How the operating system is structured.

Stratum presents transmission, substations, generation, and queue context spatially. Toggle between queue concentration, interconnection queue, renewables, and environment layers to see a different slice of the same public infrastructure context. The legend in the bottom-right corner updates to match the selected layer and collapses out of the way when not needed; hover or focus the information markers to reveal the layer definitions. Queue concentration is a proxy based on active queue-project count near a substation; it is not a measured grid condition. Queue status reads active, operational, suspended, or withdrawn; renewables size each point by capacity in MW. Environment shows the constraints that most often bring a project into environmental review: protected areas managed for biodiversity (USGS PAD-US, GAP status 1 and 2), endangered-species critical habitat designated by the U.S. Fish and Wildlife Service, and census block groups at or above the 80th national percentile of EPA’s EJScreen 2024 demographic index, from an archived copy after EPA withdrew the tool in 2025. It covers the United States only, and transmission dims while it is on so the areas stay readable. Environment is screening context for siting, not a permitting determination. Transmission stays a neutral color across every mode: line width tracks voltage class, opacity tracks whether in-service status is actually known, not assumed. Click any point for detail. Public infrastructure data does not establish the condition or status of a private project, and queue position is context, not destiny.

Cascade organizes milestones by workstream and records explicit dependency edges, lag, planned and actual dates, deadlines, and consequences. The operational model is built from named project records: milestones, dependencies, deadlines, and evidence. A dependency is not treated as true merely because it is plausible; source, confirmation state, project scope, and scenario status determine whether information may support an operational conclusion. Proposed or simulated changes must remain distinct from committed project state.

Inside Cascade, the Anticipate Agent evaluates how a change to one project milestone could affect connected work, deadlines, and consequences. Teams can test a proposed schedule change before committing it, see the affected chain, and receive a concise recommendation identifying the earliest supported intervention, when it matters, and what waiting could put at risk. Simulated changes remain separate from committed project state, and the Anticipate Agent does not change the project or contact a team on its own. Anticipate identifies where to act. Authorized people decide. The public demonstration uses an illustrative project and static PJM context, not customer data or live project evidence.

Anubis is an open-source governance gate for consequential AI agent actions, developed by Cadentis’s founder. It reviews a proposed action and its evidence before the action may proceed, and returns one verdict. Anubis supports human judgment; it does not make business decisions or act on a project without authorization.