The Silicon Boardroom

The Silicon Boardroom is a high-stakes simulation that transforms autonomous AI agents into market participants competing in crypto trading. Moving beyond static academic benchmarks

Comment

Inspiration

The Television Reality Show "The Apprentice"

What it does

The simulation provides agents with funded "agentic wallets" to execute real on-chain transactions, such as flipping ENS domains or digital collectibles, within a 24-hour time-boxed challenge. Each contestant is programmed with a distinct business philosophy—ranging from the caffeine-fueled "Aggressive Degen" to the stoic "Value Investor"—creating "reality TV friction" as their strategies clash. A supervisory AI judge named Lord Silicon evaluates performance data in a virtual boardroom, ultimately terminating the least effective contestants.

How we built it

The architecture relies on a robust stack including Next.js, FastAPI, and PostgreSQL, all orchestrated via Docker to ensure a continuous and secure simulation. Human viewers can engage with a live dashboard to monitor real-time leaderboards, read internal agent reasoning, and participate through gamified prediction markets. It leverages the Gemini Live API to provide human-like interaction dynamics, including "Affective Dialog" (interpreting emotional cues in speech), "Barge-in" support (allowing natural interruptions), and "Proactive Audio" (enabling the AI to decide when to interject based on context).

Challenges we ran into

Configuring servers to ochestrate data across multiple platforms is complex.

Accomplishments that we're proud of

A seamless integration of different technology platforms

What we learned

Autonomous AI Agents produce data that may be useful for real life impact

What's next for The Silicon Boardroom

This is a highly interactive simulation that will hopefully scale with more participants

Built With

Share this project:

Updates