John Schulman

John Schulman is a machine-learning researcher who cofounded OpenAI in 2015, where he helped develop reinforcement learning from human feedback (RLHF) and led the creation of ChatGPT, and who is regarded as one of the principal architects of the alignment methods that made conversational language models usable. He left OpenAI for Anthropic in August 2024, departed Anthropic roughly six months later, and in February 2025 became a cofounder and chief scientist of Thinking Machines Lab, the company founded by the former OpenAI chief technology officer Mira Murati.¹²

Reinforcement Learning and the Road to OpenAI

Schulman studied physics at Caltech, briefly pursued neuroscience, and took his doctorate in computer science at the University of California, Berkeley, advised by Pieter Abbeel, working on robotics and reinforcement learning. During and shortly after his PhD he developed two foundational reinforcement-learning algorithms: Trust Region Policy Optimization (TRPO), with coauthors including Sergey Levine and Michael Jordan, and Proximal Policy Optimization (PPO), published in 2017, which simplified TRPO while preserving its stability and became one of the most widely used and most cited methods in the field.³⁴

He joined OpenAI as a member of its founding team in 2015, immediately after finishing at Berkeley. PPO later became the workhorse algorithm of RLHF, the technique that fine-tunes a language model on human preference judgments to make its outputs more helpful and less harmful, and which OpenAI used to turn its raw GPT models into the aligned systems that culminated in ChatGPT. Schulman co-led OpenAI's post-training team and is widely credited as a leader of the effort that produced ChatGPT after its November 2022 release.⁴⁵

Anthropic, Then Thinking Machines Lab

On August 5, 2024 Schulman announced that he was leaving OpenAI to join Anthropic, the rival lab founded by former OpenAI researchers around AI safety. He wrote that the move "stems from my desire to deepen my focus on AI alignment, and to start a new chapter of my career where I can return to hands-on technical work," and he joined Anthropic's alignment-science team. His departure came amid a wave of exits from OpenAI's safety and research leadership, including the chief scientist Ilya Sutskever and Murati.¹⁶

Schulman's time at Anthropic was brief. In February 2025 he left to cofound Thinking Machines Lab with Murati, taking the role of chief scientist, in a company assembled largely from former OpenAI researchers including Barret Zoph and Bob McGrew. The lab raised a roughly 2-billion-dollar seed round at a 12-billion-dollar valuation in July 2025, led by Andreessen Horowitz, with Schulman among its named cofounders.²⁷