The threat is comfortable drift toward not understanding what you're doing
Summary
An astrophysics professor contrasts two paths using AI in research: Alice, who learns through traditional study and hands-on debugging, and Bob, who relies on an AI agent to read, code, and write. The piece argues that current incentives reward output over deep understanding, leading to a dangerous drift toward producing papers without true comprehension. It calls for a clear boundary between using AI as a tool and outsourcing thinking, warning that preserving the process of learning is essential for meaningful science.