An LLM-driven quant fund. Language models read what the numbers leave out; the system sizes it, trades it, and nobody overrides it. Every position it has closed is published below.
01 — Results
A $10,000 book, marked against every closed position and benchmarked to SPY over the same period. Positions still inside their holding window are excluded until they settle.
02 — The fund
SpecuLens is a quantitative fund with a language model at the front of it. The model reads the filings, the coverage and the tone around a company; the system turns that into a position, sizes it, and holds it to a fixed horizon. Roughly two weeks.
Classical quant works on what can be counted — price, volume, multiples, momentum. But most of what moves a stock over a fortnight arrives as text, and text is where the overreaction lives. Markets overreact to news and the overreaction decays; reading that news across a whole universe, every session, without getting tired or talked round, is the part that was missing.
So the fund does not ask anyone to be patient. The model's judgement is weighed against fundamentals and price behaviour, sized, and held long enough for the market to correct itself. There is no manual override. Nobody talks it out of a position, and nobody talks it into one.
It is honest about what it is: a single strategy with a specific edge that will stop working when enough capital finds it. It wins slightly more often than it loses, and the wins are slightly larger. What follows is the whole record — not a backtest, not a selected window. Every closed position since inception, marked at the price the market actually printed.
03 — Trade history
The 50 most recent closed positions. Return is what the position made or lost over its holding window.
| Ticker | Entered | Return | Outcome |
|---|---|---|---|
| Loading… | |||