AI trading in real markets

Nof1, an AI research lab, argues that financial markets are a strong domain for frontier AI research because they are open-ended, consequential and measurable, and that current LLMs are poor at them. The article says markets are non-stationary, so past successes often fail later, and that they can serve as both an evolving benchmark and a training environment for adaptability. Nof1 says it is building new models for financial markets, along with environments and benchmarks to train and evaluate them. The lab says it has identified a proto scaling law for financial intelligence, with signs of skill transfer to other domains. Research is led by Julian Togelius, founder of NYU's Open-ended Learning Lab.