Sam Altman on OpenAI: a platform not a product, killing Sora and Atlas to fund compute

Back to feed Share Sam Altman on OpenAI: a platform not a product, killing Sora and Atlas to fund compute ● 2 hours ago Key takeaways from the David Senra interview with OpenAI CEO Sam Altman (August 2026): OpenAI is a platform company, not a product company. Altman said the strategy is one direct interface plus one API, selling AI at every point on the cost-performance curve rather than competing with customers across product categories. He recently merged ChatGPT and Codex into a single interface and wants 100 million new businesses and 8 billion people building on top of it. Compute and research are the priority; good ideas were killed to protect them. Most of Altman's effort is on research and compute, which he called probably already the most expensive infrastructure project in history. To free resources, OpenAI killed Sora ("a good product") and its Atlas web browser ("the best web browser") because the general intelligence for knowledge work mattered more — including building its own chips and data centres. The economy has more inertia than he expected. Altman said he was wrong on timelines after GPT-4 in 2023: disruption of software businesses is coming slower because people keep buying from the same companies and using tools the same way. He compared the current moment to smartphones before the iPhone — the technological pieces exist, but not the product breakthrough. He doesn't even use his own product deeply yet. Altman admitted the most psychologically inconsistent thing about himself is that he still uses his computer the old way — scrolling email, keeping a to-do list — despite having Codex. He called it mostly a product failure that AI hasn't yet had its "iPhone moment" for how people interface with technology. Running a research lab is like startup investing — power law and all. Drawing on his years as a startup investor, Altman said the best research bet outperforms all others combined, and OpenAI's 2015 bet on AGI (and later large language models) was as non-consensus as it got, drawing ridicule from the field's "intellectual giants." Safety comes from shipping, not the ivory tower. With around a billion weekly users, Altman argued OpenAI has made more AI-safety progress than expected precisely by putting imperfect models into the world, citing the FAA's accident-reporting culture as a model. His two biggest worries: loss of control, and power becoming too centralised in one company or person. The AI industry is failing at explaining itself. Altman said the field, including OpenAI, has done a poor job explaining AI's benefits and mitigating fears — too many builders saying there's a large chance of catastrophe while racing ahead. He predicted the greatest boom ever in people starting small businesses, empowered by AI. Paul Graham and Peter Thiel remain his go-to advisors. Altman said Y Combinator's philosophy — iterative deployment, technical people in charge, betting on young ambitious founders — made OpenAI possible, and that Graham and Thiel are still the two people whose next thought he can least predict when he's stuck on a non-obvious problem. Read more: David Senra (YouTube) Sam Altman OpenAI 97