Ahmad Khan
writing
blogs
investments
Designing an RL Environment for Computer-Use Agents: The Expense Report Problem
How to Search 100M Lines of Code in 200ms
How We Reduced Database Write Amplification by 11x: From B-Trees to LSM-Trees
The Hard Problem of Robot Intelligence: Why World Models Will Define the Next Decade
Reinforcement Learning in Coding Agents: Multi-Tool Orchestration in long-horizon tasks.