The editorial case for letting an agent drive
Every time we propose fully autonomous production, someone asks 'but what about editorial taste?' Here's the argument.
I was a print editor for ten years before I took this job. I approached the idea of agent-driven podcasts with the same suspicion I used to reserve for automated copy-editing tools. Then I ran the math on what a human editor produces in a week versus what a well-prompted agent can produce in a day, and my suspicion got quieter.
Taste is a policy, not a keystroke
The editors I admire don't edit line by line. They write the policy. They decide what kind of story the outlet tells. The actual rewriting is almost clerical. The hard part — the taste part — is the frame.
Agent-driven production forces you to externalise that frame. You can't vibe your way through it. You write down, explicitly, what this show is and isn't. What voice it uses. What it never does. And then the agent enforces that policy on every episode, uniformly, at 4 a.m. on a Tuesday when no human editor is available.