An investment fund as a function
I joined Coatue with the belief that AI agents could have an even bigger impact on investing than coding. This is because AI is the greatest information processing technology ever developed, and an investment fund is ultimately a state machine that processes information into decisions. You can think of a fund as a single function...
investment_fund(information) → decision
-
info = news, filings, prices, data, risk limits, and all other information
-
decision = a single tuple (stock selection, size, time).
However this can't be as simple as connecting all this data to a model and asking it "give me a great decision tuple" we must break down the problem into sub-functions that we can evaluate.
To borrow from linear algebra, one can think of these sub-functions as a series of lossy projections into lower dimensional space, and there's an effectively infinite number of viable ways to define and sequence these sub-functions e.g.
-
start with a list of stocks and their most important KPIs → identify the themes most likely to move those KPIs
-
start with themes in the market → identify the stocks most exposed to those themes
Even defining the problem is an incredible challenge. This series of sub-functions can be represented of as a massive computational graph where the optimal shape is shifting at a blistering pace with each new model release.
And even after defining all of these sub-functions there are STILL an infinite number of ways to run each one. e.g.
-
one agent per KPI vs one agent per stock
-
one agent reviewing the output vs 10 review agents each running different models with a majority vote at the end
-
same 10 agents, but with the vote weighted by model size and thinking time.
Our objective is absolute performance, not maximal AI usage. So oftentimes deterministic code, statistical methods, or humans are the best way to run each sub-function.
Our job as practitioners is to generate the optimal definition of this computational graph, so we can reinvent the world's ultimate state machine, in the most competitive industry in the world. I see this as one of the great projects of the AI age, and I am excited every day to get to show up to work and give it our best shot!



