pick the kind first, and every question you pick gets asked

setting up an episode used to start with a row of quick starts and, under it, what the episode was about. it looked like one screen with two halves. people pressed podcast with ai co-host at the top, expecting to pick a topic from the half below, and were taken straight off the screen. the order was asking two questions at once.

now it asks one thing first: what kind of episode. every kind is on that screen, on camera and voice only, each a card with a picture and a line saying what happens: video interview, audio interview, just chat, podcast with ai co-host, the co-host by voice only, walk me through it, watch me work, talk about it, with a guest, and a person asks. one click picks it and moves on. there is no separate screen for camera or voice any more, because that is part of the kind.

then only the screens that kind needs

after the kind, the setup shows only what is left to decide. the interview and the co-host ask what it's about next: what to film, today's context, ask the vault, or your own words, exactly as before. then the frame, portrait or landscape, for the kinds filmed in one, and one click on the frame moves on. then here's your episode, with each line there to change.

a kind that needs nothing skips the screen. watch me work used to land on make it yours with nothing on it but a next button, and a notes box nobody needed for hours of screen. now it goes straight from the card to the summary. just chat skips what it's about, because the whole point of just chat is that you already know what you want to say. no screen in the setup is ever only a next button.

same as last time and the shortcuts you saved are still one press each. they sit under the cards as your shortcuts, and a shortcut that knows its frame does not ask for it again.

the interview is set questions now

the bigger change is what happens once you are on camera. the interview used to have a clock: about three minutes of answers on a standard episode and five at most, enforced in code, with a director deciding after every answer whether to dig in, move on or wrap. picking six things meant the interview often stopped after the third or fourth, and the ones you picked last were never asked.

now the interview is set questions. each thing you pick is one question, asked in the order you picked them, and you answer each for as long as it takes. there is no countdown under the question and no follow-up after your answer. when the last one is answered, it ends. six picks is six questions and six answers, and every answer is a clip of its own, so a long episode is also six shorts waiting to be cut.

with nothing picked it still writes a few questions from your day, and those are asked the same way. the only limit left is thirty minutes of answers, a safety net for a recording left running, not a target.

digging in is the co-host's job

an interviewer that follows up and an interviewer that asks what you chose are two different things, and trying to be both made the interview worse at each. so they are two kinds now. the interview asks your questions and gets out of the way. the podcast with ai co-host is the one that listens, pushes back, quotes your own words and asks the next thing from what you just said. on a day you want to be pushed, pick the co-host. on a day you want to say six things cleanly, pick the interview.

on the phone

the phone has the same order: what kind of episode first, with the kinds the phone films, then what it's about for the interview and the co-host, then the frame, then the summary. the interview there has no countdown either, asks every question you picked, and never follows up. walk me through it, watch me work, with a guest and a person asks are still set up on a computer, and the kind screen says so.