Yup that's true. We do benefit from massive efficiencies though, thanks to LLM codegen.