OMG this was so funny/sad/scary at the same time. It really makes me want to make blog posts on my site providing nonsensical solutions to problems just to see if I can get those solutions to propagate into questions on here.
@paintsimmon @normalized It would be amusing to see some blog posts about how the code golf answers are the best new way to write Godot with some BS about why it’s better and see if the LLMs pick them up.
Agreed. TBH, lately I’ve been skipping answering a lot more questions than usual. Even if I have the answers. I also pay attention to repeat posters who need too much hand holding and stay out of their subsequent threads.
This makes a lot of sense.
A number of us do that on here. We try to help them get there. But, it also does depend on the difficulty of their question. I’m a big fan of having people step back and telling us what they are actually trying to accomplish with their broken code, because a lot of people make things WAY more complicated than they need to be.
I used to do this a lot, but as @normalized has mentioned, I do this a lot less frequently.
Here’s the article mentioned in the video. It’s a short read, and faster than listening to this guy read it to you for 5 minutes.
I get why Rémi says that more money for code reviewers feels like the only solution. Sadly, I don’t think it’s scalable. LLMs can push out slop way faster than people can. Perhaps the ability to limit who can make PRs will help, but it’s another gate that has to be maintained by someone and a hoop that new contributors have to jump through.
Godot already has contribution guidelines that specifically state LLM contributions are not allowed and users must understand the code they are contributing. @athousandships mentioned in this thread that too few people don’t read the guidelines. What would be nice would be if GitHub created a “Doesn’t accept LLM contributions” options and then the major LLMs were to respect that. Much like meta tags in webpages to reject webcrawlers.