@nixCraft This is mostly code completion, which makes sense: they just switched from traditional IntelliSense to their own AI-powered one, and if it works well people will use it. It's technically generative AI, but it doesn't mean that it's AI slop with vibe coding. I think specialized models for language specific tooling is good, it should be possible to make it small enough to run locally too, I'm just wondering how far that path they went vs going full online LLM.