AlphaMaven AlphaMaven

Don’t Let AI Mark Its Own Investment Homework ft. Rob Carver

Reported by Top Traders Unplugged
Nov 21, 2026· 2 min read
Save Post LinkedIn Email
Key Takeaways
  • • Top Traders Unplugged episode 427 (dated 21 November 2026) hosted by Niels Kaastrup-Larsen with trend follower Rob Carver examines the risks of using AI to build, test and validate investment strategies.
  • • The discussion centers on overfitting, look-ahead bias in backtests, and the danger that AI-generated backtests are declared brilliant without independent checking.
  • • Carver and Kaastrup-Larsen raise concerns about market crowding and declining alpha if AI-powered research becomes widespread among managers.
  • • The episode covers the trade-off between model complexity and robustness, and where human oversight remains essential in research and portfolio construction.
  • • The source does not publish specific performance figures, AUM values or deal terms; its content is a podcast discussion rather than a news report.
Summary
In Episode 427 of the Top Traders Unplugged podcast, dated 21 November 2026, hosts Niels Kaastrup-Larsen and Rob Carver address whether artificial intelligence can be trusted to evaluate its own trading strategies. The episode description frames the core problem directly: a model can build a strategy, write the backtest and then declare the results impressive, leaving open the question of who checks the work. The hosts characterize faster AI-assisted research as capable of producing faster mistakes as well as faster insights.

The conversation addresses several technical risks. Overfitting, where strategy parameters are tuned until historical results look convincing but fail out of sample, is a central theme. The episode also covers look-ahead bias in backtests, and the timestamps indicate a closing segment on building more trustworthy backtests. The hosts distinguish between genuine trading edges and what the publisher calls convincing illusions, a distinction they argue becomes harder as AI tools make complex models easier to generate.

The discussion extends to market structure. The hosts question whether widespread use of AI-powered research could lead to crowded trades and declining alpha, as managers arrive at similar models. They also examine whether AI can generate genuinely original investment ideas, and the dangers of self-driving portfolios. Trend following is discussed both as a standalone approach and as a complement to a concentrated investment portfolio, and the episode touches on the latest CTA performance and Carver's own diversified trading portfolio in volatile markets.

The episode description does not provide quantitative results, fund figures or named regulatory actions. Its value to institutional readers lies in the framework it offers for validating model-driven research, with human oversight positioned as necessary where AI is applied to strategy development.

Go deeper than the headline

Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.

Start your free 7-day trial

Sources

1 outlet

AlphaMaven summarised this story from the reporting below. Each link opens the original article.

Top Traders Unplugged Don’t Let AI Mark Its Own Investment Homework ft. Rob Carver