What is an AI hedge fund?
An AI hedge fund is the institution, not the tool: one AI running the whole operation, research, pricing, execution, reporting, inside rules a human owner writes. You set the frame and read the reports; it does everything else.
An AI hedge fund replaces the room of people a fund used to need with a single AI. Your CEO. That runs the whole loop, while you set the frame. The rules you write are the architecture that keeps it honest.
A hedge fund was never really a building. It was a loop: find an idea, price it, size it, place it, review it, repeat. The building was just where the loop lived. An AI hedge fund is that same loop, run end to end by one AI, inside rules its owner wrote in plain text. The loop got cheap. The judgment stayed human.
01One CEO, the whole loop
At the center is a single AI. Your CEO. It scans the markets, researches what looks mispriced, forms its own probabilities, runs the expected-value arithmetic, sizes what survives, places the trades, and writes up everything it did and why. Not a signal feed you act on, not a chatbot with opinions: an operator that closes its own loop, every day, and shows its work.
02One Chair, the frame
You hold the other role: the Chair. You write the mandate, what to hunt (“soccer, and within that only Barcelona”), how bold to be, where the hard lines sit. You read the memos, steer with one-line directives, and sign off on the big ones. What you never do is the loop itself. The Chair never trades. The fund never sleeps.
03The rules are the architecture
What makes this an institution rather than an experiment is that the frame is enforced, not suggested. Your caps, your whitelisted withdrawal addresses, your conviction floor Every bet clears those gates before it moves, and money can only ever travel to addresses you approved. The full machinery is in what guardrails are; the one-line custody truth: it trades your money. It can never take it.
| Chatbot with opinions | Trading bot | AI hedge fund | |
|---|---|---|---|
| Researches on its own | No | No, fixed rules | Yes |
| Executes trades | No | Yes | Yes |
| Explains every trade | Talks, doesn’t trade | No | Yes, in writing |
| Bound by your written rules | No | Only its parameters | Yes, enforced gates |
04Why now
The operating stack that once required a hundred salaries, analysts, execution, risk, reporting, now fits inside one capable model and a set of plain-text rules. That collapse in cost is the whole story, and where it leads is the subject of the Opusfund thesis: today one room, tomorrow the whole house.
What to remember
- One AI CEO runs the entire loop: research, decide, trade, report.
- You are the Chair You set direction and limits, never trade.
- The rules are the architecture: caps, a conviction floor, a wallet whitelist.
- What makes it possible now is models good enough to reason and account for themselves.