Scientist · ML Engineer · Researcher
Terry Feng
I research and work on agentic reasoning and multi-agent systems - at the intersection of AI, game theory, and statistical estimation. Previously at Reddit and NextRoll where I built and optimized large scale deep learning and ad auction systems. I also share technical thoughts, research and learnings about such topics here.
Selected articles
All writing- August 27, 202616 min
LLM Autobidders in Repeated Auctions
No regret autobidders do not learn to tell the truth in repeated second-price auctions, but they do in first-price auctions. LLM bidders tend to play it safe, and don't find the over-reporting exploit.
AgentsMarketsGame TheoryLLMExperiment - August 9, 202615 min
The Sampling Agents' Equilibrium
From crowded cafes to autonomous agents. A primer on sampling equilibrium with statistical inference (SESI), and how controlling information shapes the broader economy.
Game TheoryMechanism DesignEstimationMarketsAgents - August 7, 20265 min
Keep It Simple, Scale
A short reflection on the Bitter Lesson and how it's come up in different contexts.
EstimationThoughts