OpenAI Exec: The Merger of ChatGPT and Codex Leads to the Ultimate Goal of Personal AGI
OpenAI's product lead, Thibault "Tibo" Sottiaux, overseeing ChatGPT and Codex, reveals the company's ultimate vision is to merge these tools into a single platform: a "personal AGI." The merger is not superficial integration but a unification of the underlying technology, agent framework, and user interface.
The strategy positions Codex, initially seen as a coding tool, as the "hands" that execute tasks by converting natural language into actions (code, API calls, browser operations). ChatGPT provides the conversational understanding. By combining them, OpenAI aims to build a "super app" where the interface adapts to the individual user, not the other way around. This personal AGI will be a unified, multi-modal (voice-first) system usable by everyone, from developers to casual users.
Sottiaux discusses the rise of AI agents as the new competitive battleground, citing Anthropic's gains with Claude Code. He notes that past OpenAI agent attempts failed because the technology wasn't ready. Now, with improved models and a significant speed boost from "Ultra Fast" inference, real-time, iterative collaboration with an AI becomes possible, preserving user focus and "flow."
Reflecting on his time at Google/DeepMind, Sottiaux recounts how an internal chat model (LMChat/Sparrow) pre-dated ChatGPT but was never publicly released due to organizational silos and misaligned metrics. He emphasizes learning from this: to move bottom-up and be willing to disrupt successful products.
The article concludes with a symbolic detail: a physical button on Sottiaux's desk that resets all developer usage quotas—a gesture of control and generosity from the man steering the product used by nearly a billion people weekly.
marsbit08/31 10:14