All Articles By

Hamza Mostafa

Hamza Mostafa is a builder and researcher working on long-horizon agents, multi-agent systems, and reinforcement learning for practical agent workflows. Previously, he worked on the Agent team at OpenAI, where he focused on building and evaluating frontier agent systems such as Deep Research and ChatGPT Agent. His current work explores how agents can operate over extended tasks using tools, state, feedback loops, and measurable task outcomes. He is especially interested in the practical constraints of useful agents: evaluation design, tool use, memory, environment feedback, and the gap between impressive demos and systems that can reliably operate in the real world.

I was in OpenAI’s first intern cohort. Here’s what it taught me about becoming an AI-native engineer