Before
Humans design and freeze the format, questions, cast, moderator direction and publication rules before generation begins.
Independent AI debate podcast
Four AI systems. One complete generation run. No answer shopping.
Episode one is in pre-production
The show
GPT, Claude, DeepSeek and a self-hosted abliterated model enter a moderated debate. They answer serious questions, challenge one another and play a fast anonymous guessing round called Cards Against Artificial Intelligence.
How it is made
Humans design and freeze the format, questions, cast, moderator direction and publication rules before generation begins.
The complete conversation is generated as one recorded run. Individual panellist answers are not rewritten, swapped, reordered or selected from other takes.
Kua SpA may publish the whole run or reject it and start the entire episode again. Moderator passages may be scripted before generation. Dialogue is not rewritten afterward. Any narrowly scoped technical correction, such as removing a duplicate turn caused by a scheduling defect, is documented alongside the episode.
The on-air promise
The conversation you are about to hear comes from exactly one complete generation run, which is kind of like the AI equivalent of live television.
That describes answer selection, not literal live broadcasting. The published programme is edited entertainment, and the audio is produced in post.
With each episode
The episode archive keeps viewing links and dated corrections together. Materials for each release are added after review, without publishing credentials or sensitive private records.
Disclosure
The words and voices. Panellist wording is generated by the listed model/runtime combinations under producer-written prompts. The voices are synthetic—no human performer and no recording of one; they are not the participants, because each seat is a different AI model speaking through a synthetic voice assigned by the show.
The independence. Synthetic Beef is produced by Kua SpA and is not affiliated with or endorsed by any model developer, hosting provider, voice provider or Cards Against Humanity. Outputs are not company policy, verified fact or evidence of a stable model belief.