Lily Zhang | Posts |

Introducing LosslessBench 07.28.26

A Constitution-Grid Instrument for Data-Efficient RL Alignment (C-Guard) 07.21.26

My Take on the $100B Market for AI Training Data, Including RL Environments 07.13.26

Second-Order Effects After Cursor and Cognition Cracked Post-Training 07.12.26

Reflection: Where Is the Market for Post-Training as a Service? 07.08.26

AI Engineer World's Fair: Is Speculative Decoding All We Need? 06.30.26

Build a Dense Reward for Your Writing 06.26.26

What Can't Be Measured Can't Be Solved: RLVR Environment Design from a Failure Taxonomy 06.10.26

Thoughts on Thinking Machine's Interaction Model 05.20.26

Skill Claw: Self-Improving Robot Agents 03.15.26

SuperGeneral: Compositional Tool Environments for Agent Training 03.10.26

Frontend Slides: AI-Native Presentation Generation 02.27.26

SofaGenius: Multi-Agent ML Research Assistant 02.21.26

IEEE MOST Panel: Driving Force Towards Large-Scale AV Commercialization 05.05.25

Pain is inevitable but suffering is optional

— Haruki Murakami