Scaling Laws for Market Prediction
We examine how model capacity, data diversity, and compute interact when forecasting high-frequency and medium-horizon returns.
Read more →
RazorLake is a frontier lab advancing the science of markets through AI, rigorous experimentation, and deep domain expertise.
Peer-reviewed thinking and technical notes on foundation models for markets, reinforcement learning in execution, and the limits of predictability.
We examine how model capacity, data diversity, and compute interact when forecasting high-frequency and medium-horizon returns.
Read more →A framework for learning execution policies that remain robust under regime shifts and changing liquidity conditions.
Read more →Why research taste, hypothesis generation, and scientific skepticism remain decisive even as models grow more capable.
Read more →We treat markets as complex adaptive systems. Our work combines large-scale foundation models, causal inference, and carefully designed experiments to produce trading systems that are both high-performing and interpretable.
Every model is stress-tested across regimes, subjected to adversarial evaluation, and continuously monitored in live environments.
Learn more about our processPre-training and fine-tuning large models on multi-modal market data.
Moving beyond correlation to understand what actually drives returns.
Continuous validation, risk constraints, and human-in-the-loop oversight.
Publishing methods and findings that advance the broader field.