Sam Altman at Stripe Sessions: What AGI Means for Software

Altman's Stripe Sessions fireside chat (April 2024) was his most developer-focused public appearance — aimed at the infrastructure and fintech builders who are among OpenAI's most important enterprise customers. The conversation's key themes: how AI will change software development (coding will move from writing code to specifying what you want at an increasingly high level); why the model capability ceiling keeps moving faster than most people's model of it (test-time compute scaling changes the relationship between model size and capability); and OpenAI's competitive moat as an API provider (not architecture, which is increasingly commoditized, but data, trust, and the feedback loop from millions of deployment environments). Altman was unusually candid about the competitive dynamic: he described the AI race as a genuine sprint with multiple capable players, and said OpenAI's ability to stay ahead depends on maintaining technical leadership while scaling the enterprise business fast enough to fund the research. He also gave his clearest statement of the "AGI timeline as software question" — that the remaining obstacles to AGI are engineering challenges, not fundamental scientific unknowns, and that the pace of progress on those engineering challenges is faster than it looks from outside OpenAI. The Stripe audience's focus on revenue and reliability produced questions about API stability, pricing, and uptime that revealed the tension between OpenAI's research mission and its commercial obligations.