I researched on using interpretable AI systems with controlled memory, then stress-tested how memory fails to better understand how models can use it more effectively. My research interests right now are mostly model architecture, mechanistic interpretability and agentic AI.

I run Speculative Commons — we find asymmetric philanthropic bets that matter most before consensus, and make sure the right money follows.

May or may not be doing something in stealth. It may or may not involve giving drone swarms a mind of their own and letting them figure out the rest. If that sentence didn't make you nervous and excited at the same time, we probably can't be friends. Onwards and upwards, literally.

I write Silicon Macro on Substack, A research newsletter covering global macro and equities. My focus spans the entire Technology, Media, & Telecom value chain, alongside monetary policy, capital flows, fixed income, FX dynamics, and broader equity indices.