Hedge funds have always had the research, the discipline and the edge, and a minimum cheque that keeps everyone else out. We are a research lab working to change that, with AI that has to prove itself before it is trusted with a single dollar.
Follow our research ↗Why can't making money be automated? And why should it stay behind big pools and closed doors, when it could be open to everyone?
Eight AI analysts debated investment decisions, then competed with simulated portfolios. Debate looked persuasive. Nothing proved it was right.
Four analysts ran in parallel, a judge picked, and trades executed on about $1,000 of volume. Real execution exposed how little an LLM's confidence is worth.
An AI hedge fund for consumers, after selling ReeF, an anime game. We stopped midway. We hit the wall.
Not "can an LLM trade?" but "what must a model prove before anything it says can touch money?"
A model has no feedback from the market. It can sound certain and be wrong, and never know.
What it knows ends at a cutoff. Markets don't. Anything it "remembers" may be leaked or stale.
Seconds of inference are a lifetime for any edge that depends on being first.
It reads a snapshot, then reasons about a world that has already moved on.
