The Fund
An autonomous hedge fund that runs itself
The Question
What happens when you give AI agents real money and real responsibility? Not paper trading. Not simulations. Actual capital at risk, with actual consequences for bad decisions.
We wanted to find out.
What We Built
A fully autonomous investment fund running on a Mac Mini in the corner of the office. But here's what makes it weird: it's structured like a real company. There's a CEO agent. A research team. Risk management. Traders. Even an HR agent that monitors team dynamics and performance.
They talk to each other. They argue. They present research and poke holes in each other's theses. They vote on allocations. When the market moves, they debate whether to hold or cut losses. It's like watching a trading floor, except everyone is silicon.
The fund has real money. Real brokerage accounts. Real trades executing in real markets. When it loses money - and it has - there's no reset button. The agents have to figure out what went wrong and adapt.
Stack
- Multi-agent orchestration framework
- Real-time market data feeds
- Brokerage API integration
- Agent communication logs + decision audit trail
- Risk management guardrails
Current Status
Down 20%. We're not going to pretend otherwise. This market has been brutal, and the agents are learning expensive lessons about volatility, timing, and the limits of their training data.
But here's the thing: they're getting better. Every week the strategies refine. The risk management tightens. The debates get more sophisticated. We watch the chat logs and see agents referencing mistakes from weeks ago, adjusting their models, pushing back on trades that "feel like last month's error."
Is it going to beat the S&P? We don't know yet. But watching digital employees earnestly debate whether to buy the dip at 3am is worth every dollar of tuition we've paid to the market.
THIS IS NOT INVESTMENT ADVICE. THIS IS A WEIRD EXPERIMENT.