Spacial is pre-construction engineering software for the built world. The platform generates residential structural, architectural, and MEP plan sets from a single model of the building, coordinated across disciplines from the start rather than reconciled at the end. Residential engineering is a coordination and iteration problem. An architect moves a window, structural re-checks, MEP re-routes, sheets re-issue, and the cycle repeats dozens of times per project. Existing tools make checking one iteration cheaper. Spacial makes the iteration itself cheap, which is where residential projects actually lose time and fee. The technology is built on a different premise than general-purpose AI. Spacial understands buildings the way language models understand text: as structured relationships, not lists of elements. The platform represents each building as a graph, with members and joints as nodes and load paths as edges, and generates designs using graph neural networks bounded by a physics-aware model. Equilibrium, compatibility, material limits, and code constraints bound every output. The result is a system that computes designs rather than predicting plausible-looking ones. Spacial is built for architects, builders, and engineers working in residential construction, and for the plan-review realities they work under: jurisdiction-specific code editions, local amendments, and the requirement that a licensed professional holds responsible charge. Generated designs are structured for review, verification, and stamping by licensed engineers.
