I use Jules myself, and it hallucinates like hell, but I can't afford Claude code. Honestly, I feel it struggles to understand my prompts. No wonder how good they are. Yeah, that's basically it. But why are you doing your analytics on agent coding experiences?