A working architecture,
on a charted path.

PCM is not a research thesis. A working architecture has been demonstrated, and pilots are running.

In April 2026, pCurve demonstrated the Persistent Cognitive Machine to the U.S. Navy at NIWC Pacific. The demonstration exercised the architecture on a static manifold: knowledge organized geometrically, queried efficiently, and committed only when the system could structurally defend its answer. Pilots with a small set of partners are continuing under non-disclosure.

The patent estate underlying the architecture has reached 253 filings in the geometric AI category. We believe an architecture worth deploying is also an architecture worth protecting.

What has been demonstrated, where the work is going next, and the destination it is built toward. Intermediate milestones are disclosed to engaged partners under non-disclosure.

We believe a credible roadmap names what has been done, names the destination, and is honest about which middle stages it is appropriate to discuss publicly. The destination here is a federation of cognitive agents that share what they have learned without surrendering what they know. The path to that destination is sequenced and underway.

Every stage in the roadmap is the same architecture with more of its capabilities enabled. We are not rewriting the foundation as we go.

Every release expands the same underlying architecture. The foundation is not being rewritten as the work proceeds. Knowledge is curvature on a manifold at every stage; what changes is how much of the manifold is alive in real time, how much can be exchanged across agents, and how much the system manages on its own behalf. The version demonstrated today and the federation at the end of the path are the same machine at different levels of operation.

We believe this is the operationally important fact about the path. Each release is an expansion of an architecture already understood, not a replacement of one that did not quite work.