The editorial decisions behind Podcapp, written down as we go. Sentence-by-sentence checking, airtime budgets, sources we drop: what separates a real briefing from a machine reading your tabs out loud.
A model that writes well also writes false things with confidence. We put a step between the writing and the microphone: every factual sentence has to produce its quote, or it gets rewritten or cut. On the pilot episode, four sentences failed the check.
Read the article →150 seconds for private credit, 90 for a photography show, zero for the rest. Fixing durations before anyone writes is what turns a pile of links into a programme. Without that constraint, everything matters equally and everything runs long.
Read the article →A paywall, a video with no transcript, a deleted page. The temptation is to fill the hole with whatever the model thinks it knows. We chose the opposite: the source is named, dropped, and the listener hears about it in the outro.
Read the article →