Raw LLM Responses

I can see this, I’m currently working on translating python code to High performant Rust code for a HFT, maybe this is just my experience, but development with Copilot (I tried pretty much all the models) slows me down immensely sometimes, it writes unnecessarily complex code to accomplish simple things. I’m not saying it’s stupid all the time but the amount of time spent fixing the code based outweighs the benefits you get from using an AI assistant. LLM are capable to producing great results at smaller scale, at scale they produce absolute garbage with confidence, that’s the distinction between humans and Ai as “senior engineers” , a senior engineer would never do that.