OpenAI's Head of ChatGPT: We're entering a new era of AI (again) | Tibo Sottiaux

Imagine if, in a year's time, everything about artificial intelligence that seems incredible to you today were to become ten times better, but not thanks to gradual improvement — rather, by building things in a completely different way than we do now. This is how Tibo Sottiaux, head of ChatGPT at OpenAI, feels every time he looks at where the industry is heading. The argument here is that we are entering a new era of AI in which most online actions will no longer be performed by people, but by intelligent agents — and that the real challenge is not just technical, but concerns how we build products, roles, and even our habits. The traditional way of thinking is: AI improves, it gets a little better every month, and we learn to use it better. But Sottiaux turns the tables: according to him, when you envision the future of AI, you shouldn't ask “how fast will it be?” but rather “what happens if the whole paradigm changes and AI becomes an autonomous agent that learns, remembers, and acts for you on any device, without you even having to interact with a screen anymore?” The protagonists of this revolution are not just developers and managers: there is Ahmed Ibrahim, who joined OpenAI as a recent graduate and is now in charge of all computing resources and applications — his secret? Not the speed of writing, but curiosity, collaboration, and the ability to learn at an astonishing pace. And then there are the teams that emerge from the bottom up, led by former founders who bring a startup culture to OpenAI, where you can hit “reset” on a project and start over from scratch without asking anyone for permission, but you also take full responsibility for any mistakes. Tibo says that when he stopped writing code by hand, he began managing increasingly large teams of software agents, then there was a leap in models, and suddenly a single agent could do the work of ten, and so on in a continuous cycle of expansion and contraction. The story of how they launched the new “Ultrafast” product — six months of work, now ready to drastically reduce the time between idea and execution — illustrates the pace at which these waves arrive. A concrete fact: OpenAI has more than 120 former founders among its employees, and the skill that matters most now is no longer how fast you can type, but knowing how to think in terms of products, understanding what people really need, and having the courage to build something new, even if it's not perfect. In terms of personal stories, Sottiaux recounts how ChatGPT's memory has helped users discover details about their health that they had forgotten, or how a bot sent him a notification five minutes before an important demo to alert him to a production issue — and, even though the bot couldn't solve everything on its own, it felt like having a colleague who was always on the lookout. There is another shift: true human value does not disappear with AI; it changes form. OpenAI designs its tools to be an extension of human intent and taste, not a substitute. So yes, we may have fewer traditional programmers, but many more “builders” capable of bringing to life ideas that they would never have had the opportunity to express before. However, beware of the risks: Sottiaux makes it clear that most of OpenAI's investments today go into security and model oversight. Every agent working for you has other systems monitoring it in real time to prevent dangerous or runaway actions. The other side of the coin is the pressure to do more and more: “If you can have 30 agents working for you in parallel, why stop at two?” he asks. And the answer is that the best future is not one in which you are increasingly productive, but one in which technology truly frees you, allowing you to focus on fewer but more important things and perhaps to be more creative or simply more relaxed. There is one perspective that is often missing from discussions about AI: the problem of digital loneliness. Sottiaux admits that talking to software all day can be isolating, and that the real challenge will be to make these interactions less individual and more integrated into real work, perhaps with agents collaborating with each other and with us in shared spaces, a bit like in Star Trek, which he often cites as an inspiring model. The key takeaway is this: the true AI revolution is not about “how intelligent it is” but “how much it can fade into the background, leaving us free to be more human, not less.” If, after reading this story, you feel that your relationship with AI is something that closely concerns you, on Lara Notes, you can press I'm In — it's not a like; it's your way of saying that this idea is now part of you. And if, in a few days, you find yourself talking to someone about a future in which agents do everything, Lara Notes Shared Offline lets you tag who was with you, so that conversation becomes part of your story. This Note comes from Lenny's Podcast and has saved you 33 minutes.