Sam Altman has a confession: he's the CEO of OpenAI, the company building some of the most capable AI models on Earth, and he still manually clears out his email and to-dos. Even though he has Codex.
He blames the awkwardness on a "product problem." The models are already powerful, but humans haven't figured out where they should click the computer themselves and where they should hand it over to the AI.
In Altman's view, today's AI is a lot like smartphones before the iPhone. The necessary pieces are mostly there, but the product that truly changes how we interact with the machines hasn't arrived yet. AI, he says, hasn't had its "iPhone moment."
What he wants to work on now isn't just model intelligence — it's context. Altman wants AI to read through all the OpenAI internal Slack messages he'll never get to, the customer success and failure stories from ChatGPT usage, and more papers — then surface the genuinely relevant information when he's making decisions.
His read: the industry has spent years making models smarter, while underestimating something else. Getting AI to read tens of thousands of pages in seconds and then use that knowledge to help people make decisions. That's a capability humans simply can't match.