Sep 12, 2026 · 1h 9m
AI and discretion test systematic trading’s limits
SI417: What Happens When AI Starts Trading the Markets ft. Rob Carver
The episode examines how robust rules, investor communication, and market structure shape outcomes as trading systems become more complex and AI-assisted.
- 1Rob Carver argues that systematic trading should isolate any discretionary skill while keeping sizing and exits rule-based.
- 2AI may accelerate quantitative research, but widespread misuse could amplify speculative behavior and reinforce existing market narratives.
- 3Trend-following speed depends on the shape of market moves, making volatility alone an unreliable guide to strategy design.
Don't miss
Rob Carver explains why AI could improve quantitative research while also encouraging crowded, speculative behavior when traders use it poorly.
The brief
Rob Carver and Niels Kaastrup-Larsen begin with bond-market pressure, Treasury buybacks, and short positions before turning to what changing flows mean for systematic investors.
A discussion of Simplify’s Managed Futures Strategy ETF becomes a broader warning: reducing diversification or changing a strategy can reset investor expectations as much as performance does.
Carver defends robust, testable systems over intuition, arguing that genuine discretionary skill should be isolated while position sizing and exits remain rule-based.
The sharpest tension arrives with AI: it could speed idea generation and testing, yet widespread misuse may amplify speculation and produce crowded, derivative signals.
The closing volatility debate rejects simple rules about speed, showing how V-shaped reversals and sustained trends can reward opposite trend-following systems.
What was said on this episode
13 statements · 4 positive · 6 negative · 3 neutral
AI may overfit quantitative research during the research phase.
“in the research phase, I think the biggest risk is potentially that it will just overfit”
Listen at 46:22
AI-generated quantitative research needs substantial non-AI code guardrails.
“You need to essentially wrap it in a... in a lot of code that isn't written by AI”
Listen at 46:35
An AI code review removed short positions, leaving only long positions.
“the code essentially would no longer have short positions, just long ones”
Listen at 47:49
AI use is acceptable when users diligently check everything it writes.
“if you're using it and you are diligently checking everything it writes yourself”
Listen at 48:08
Oversight and governance processes are essential for automated trading code.
“the oversight governance and processes around code implementation are absolutely key”
Listen at 49:10
The market effects of widespread consumer-grade AI trading remain unknown.
“What will be the effect of a wave of people using, you know, consumer-grade AI models on the markets? Will it, for example, will it increase turnover in markets? Will it increase risk? Will it increase volatility? I don't know, is the answer.”
Listen at 50:02
Widespread AI trading will affect financial markets.
“I think it will have an effect on markets.”
Listen at 50:40
AI-driven trading could temporarily make markets irrational or disorderly.
“markets just could get quite silly”
Listen at 50:51
Tesla is currently the world’s biggest meme stock.
“tesla is the biggest meme stock in the world”
Listen at 51:05
A large group of uninformed AI-driven retail investors will emerge.
“I think we're going to see a herd of uninformed AI-driven retail investors.”
Listen at 51:33
Informed traders may find opportunities against AI-driven retail investors.
“I think there'll be opportunities there.”
Listen at 51:39
AI systems recommending trend following learned from internet content.
“it's been trained on reading the internet”
Listen at 52:17
Online commentary advocates allocating more to CTAs and trend following.
“there are lots of people saying, including myself, and you, that people should have more in there, more CTAs, more trend following in their portfolio”
Listen at 52:20
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.
Books & mentions
Some links are affiliate links — PodLume may earn a commission if you buy.
