Quick Overview
In this demonstration video, creator Nate Herk presents an automated business advisory framework called Roast Room. He explains how he configured Claude Code with multiple specialized subagents modeled after prominent entrepreneurs and investors. The video shows the technical structure of the project repository, the debate and voting mechanics, and how users can run the workflow.
Key Points
- 1.Nate Herk created an AI advisory board called Roast Room comprising seven distinct subagents modeled after Jeff Bezos, Paul Graham, Alex Hormozi, Gary Vaynerchuk, Dan Martell, Gino Wickman, and Warren Buffett.
- 2.The council members debate business prompts among themselves, challenge each other's viewpoints, and cast weighted votes on potential options.
- 3.Each simulated advisor runs on a dedicated markdown brain folder compiled from public materials, including frameworks, principles, stories, and voice profiles.
- 4.For Alex Hormozi alone, the knowledge base was populated from 124 public videos and talks containing over 1.3 million words.
- 5.The system produces a structured verdict and summary report that can generate a formatted PDF deliverable detailing advisor scores, recommendations, and debate rationales.
- 6.The setup operates as a Claude Code skill and can be executed entirely through the command-line interface without requiring the graphical web dashboard.
Summary
Nate Herk introduces Roast Room, a multi-agent system designed to act as an advisory council for business decisions. The council consists of seven simulated business personalities: Jeff Bezos, Paul Graham, Alex Hormozi, Gary Vaynerchuk, Dan Martell, Gino Wickman, and Warren Buffett. Herk demonstrates the platform by submitting a prompt regarding an upcoming in-person conference event. Once initialized, the subagents interact directly with one another, debating competing options, challenging assumptions, and defending distinct business philosophies.
During the demonstration, the visual dashboard highlights interactions between the advisors. Alex Hormozi argues against Jeff Bezos, Gary Vaynerchuk debates Warren Buffett, and Dan Martell pushes back on Gary Vaynerchuk. After three rounds of discussion, the council members evaluate the proposals and cast weighted votes. The resulting consensus is compiled into an exported PDF report. The document outlines the chosen course of action, presents weighted average scoring across each evaluated option, notes specific objections, and summarizes why the council reached its final recommendation.
Herk then inspects the project directory on GitHub to explain how Roast Room is structured behind the scenes. Under the project directory, a dot-claude folder contains agent definition files corresponding to each council member, along with dedicated brain folders. Inside each advisor's brain directory, markdown documents store structured knowledge such as frameworks, decision principles, stances, source materials, stories, and voice guidelines. For Alex Hormozi, Herk shows diagrams indicating that the persona was constructed by processing 124 public videos and talks comprising more than 1.3 million words.
Finally, Herk shows that the system does not depend on a graphical frontend to function. He displays a terminal window executing the multi-agent session directly through Claude Code. Running the council skill in the terminal launches the subagents, evaluates the arguments across debate rounds, and outputs the completed verdict markdown file alongside all advisor scores and rationale without requiring the web dashboard.
The Roast Room Council and Debate Process
Nate Herk demonstrates a system called Roast Room where seven virtual advisors debate business choices. After a business dilemma prompt is submitted, simulated personas for Jeff Bezos, Paul Graham, Alex Hormozi, Gary Vaynerchuk, Dan Martell, Gino Wickman, and Warren Buffett debate back and forth. The video illustrates these agents arguing with one another across multiple debate rounds before casting individual weighted votes on proposed options.
Decision Outputs and Final Verdicts
Following the debate phase, the council compiles a structured verdict that can be exported as a PDF summary. This document details the final recommendation, weighted scores across options, justifications from individual advisors, points of friction, and actionable caveats identified during the discussion.
Architecture and Agent Brains in Claude Code
The project is implemented as a skill within Claude Code using subagent configuration files stored under a dot-claude directory. Each advisor relies on a dedicated brain folder with markdown files defining principles, frameworks, positions, voice characteristics, and source materials. In Alex Hormozi's case, this knowledge base was constructed by scraping 124 public talks and videos totaling more than 1.3 million words.
Terminal Execution Versus Visual Dashboard
While Herk displays an animated web interface depicting advisor avatars talking and connecting through debate lines, he emphasizes that the visual layer is optional. Claude Code can run the exact same multi-agent debate session and generate all summary markdown files, votes, and final recommendations entirely within the terminal.
The Bottom Line
The video demonstrates how Claude Code can coordinate multiple specialized subagents to deliberate, vote, and generate structured strategic decisions. It establishes the folder organization, personality knowledge structures, and execution methods required to build an advisory board workflow. The demonstration leaves open how the system resolves persistent deadlocks among advisors or how users can expand the council to include custom, private data sources.
FAQ
What is the Roast Room AI board of advisors system created by Nate Herk?
Roast Room is a multi-agent framework built with Claude Code that simulates an advisory board of prominent business figures to debate strategic dilemmas, vote on options, and output actionable business recommendations.
Which business figures are simulated as council members in the Roast Room Claude Code skill?
The council consists of seven simulated advisors: Jeff Bezos, Paul Graham, Alex Hormozi, Gary Vaynerchuk, Dan Martell, Gino Wickman, and Warren Buffett.
What types of files and information are stored in each advisor agent's brain directory?
Each advisor's brain folder contains structured markdown files that document their specific books, frameworks, positions, decision principles, sources, stories, and voice style.
What deliverables does the Roast Room council produce after completing its debate and voting?
The system generates a detailed verdict document and a formatted PDF detailing the council's recommended decision, weighted average scores, advisor breakdowns, caveats, and next steps.
Can the Roast Room multi-agent advisory system be executed without the web visual interface?
Yes, the system is fully functional within the Claude Code command-line environment, which can run the debate rounds, compute the votes, and produce all deliverable files directly in the terminal.
Worth watching for
Entrepreneurs, software engineers, and AI developers interested in building multi-agent systems, simulated personality models, or automated decision-making workflows using Claude Code.
- claude-code
- ai-agents
- prompt-engineering
- multi-agent-systems
- business-strategy