5 comments

[ 3.5 ms ] story [ 24.5 ms ] thread
The real deal is to have deterministic quality gates. Nice to know that experienced programmers are terrified by the upcoming generation of savvy wildcoders.
LLMs are just a tool. Bad developers will find a way to use any tool to generate bad code. The right solution is robust review of PRs. If the developers can't engage in that, drop their PR on the floor.
I think the policy makes sense, but I don't like the evolution of a false dichotomy in the conversations about AI.

I'm a professional developer like many here. Based on my career achievements, I believe I'm at least decent at that. I completely understand the revolt against vibe coded solutions because I've seen many questionable things coming from AI. But the people who copy and paste things straight out of the LLM are simply poor developers, the same ones who would blindly copy and paste snippets out of Stack Overflow into commercial systems.

I think partnering with AI makes me a much better and more productive developer. I'm completely dispassionate about the thing.I use it because it saves me time. The majority of the code is still mine but the focused AI contributions under my supervision are better than anything I could have come up on my own in the same period of time.

In other words, the false dichotomy is 100% AI vs 0% AI. I understand Godot because if they don't discourage AI, they know they will be in practice inviting 100% AI. This said, given how good Godot developers are, I know that many if not most of them know that AI can be extremely useful if applied properly.