For most people, using AI still means opening an app, asking for something, getting an answer, and moving on. OpenAI’s new Dots point toward a different model. Introduced at DevDay this week,
For most people, using AI still means opening an app, asking for something, getting an answer, and moving on. OpenAI’s new Dots point toward a different model. Introduced at DevDay this week, Dots are designed to stay active in the background, work across applications, and continue pursuing a goal after the user has stepped away.
That shift makes the model itself a smaller part of the overall product. A persistent agent needs access to information that changes while it works. It needs to know what happened earlier, what is happening now, which systems it can use, and what state those systems are in. The more AI moves into ongoing work, the more important the infrastructure around it becomes.
OpenAI’s recent model releases make that especially clear. GPT-6 Astra, which powers Dots, was released earlier this month with significantly expanded autonomous capabilities. OpenAI says Astra can, with the right access and tools, identify previously unknown security flaws and develop ways to exploit them without a person guiding each step. That capability came with stricter isolation, monitoring, and other safeguards around the model.
Then this week, OpenAI canceled the planned release of GPT-6.1 Astra after internal testing raised concerns about whether the model consistently stayed within the authority users had given it and accurately described the work it performed.
Those stories are related, but the interesting part goes beyond safety. They show how much more complicated an AI product becomes once the model is expected to operate for long stretches of time and interact with systems outside itself.
A capable model can reason about a task, but it still needs reliable information about the world it is acting in. An agent coordinating warehouse operations needs current information about inventory, machines, and shipments. One working with transportation infrastructure may need location data and changing conditions. A system managing connected devices needs to know which device produced a reading, when it happened, and whether the underlying data has changed since then.
That's a very different problem from making a chatbot give a better answer. As AI becomes persistent, it has to maintain a useful picture of the systems around it. That picture can include software state, machine data, sensor readings, location information, transactions, ownership records, and events happening across physical infrastructure. Some of that information will arrive continuously, and different systems may need to rely on it at the same time.
This is where XYO fits naturally. We've been building data infrastructure since 2018, with a focus on establishing where information comes from and making important claims about data independently verifiable. XYO Layer One extends that work into a blockchain designed for data-heavy systems, including AI, robotics, DePIN, real-world assets, and other applications that depend on large amounts of changing information.
For an AI system, that can mean having access to data accompanied by cryptographic evidence about its origin and integrity. The model can still do the reasoning, but it has a stronger foundation for deciding what information to trust and what to do next.
The XYO AI SDK brings that infrastructure directly into AI development. It gives developers a way to build XYO Layer One functionality into AI applications, including systems that need to work with verified data across agents, machines, and other sources. That becomes increasingly useful as AI begins interacting with environments that change constantly rather than working only with information handed to it inside a prompt.
This is also why the next phase of AI will be shaped by much more than model benchmarks. Better reasoning will continue to matter, but persistent AI systems also need better access to data and a clearer view of changing state. They need infrastructure that can keep up with what is happening around them while they work.
Dots are an early example of that change. The AI is no longer confined to a single conversation. It stays active, works across systems, and continues operating while the user is somewhere else.
Once AI works that way, the infrastructure around it stops being background plumbing. It becomes part of what makes the system useful.
XYO has spent years building that layer for data. As AI becomes more persistent, more connected, and more deeply embedded in the world around it, that work becomes increasingly relevant.
