OpenAI held its annual developer conference in San Francisco last week and announced ‘dots,’ a new AI assistant designed to help users connect information across multiple sources. The timing is impeccable: the company is still explaining why its latest flagship model won’t ship because safety reviews keep finding problems it can’t solve.
Sam Altman took the stage to describe dots as a breakthrough in reasoning and synthesis. The demo showed it connecting disparate data points into coherent narratives. Impressive. Also irrelevant. The real story is that OpenAI’s next-generation model—the one everyone actually wants—remains stuck in limbo because internal safety testing revealed issues the company doesn’t know how to fix. So they pivoted to announcing something shinier.
Why announce a mid-tier assistant when your flagship product is broken? Because the narrative matters more than the product. The Valley’s playbook never changes: when safety concerns threaten to dominate the news cycle, release something that sounds innovative enough to reset the conversation. Dots will be fine. It will be useful. It will also be a perfect distraction from the fact that OpenAI’s safety infrastructure is apparently a bottleneck on its own products.
The press release called dots ‘a leap forward in contextual reasoning.’ It is not. It is a competent feature that existing systems could theoretically handle. But it has a name, a demo, and a story about connecting things—which, in 2026, passes for innovation. The real dots OpenAI needs to connect are between what it promises and what it can actually deliver safely. Those remain stubbornly separate.