๐ช Our paracingulate advantage
AI Summary
A new study by economist Spyros Galanis tested whether LLM-based AI agents possess a 'theory of mind' by competing in simulated prediction markets. The results show AI agents can mimic deceptive trading behavior like humans but fail to reason about what other traders know from their actions. Frontier models released as recently as late 2025 performed worse than older models, suggesting a persistent upper bound on AI interactive reasoning.
Key Facts
Author Takes
AI theory of mind in trading
LLMs do not have a theory of mind and likely face a persistent upper bound on interactive reasoning, meaning they may never replicate the human ability to deduce what other traders know from observing their behavior.
Related topics
More from The Breakdown
๐ช Friday Charts
Alphabet's blockbuster Q1 2026 earnings added $450B in market cap, with the three major cloud providers collectively holding $1.5T in order backlogs,
๐ช Friday Charts
Byron Gilliam's Friday Charts edition examines whether AI tokens will become more productive and cost-effective than human workers, using data from Me
๐ช Thursday links
This newsletter covers cybercrime statistics showing Americans lost $21 billion to cybercrime in 2025 with crypto used as payment 72% of the time, pre