Fabric is Seth’s idea for an agent-native database where the database itself becomes the platform. Its near-term loop is simple: publish immutable modules or artifacts, call governed reducers that change state, then query or subscribe to the results. The long-term goal is a typed, time-aware semantic system that connects code, documents, processes, CAD, spatial data, agents, organizations, simulations, and physical work—while preserving the history, evidence, authority, and provenance behind every change.
This could support large engineering efforts where a change in one discipline affects many others. A building, factory, bridge, spacecraft, or energy project could connect requirements, geometry, materials, procurement, schedules, cost, sensors, inspections, simulations, permits, and field observations in one causal model. An agent could evaluate a design change, determine which drawings and components are affected, run structural or thermal analyses, check constraints and approvals, update downstream schedules, and present a reviewable proposal with the evidence behind it. CAD models, GIS data, IoT telemetry, maintenance records, and construction progress would become linked views of the same evolving project rather than isolated files and systems.
The intended scale is universal: local machine coordinates can connect to a facility, construction site, city, region, planet, or eventually orbital and planetary systems, with time-aware coordinate frames and explicit uncertainty. Fabric would distinguish observed reality from planned, simulated, predicted, or disputed states, allowing engineers to explore alternate designs and merge approved results into operational reality without confusing a simulation with a fact. In that model, software agents, human teams, robots, sensors, documents, physical assets, and processes all participate as governed entities in one semantic environment—one that can explain what changed, why it changed, what evidence supported it, and what real-world consequences follow.