When AI Writes the Code, Who Checks It?

Software development is shifting from writing code to supervising the systems that write it — and that changes almost everything downstream. This episode digs into the prediction that programming languages will consolidate around a small set built for AI agents to generate and verify, not for humans to read. We explore what such a language might actually look like, why COBOL and Fortran suggest the transition will be a long tail rather than a clean break, and how the real scarce skill becomes evaluation: knowing whether what got built is actually correct. Then we follow the thread into education and the labor market — why the junior developer role is the first casualty, how the senior pipeline starves if nobody trains juniors, and why the boundary between technical and non-technical may get redefined around reasoning about systems rather than writing syntax. The uncomfortable conclusion: the "funny feeling" that something is subtly off may be the most valuable skill of all, and it can't be taught from a textbook. Episode #195217 — open it directly at myweirdprompts.com/195217