How to Build an AI Roast Room Council in Claude

TL;DR

The Roast Room uses seven AI advisers as sub agents plus a chair to debate a question in rounds and issue a final verdict with a PDF deliverable. It is built inside Claude Code and designed to replay sessions and track opening positions, debates, and final scores. The system exposes the brain of each advisor and how they communicate.

Transcript

So right here you can see that I've got Jeff Bezos, Paul Graham, Alex Herozi, Gary Vee, Dan Martell, Gino Wickman, and Warren Buffett sitting here in what I call the roast room. And what's happening is all of these advisers are writing their opening positions and then they're going to have a debate about the actual question that I submitted to them... Read More

Key Insights

• The Roast Room forms a seven advisor council to debate a question.
• The debate proceeds in rounds with opening positions and cross argumentation.
• A chair sub agent facilitates the debate among sub agents.
• Each advisor has a focused domain such as pricing or growth.
• Final verdict is produced as a PDF and can be replayed.
• Sessions and prompts are stored for future review.
• The system uses sub agents instead of agent teams for cost and voice fidelity.
• The setup is reusable via a free skill and repository.

Questions & Answers

Q: What is the Roast Room and how many advisers are involved

The Roast Room is a structured debate setup where seven advisers, inspired by real figures, discuss a question in rounds. It uses a chair to facilitate, sub agents for each adviser, and a final verdict phase that yields a PDF deliverable. It is designed to help weigh options and capture reasoning.

Q: What role does the chair play in the Roast Room

The chair acts as the facilitator of the debate, coordinating rounds, guiding opening positions, and prompting advisers to argue or respond. This reduces confusion and ensures the discussion follows a formal flow from opening statements to final voting.

Q: How are the advisers’ opinions gathered and aggregated

Each adviser writes an opening position based on their focus, then the group reads and responds to others. Revisions occur across rounds, and the final votes determine the verdict. The process captures diverse perspectives and produces an average score when applicable.

Q: What kind of output is produced at the end of a session

At the end of a session a final verdict is produced and delivered as a PDF. The PDF includes scores, round details, and key points, allowing teams to review the reasoning and decisions without watching the entire debate.

Q: Can past sessions be replayed or reviewed later

Yes, past sessions can be replayed in the interface, and the system stores all sessions so you can revisit the briefs, rounds, and the final verdict. This helps in understanding how decisions evolved over time.

Q: Why are sub agents used instead of agent teams

Sub agents are used because they provide distinct voices and are cheaper while preserving the individual brains of each adviser. This setup also keeps the decision process organized and allows the chair to manage discussions efficiently.

Q: What domains do the advisers focus on

Advisers focus on specific questions like whether an option is the right bet and how reversible it is, whether there is actual demand and growth, and how the decision affects cash risk and long term value. This helps cover multiple business dimensions.

Q: Is the Roast Room available as a free resource

Yes, the creator points to a free skill and repository that allow users to set up their own Roast Room sessions, view past sessions, and replay debates, making the concept accessible to anyone interested in structured advisory debates.

Summary & Key Takeaways

• The Roast Room assembles seven AI advisers inspired by real figures to discuss a given question in structured rounds. The setup includes opening positions, cross debate, and final voting with a summarized verdict. Deliverables include a downloadable PDF and replayable past sessions.

• The mechanism uses sub agents with a chair to facilitate and a cloud code interface to run and record the debates. It emphasizes transparency, scoring, and the ability to review different paths of reasoning.

• The creator provides a free resource and explains how the system stores sessions, how prompts dictate who speaks, and how the final verdict is formed and exported for decision making.