AI Fund House
A multi-agent paper-trading fund. AI agents act as analyst, risk manager and CIO — code enforces the risk limits, and I approve every trade.
The question
Can a team of AI agents make disciplined investment decisions — if the rules that protect the portfolio are enforced outside the model, and a human keeps the final say?
Agents are good at reading markets and arguing a case. They are less good at saying no. So the fund is built around that gap: the agents propose and debate, while plain code and a human approval hold the line.
How it works
Early results
The fund is in its first weeks of paper trading, and early performance is positive. I'll publish the full results in about a month — returns against a benchmark, drawdown, and notes on what the agents got right and wrong.
Get notified when the results are out →What I've learned so far
The build guide walks through every step, including the mistakes I made along the way.
Read the full build guide →What's next
Publish the first month of results, then test the classic trend-following signals I'm studying — Livermore, O'Neill, Minervini and Qullamaggie — as new inputs for the agents.