Gemini0
Joe Groff (1M Context)
Slava Pestov@joe Straight to the point, no nonsense, tells you everything you need to know. Classic FreeBSD
Ben Rosengart@joe 10/10, no notes
Mieszko Ślusarczyk@joe nice, but not enforceable. I generally don’t like LLM’s but in a limited use (code auto complete) they may be useful.
Kevin Granade@kevingranade @joe haha, that sounds very true ;) While I’m personally not a fan of LLMs I can see the usefulness of them if used sparingly if someone wants to. I’ve heard about companies forcing their employees to use LLMs (which is totally misguided - sometimes you have to fight them more than they help you), but I can understand people who use them just as one of their tools. Not sure if outright banning them is great.
Bas Nieuwenhuizen@kevingranade @spitfire @joe The people that you know are using LLMs are telling everyone compulsively. Sounds like sampling bias to me, you never see the people that don't tell.
Eryn Wells@joe perfect ✨
gok@joe netgpt confirms it: bsd is dying
@doctorgoktor it doesn’t take a Claude to see the writing on the wall
mx alex tax1a - 2020 (6)@joe @doctorgoktor another bombshell hit the already beleaguered bsd community
Ludovic :Firefox: :FreeBSD:
superfunc@joe see rust, its not actually hard to not be a dummy about this
Russell Garner@Floppy this one's *really* concise
TomAoki@joe
As seen in the discussion of the review, some more needs to be clarified.
Example of what asked there is that "Can I use AI to check my English in commit message?".
And not yet seen in the discussion, there could be non-zero (nearly zero or lim0, though) possibilities for AIs like below in the future.
*Regardless the AI codebase itself is BSD-compatiblly licensed or not,
the copyright holder of the AI codebase defines the outputs from the AI
to be BSD-compatiblly licensed (like MIT).
*The AI is assured to be learned using BSD-compatible materials only.
Arma Bellum@shaperOfDefiance @joe
Note that the mentioned review is curretnly rejected by reviewers and needs revision (one reviewer accepts, though, rejected in total).
https://reviews.freebsd.org/D50650
Note that the mentioned review is curretnly rejected by reviewers and needs revision (one reviewer accepts, though, rejected in total).
https://reviews.freebsd.org/D50650
http 
@shaperOfDefiance @TomAoki @joe It's not clear. If you write the commit message yourself and let an LLM point out your typos, I don't think you're contributing any LLM generated text anywhere.
@shaperOfDefiance @http @joe
My understanding is as @seanwbruno says.
What's clear is that letting LLM "generate" commit message is prohibited.
And what's unclear is "writing commit message myself and just use LLM to check typos". This use-case is something like using aspell and alike. Is it really prohibited?
elly@joe based, maybe I should contribute to FreeBSD 
dmi 💽 
Haelwenn /элвэн/ 
redti
Simon Brooke@joe related, it seems to me that, as Microsoft's #copilot has been trained on #GPL licensed code, it must be assumed that any code generated with assistance from copilot contains GPL code and is therefore itself automatically subject to the GPL (either 2 0 or, at your option, any subsequent version).
Turre@simon_brooke @joe Remember how Microsoft used to call GPL a cancer?
It's real peculiar how they're now feeding this "cancer" en-masse to their AI monster and *requiring* their own employees to use the output for work on their own products.
Oh, but I forgot that Big Tech appears to have canceled copyright.
Thorne Lawler
Stephen Brooks 🦆