LLMs are already nearly as smart as humans. Whatever needs to be known should be able to inferred from the documentation