OpenAI engineer Thibault Sottiaux, who helped build Codex into one of the company’s fastest-growing businesses, is now leading a broad overhaul of ChatGPT’s behavior.
The shift matters because model behavior is becoming a product discipline, not just a research preference. How an AI system asks, refuses, explains, and takes action can decide whether users trust it in daily workflows.
It also suggests OpenAI is pulling lessons from coding agents into the wider ChatGPT experience, where reliability and task execution are becoming central.