Embed this notice
     tusharhero (tusharhero@mathstodon.xyz)'s status on Saturday, 09-May-2026 02:52:03 JST tusharhero

     in reply to

     • PuercoPop

     @PuercoPop recently heard that GNU is going to put on hold all AI patches. And I am sure RMS isnnto letting it get merged in Emacs (he is still personally active in devel mailing lists)

   • Embed this notice
     PuercoPop (puercopop@mastodon.social)'s status on Saturday, 09-May-2026 02:52:04 JST PuercoPop

     I already held Eric Raymond in the worst regard (and him calling integration tests "front-end tests" does not help). But they are now trying to upstream LLM-assisted code into #Emacs and they are upset they can't do it all in one go.

     > I know, but I've been systematically auditing the code for those
     using chatGPT 5.5, which is very good at that kind of deep dive.
     It tells me I've covered almost everything that can be tested in
     batch mode in my 42 patches.

     https://lists.gnu.org/archive/html/emacs-devel/2026-05/msg00138.html