Stoic Point co-founder Raj Shah says AI lets his two-person, concentrated hedge fund do research a lean firm could not otherwise afford.
Raj Shah says LLM-native screeners give inconsistent results, so Stoic Point starts with a Bloomberg quantitative screen before passing a reduced universe to qualitative tools like Portrait and AlphaSense.
Raj Shah argues AI makes junior analysts more valuable and could make an AI-native hire a better next step than a back office role.
AI Gives Two-Person Hedge Funds the Reach of Large Teams
AI is giving small, concentrated hedge funds the research capacity that once required a large team. Raj Shah's two-person firm uses it for screening, modelling and monitoring, while keeping stock selection and judgment human. The open question is whether alpha survives once every investor runs the same tools.
Takeaways
AI lets a two-person fund do more research. Stoic Point holds 10 to 15 names, with eight or nine making up about 80% of exposure and a median hold near three years. With two people, time is the costliest input, so AI covers screening, deeper research and monitoring.
LLM screens are inconsistent, Bloomberg is not. Two partners running the same prompt got different dozens of companies. Raj says LLMs look for the shortest path to an answer. Bloomberg's EQS screener returns identical results each time and covers the whole investable universe.
Automated monitoring catches signals a lean firm would miss. Claude takes the portfolio file and asks Portrait for the next five to seven peers of a new holding, and AlphaSense watches them. An alert showed Brunello Cucinelli naming Luxe Experience, a Stoic Point holding, as a growth avenue.
Junior analysts become more valuable with AI fluency. Raj says a junior's added value can far exceed the past. Juniors need AI skills from day one, while judgment, which AI mostly lacks, still comes from experience. His next hire may be an AI native who automates tasks, not the planned back office role.
Widespread AI use could erode alpha. Brett cites rumours that LLMs could reach a 55% batting average on earnings calls. Raj says that would undermine itself as investors converge on one answer, as happened in the earlier data-driven era when early adopters of new data sets generated alpha.
Humans pick stocks, AI manages risk, but this is contested. Raj would reverse the usual split, using AI for portfolio and risk management and keeping human stock picking as the source of alpha. Brett argues idea generation through pattern recognition can be automated too.
Rising volatility favors lean, long-horizon managers. Raj expects volatility to rise with 24-hour trading and retail options activity, rewarding cool heads with long horizons. Regulation has made funds harder to run, but he predicts the next billion dollar company could be a one or two person shop.
Stoic Point's screening funnel
Run a quantitative screen on Bloomberg, which Raj calls the sleeper in the AI race. Claude is strong at writing the BQL code, though two-factor authentication makes automation harder.
Cut the universe, for example 500 names from 1,500, and pass it to qualitative tools such as Portrait and AlphaSense.
Tap a name to go there · tap a line to see why they are linked
On this episode 3
Show 1
Portrait turns natural language queries into screens. Data providers act as walled gardens, so the interface between tools stays disjointed.
Run about 50 screens, then use a meta screen that flags companies appearing on seven of them.