Andrej Karpathy walks through his personal workflow with modern LLMs — the tasks he delegates, the prompting patterns that work and the guardrails he keeps in place when models are wrong.