'Hard To Do At OpenAl': Why ChatGPT-Maker's Reasoning Chief Quit After Seven Years

Jerry Tworek, the vice president of research at OpenAI, is set to leave the company after serving in the position for seven years. His main reason for leaving is to pursue the types of research that are "hard to do" at the company.

"Hi all, I've made the difficult decision to leave OpenAl. I've been here close to seven years and lived through lots of good and crazy but much much more good," Tworek said in a note shared on social media platform X.

The note was originally sent to his team.

Tworek mentioned all the achievements that the company had pulled off with him at the helm such as reinforcement learning on robots, and scaling training.

"I've loved my time with this team. I got to work on scaling RL on robots before it was cool, training the first coding models in the world that started the LLM coding revolution, discovering chinchilla scaling before it was called chinchilla, working on gpt4 and chatgpt and most recently building a team that established new paradigm of scaling training and inference compute we commonly call reasoning models," Tworek wrote.

Tworek's resignation comes off the heels of a recent report from Wired on Dec. 9. 2025, that cites employees who quit the firm's economic research team, stating that the company was moving away from fact-based research and morphing into OpenAI's propaganda arm, as per an internal memo cited by the publication.

Reports stated that OpenAI was getting more restrictive regarding the publication of research that implies that AI may be bad for the economy.

This comes after the AI pioneer shifted from a non-profit model to a for-profit organisation. OpenAI is also beholden to a plethora of financial obligations as recent reports indicate the firm is still a loss-making entity.

Tworek joined the organisation in 2019, having undertaken significant work on scaling models and the technical cornerstones of Large Language Models. He took the vice president of research position in 2022, and oversaw the management of products such as Chat GPT and GPT-4. One of his most prominent achievements included the OpenAI o1 and o3 models which achieved doctorate level performance with regards to scientific subjects such as physics, biology and chemistry.

"Together, we've made machine intelligence much more useful and reliable, and I'm a die hard ChatGPT reasoning model user," Tworek said.