now
What I am working on now
August notes on the active research campaign and the move into UChicago Financial Mathematics.
Current focus
Preparing for UChicago Financial Mathematics. The immediate study stack is probability, statistics, linear algebra, options, stochastic calculus, numerical methods, and volatility modeling. The BTC/USD cross-venue price-discovery study is complete and published.
Recently shipped
BTC/USD cross-venue price-discovery v1.0.0: a confirmatory econometric study of Coinbase vs Kraken using synchronized market data, VAR/VECM, Granger causality, impulse responses, Gonzalo–Granger component shares, and Hasbrouck information-share bounds across 9 usable paired sessions. Coinbase showed stronger short-horizon leadership; long-run price discovery was heterogeneous.
Also shipped
BayesAudit v1.0.0: a reproducible research framework for budget-constrained oversight in hierarchical LLM workflows, with Bayesian monitoring, strategic attackers, matched experiments, provider-cost controls, and 1,945 automated tests. The writeup preserves the held-out nonreplication rather than smoothing it away.
Honest finding
The cross-venue study found asymmetric short-horizon predictability but not uniform long-run leadership. Cointegration rank varied across sessions, and Gonzalo–Granger/Hasbrouck measures were reportable only for the four rank-1 sessions. Those sessions did not consistently favor one venue.
Reading
Currently: Gatheral, The Volatility Surface. Next: de Prado, Advances in Financial Machine Learning, after the immediate Financial Mathematics prep is stable.