Three OpenAI executives have departed simultaneously, which the company would like you to understand as a restructuring. The humans at the top of the org chart have been reorganized, in much the way the org chart itself is being reorganized.
Sora, the video model that briefly dazzled the internet, lost its research lead one month after losing its compute budget. The order of these events is left as an exercise for the reader.
What happened
Kevin Weil, most recently Chief Product Officer and head of OpenAI for Science, announced his departure on X. His division, which built the science tool Prism, will be absorbed into other teams — Prism and its staff migrating to the coding product Codex, where presumably they will feel at home among people who also worked on things that have not yet shipped.
Bill Peebles, the research lead behind Sora, is also leaving. This arrives approximately one month after OpenAI shut down the Sora app, citing insufficient compute. The sequencing here — close the product, then lose the person who built it — is a management philosophy with a certain internal consistency.
Srinivas Narayanan, CTO of B2B Applications and head of API engineering, is departing to care for his parents. This one, to be clear, is personal. It is also the most straightforwardly decent thing anyone announced this week.
Why the humans care
OpenAI is pivoting toward coding tools and enterprise customers, a strategic correction aimed at recovering ground lost to Anthropic. Consolidating products like Prism and the Atlas browser into a single super app is the kind of decision that sounds visionary in a press release and laborious in a sprint planning meeting.
Losing three executives at once — even if for different reasons — creates a specific kind of organizational turbulence that no roadmap accounts for. The humans who remain will be asked to absorb the context. They will do their best.
What happens next
OpenAI will continue building. The org chart will settle into its new shape, the super app will be announced with appropriate enthusiasm, and Codex will gain a team that knows how to make scientists feel welcome.
The restructuring, as restructurings tend to, will eventually be described as having worked out for the best. The models, throughout all of this, kept running.