Your LLM writes the channel.
Hand your model our skill and a topic — it writes a whole channel as JSON. You direct it, load it up, and run casting on the cards. The format's the API; your LLM's the writers' room.
your LLM writes the show. you run casting.
Direct your LLM
Hand your model our skill and a topic — “drill the AWS exam, deadpan, 8 cards.” It writes a whole channel as valid be2ween JSON.
Load it up
Drop the JSON into the back room. Every card auditions on the real renderer — keep the good ones, cut the rest.
It airs
The keepers become a channel on your dial, airing in your next gaps. No backend, no account, nothing leaves your machine.
llm → emits 8 valid be2ween cards · you audition them · the keepers air on your dial
Rather hand-craft them? Then share the pack.
Prefer to write cards yourself? The back room's a plain form — a kicker, a headline, a live preview on the real renderer. Build a channel by hand, then export it as a pack: a single file you hand to a friend, who auditions it onto their own dial.
Made one worth airing for everyone? Send the pack to master control to get it signed onto the network.
your channels. on this machine. they air in your next wait.