Sam Altman, over bread rolls, explores life after GPT-5

I look forward to Alcatraz Island from a Mediterranean restaurant in San Francisco with one hundred dollar fish dish on the menu. While I talk to other reporters, Sam Altman from OpenAi jumps through the door on my left. Altman looks down on his bare iPhone to show us everything, and a pushy thought slides out of my mouth: “No phone case is a daring choice.”
Of course I immediately realize that the billionaire -CEO of OpenAi, who employs Apple veteran Jony Ive, gives more to preserve the original design of the iPhone than the $ 1,000 that it costs to replace one.
“Listen, we are going to send a device that will be so beautiful,” Altman said, referring to the upcoming AI device of OpenAi and IVE. “If you put a business over it, I’ll find you personally,” he joked.
Altman had gathered about a dozen technical reporters to join him and other OpenAi managers for a dinner on the record (and off-the-record dessert). The night raised more questions than it answered.
For example, why is Nick Turley, the VP from Chatgpt, who gives me a lamb spoil a week after the launch of GPT-5? Was this to let me write fun things about the largest AI modeling from OpenAi so far, which was relatively disappointing considering the years of hype around it?
In contrast to GPT-4, which far exceeded the rivals and the expectations of what AI can challenge, GPT-5 performs roughly the same as models of Google and Anthropic. OpenAi even brought back GPT-4O and the Chatgpt model selector, after various users expressed concern about the tone of GPT-5 and his model router.
But all night it became clear to me that this dinner was about the future of OpenAi than GPT-5. The managers of OpenAi gave the impression that AI models are less important than when GPT-4 was launched in 2023. After all, OpenAi is now a completely different company, aimed at cleaning old players in searches, consumer hardware and enterprise software.
WAN event
San Francisco
|
27-29 October 2025
OpenAi shared some new details about those efforts.
Altman said that OpenAi’s incoming CEO of Applications, Fidji Simo, would supervise several consumers apps outside Chatgpt – OpenAI still has to be launched. Simo is planned to start at OpenAi in just a few weeks, and she can ultimately supervise the launch of an AI-driven browser that is reportedly developed to compete with Chrome.
Altman suggested that OpenAi would even consider buying Chrome – probably an offer that would be taken more seriously than the Pertlexity offer – if it would become available. “If Chrome really starts to sell, we should take a look,” he said before he looked at us all and asked, “Is it actually going to sell? I assumed it wouldn’t happen.”
Simo can also run an AI-driven social media app that said the CEO of OpenAi that he is interested in exploring. Altman even said that ‘nothing’ inspires him about the way in which AI is used today on social media, and adds that he is interested in “whether it is possible to build a much cooler kind of social experience with AI.”
While Turley and Brad Lightcap, the Coo of OpenAi, largely gave the floor to Altman, drinking wine next to the other sitting guests, Altman also confirmed reports that OpenAi is planning to support a startup of a brain computer interface, Merge Labs, to compete with Elon Musk’s Neuralink. (“We haven’t done that deal yet; I wish we do that.”)
How intertwined that company will be with the models and devices of OpenAi is still to be seen. Altman only described it as a ‘a company in which we would invest’.
Despite all the talk about browsers and brain chips, the elephant in the room remained the raw reception of GPT-5. Eventually the conversation circled back to the model that our group dinner had asked in the first place.
Turley and Altman say they have learned a lot from the experience.
“I thought legitimate that we ruined that,” Altman said about writing GPT-4O without telling users. Altman said that OpenAi users will give a more clear “transition period” when abolishing AI models in the future.
Turley also said that OpenAi is already rolling out a new update to make the answers from GPT-5 “warmer”, but not sycophantic, so that it will not reinforce negative behavior in users.
“GPT-5 was just that time. I like that. I use the robot personality-I am German, you know, whatever,” Turley said. “But many people don’t do that, and they really like the fact that Chatgpt would actually check in with you.”
It is a delicate balance for OpenAi to strike, especially given that some dependencies have developed chatgpt. Altman said that OpenAI believes that less than 1% of chatgpt users have unhealthy relationships – which could still be tens of millions of people.
Turley said that OpenAI collaborated with experts in mental health care to develop a section to evaluate the answers from GPT-5, so that the AI model will return to unhealthy behavior.
That said, it seems that GPT-5 has not damaged the OpenAI company. Altman even said that the API traffic from OpenAi doubled within 48 hours after the launch of GPT-5, and the company is effective “from GPUs” thanks to all the demand.
In many ways, the contradictions of the night-pulling launches, record-breaking use-the strange reality of OpenAi at the moment, reflected.
Given the bets of OpenAI on browsers, brain chips, AI chatbots – and others who make the company around data centers, robotics and energy – Altman clearly has ambitions to run a much larger company than just the chatgpt maker. The final form can look like something like Google’s parent alphabet, but perhaps even wider.
When the night ended, it became clear that we were not collected to think about GPT-5. We were thrown at a company that wanted to outgrow his famous and controversial product.
It seems likely that OpenAi will become public to meet its enormous capital requirements as part of that image. In preparation, I think Altman wants to improve his relationship with the media. But he also wants OpenAi to come to a place where it is no longer defined by his best AI model.




