On Wed, 5 Nov 2025 at 22:50, Francesco Bonazzi <[email protected]> wrote:
>
> Maybe it should be made mandatory to disclose any usage of LLM when opening 
> PRs.

There needs to be at the very least a policy that bans using LLMs
without saying that they were used. There should be checkboxes when
opening a PR:

[x] I have or have not used an LLM
[x] I have checked the code from the LLM
[x] I have tested the code myself
[x] I do/don't understand all of the code that was generated by the LLM.

> Banning usage of LLM completely is a bit extreme, but it may be necessary if 
> vibe spammers keep flooding github with useless PRs.

Using LLMs to generate comments and PR descriptions should just be
banned outright with an exception for google translate type use only.
Obviously this cannot be enforced but the messaging needs to be clear:
don't dump LLM output as if it is a comment from yourself.

As for use of LLMs for writing code I don't necessarily object to
simple autocomplete type LLMs for convenience but I think if you look
at typical PRs and contributors right now they really should not be
using LLMs at all. The LLMs only seem to help people to make
time-wasting vibe-code PRs where the "author" has no understanding of
the code they are submitting and has not even done basic testing.

--
Oscar

-- 
You received this message because you are subscribed to the Google Groups 
"sympy" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion visit 
https://groups.google.com/d/msgid/sympy/CAHVvXxTfs2xHzfzWz66xRor_o0ZOg2JLRdXNqBvC6RWve81A-A%40mail.gmail.com.

Reply via email to