8 July 2026

2 clips · 5 minutes of listening · 1 show

3 shows published 3 episodes in the period, 2.4 hours in all. These are the 5 minutes worth hearing.

Investment

Flirting with Models

Ben Wellington, head of complex feature engines at Two Sigma.

Episode

Ben Wellington – Complex Feature Engineering at Two Sigma (S7E32)

Wellington did a PhD in natural language processing at New York University and joined Two Sigma in 2007, when it was about 125 people in a Soho loft, starting on the data engineering team. He spent more than a decade on how text predicts markets and now runs the teams that build the feature layer that Two Sigma's models forecast from.

Hosted by

Corey Hoffstein, co-founder and chief investment officer of Newfound Research.

6 July 2026 · 55m 54s · 2 clips below

1 of 2 contrarian

Two Sigma's Ben Wellington argues that a push-button AI research tool lowers the entropy of what a quant team produces, and rising correlation between models is the one thing a multi-model portfolio cannot absorb.

Automation suits a cardboard box coming off an assembly line and not a product whose value is orthogonality; what keeps Two Sigma's hundreds of models apart is the individual researcher behind each hypothesis. Their answer is tools that carry each person's own context, so a physicist and a computer scientist on the same dataset still land somewhere different, with two researchers getting the same answer treated as a defect.

2m 26s · from 33:37
My observations
- The test he applies to a tool transfers directly: give it to two people with different backgrounds and the same data, and if the answers converge, the tool is flattening them rather than amplifying them.
- This moves the diversification argument up a level, from correlation between signals to correlation between the researchers who generate them. It also cuts against the headcount case for these tools, since the value he describes comes from the people being amplified.
- He runs a team whose worth rests on human ingenuity in the feature layer, so the conclusion suits him. The claim is checkable from outside: if cheap feature generation is commoditising alpha, decay rates on newly deployed signals should be shortening across the industry.
2 of 2 framework

Two Sigma's Ben Wellington describes using AI to run company-specific analysis across 3,000 companies at once, aiming systematic research at the deep single-name work that has belonged to discretionary analysts.

A feature covering one name has always looked worthless next to one covering a hundred, so the sample-size instinct pushed quant teams towards cross-sectional data. What Wellington calls idiosyncrasy at scale is the claim that AI removes the trade-off: reason down to what is peculiar about one business, then generalise the shape of the reasoning, so what is run for each name stays specific to it.

2m 08s · from 48:17
My observations
- For an allocator this is the line to watch: the deep single-name read is most of what justifies a discretionary fee, and a systematic shop is now pointing scale at it.
- Company-specific reads still have to be right often enough to survive aggregation, and he offers no evidence that they are. He describes this as an expansion they are thinking about rather than a result.
- Put the question to any systematic manager pitching AI capability: how many live features are company-specific rather than cross-sectional? That number separates the claim from the demonstration.
- quantitative-investing
- manager-selection

8 July 2026

Investment

Two Sigma's Ben Wellington argues that a push-button AI research tool lowers the entropy of what a quant team produces, and rising correlation between models is the one thing a multi-model portfolio cannot absorb.

Two Sigma's Ben Wellington describes using AI to run company-specific analysis across 3,000 companies at once, aiming systematic research at the deep single-name work that has belonged to discretionary analysts.