When LLMs write code to accomplish a task, that code has to actually run somewhere. And right now, the options aren't great. Spin up a sandboxed container and you're paying a full second of cold start overhead plus the complexity of another service. Let the LLM loose on your actual machine and... well, you'd better be watching. On this episode, I sit down with Samuel Colvin, creator of Pydantic…