Mastodon Feed
Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
fred@fthevenet.eu ("Frederic Thevenet") wrote:
@cstross @zzt @davidgerard the best way to enforce an LLM ban in a FOSS project is... to do strictly nothing outside of banning the use of LLMs, IMO.
If a careful reviewer gets fooled by someone secretly using an LLM i.e. the submitted code is reasonable and purposeful, the general use of language around it -- comments, docs, commit messages, etc... -- not hell bent on wasting everybody's time, and the submitter answers questions with a demonstrable understanding of the problem they're claiming to fix, then fine. Who cares?
The thing is, it's not even remotely believable that this could happen outside of the most trivial of cases without requiring the submitter to work harder massaging the slop than doing the work in the first place. Then, they'd invariably get both so goddamn proud they fooled you and gutted no-one even noticed, that they wouldn't be able to strop bragging about it...
A which point you'd just throw out their PR and ban their arses for breaking the clearly stated rules ¯\_(ツ)_/¯