Phasis builds World API to orient AI agents before they act

The company began by using graphs to measure culture, seeking to understand relations not just between people but between a person and their values, beliefs and subcultures. The goal was to map culture in ways agents could understand what culture is composed of and make decisions based on its fast-changing nature, Farley told John Furrier for theCUBE + NYSE Wired: AI Luminaries.

After that work, Farley said, Phasis found that the same process it applied to cleaning, verification and ontology creation could map to other domains. That led to what the company calls the World API, a representation of reality and the real world that covers many different domains for agents to orient against. Farley said reasoning has scaled rapidly and advances weekly, while expertise and understanding of the world have not moved as fast.

The Phasis World API is designed for production agents. A World is the maintained, evidence-backed state of a domain, and Phasis builds and deploys Worlds so agents can be oriented before work begins and remain current as circumstances change. Farley said the API is designed to be built into systems as they are now, along with the primitives needed to build a reality layer and expertise.

He said companies whose agents are oriented against reality will compound their advantages, while agents that piece together context each time will lag. The durable asset is the World model, Farley said, and companies able to use it can move ahead quickly.

The primary endpoint is called /orient. A user can send a task to /orient, and Phasis returns the relevant World frames an agent needs to start, including current state, material changes, supporting evidence and uncertainties. Farley said the endpoint adds an orientation step before an agent searches, plans or strategizes. It provides a quick check against reality across the shared world or a private world and is meant to give any agent orientation before it starts, so it knows where to look and what it does not know.

Phasis uses the tagline “Models Reason. Worlds Orient.” and has positioned itself as a resource for agentic deployment. Farley said the AI future is expanding quickly and more advances are coming. “We’re seeing smaller models that are incredibly capable,” he said, adding that over the next year he expects to witness scaling of agent swarms and auto research. He said the ways intelligence will be applied will be remarkable and that new things never possible before will surface.

The interview was part of theCUBE + NYSE Wired: AI Luminaries series. TheCUBE is a paid media partner for the series, according to the disclosure accompanying the interview, and neither Neo4j, the sponsor of theCUBE's event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.