An object-oriented building model
Translates any 2D drawing file, CAD, DWG or SVG, into a physics-constrained, IFC-native object model.
Layer 1The FoundationThe construction world model.
A physics-aware neural substrate for the built world. Upload any 2D drawing set and Spacial rebuilds it as a working building model grounded in real physics. That model becomes a shared substrate for structural, energy and MEP, so a single wall gets checked for the loads it carries, the heat it lets through and the wiring inside it, all at the same time.
- 97.2%of results within tolerance of reference solutions
- 100–1,000×faster physics evaluation per element
- $15.8Bannual U.S. cost of coordination failure (NIST)
No drawing holds the whole building.
A typical set of Construction Documents runs to thousands of lines and symbols across at least five disciplines: architectural, structural, mechanical, electrical and plumbing. Each one is drawn in a different tool, and there’s no common substrate that holds the full building. That’s why most coordination errors show up between the drawings, where reviewers are least likely to catch them.
No shared model
A single building usually comes with five discipline drawing sets, each in its own file. Every sheet can be accurate on its own and pass review by itself, but there’s no substrate underneath that models the building. Together they’re just a record of what each discipline drew.
Between the drawings
A duct and a beam occupy the same volume; a plumbing stack passes through a shear wall. Clashes between disciplines appear where trades meet, and existing clash-detection tools already catch many of them.
Inside a single element
When insulation, conductors and a vent stack all need the same wall cavity, the conflict stays invisible. No drawing shows the inside of an element, and no clash tool looks there. Spacial’s patent-pending modeling was created to find these hidden overlaps.
The file is not the building.
There are two ways to represent a building: as files of lines drawn one discipline at a time, or as one model of the building itself.
Every element carries what it is, what it holds and what it touches.
Lines and layers in separate files. The conflict only shows where two of them happen to cross.
Building-native elements
Walls, beams, systems and openings are reconstructed as elements with attributes, constraints and behavior.
One shared substrate
Structural, envelope, HVAC, electrical and plumbing layers reason over the same underlying model.
Continuous coordination
Every change is evaluated across the building, surfacing conflicts while decisions can still be resolved.
One substrate, five layers, every discipline.
Spacial rebuilds every 2D drawing set as an IFC-native object model, the substrate every discipline works from. Each element carries its own identity, the loads acting on it and the code requirements that apply to it.
An attributed bond graph
Assigns every element what it can and cannot do, and builds the attributed bond graph that drives all downstream reasoning.
Layer 2The Substrate White paper
Intra-element behavioral modeling
A single wall is a structural member, thermal envelope, electrical host and plumbing chase at once.
Layer 3Micro-scale physics White paper
Whole-building graph
Reads the whole building as a graph, predicts needed connections across disciplines and catches conflicts where trades intersect.
Layer 4Macro-scale topology
Generation and review loop
Coordinates generation, in-loop validation and cross-discipline resolution, then hands a reviewed package to the licensed engineer of record.
Layer 5The OrchestratorWhite paperThe Substrate and Micro-scale physics are the two layers the white paper focuses on: the first decides what each element can do; the second evaluates what follows when several disciplines act on it at once.
Model elements by their attributes, not their geometry.
Conventional building data models sort everything by type. An element is a wall or a window or a beam, and its behavior comes from that label. Rules written this way can only say yes or no, so they can’t capture a conduit passing through a shear wall, which is allowed but reduces the wall’s capacity and only works up to a point.
Spacial describes each element by its capabilities rather than its label. Those capabilities are encoded in an attributed bond graph, and every layer built on top of it inherits them.
The conflict no drawing can show.
A 2×6 exterior wall, three disciplines, one finite cavity.
Wall cavity volume per stud bay
The same 2×6 bay, drawn as it is built: R-21 batt insulation, branch-circuit conductors and boxes, and a vent run, all sharing one finite volume that no single discipline’s drawing represents.
105%of the cavity
Total demand exceeds capacity. Spacial flags a 5% over-allocation across thermal, electrical and plumbing, a conflict no drawing review would reveal.
Measured against reference physics, not marketing copy.
We test every headline number against classical reference solvers, using real plan sets held out from training. The scoring criteria are locked in before each run.
Every prediction comes with a tested confidence range
Conventional tools give you a single number with no sense of how far off it might be. Spacial gives an estimate and a range, both tested on real plans the model never saw during training.
R-17.5Effective assembly R-value
- 97.2%
Tolerance-band agreement
Predicted engineering quantities agree with classical reference solutions within registered tolerances.
- 5×10⁻²s
Per-element evaluation latency
Fast enough to close a design loop inside a working session, not overnight.
- 0.00%
Hard-constraint violation rate
Structurally enforced laws are guaranteed by graph topology. They cannot be violated by construction.
Within registered tolerance
- Baseline
- A predefined reference and acceptance criterion.
- Measured result
- Reported against a criterion registered before the run.
- Acceptance criterion
- Registered in advance, never fitted after the fact.
Every figure names its baseline, its corpus and its criteria.
We measure accuracy against an acceptance criterion registered in advance, using a held-out set of real plan sets split by project, jurisdiction and firm. None of that data is ever used to train the model. If a figure is a target and not yet a measured result, the white paper labels it that way.
Every package is still reviewed and stamped by a licensed engineer. Spacial speeds up generation and coordination, while professional accountability stays with the engineer of record.
One topology, two readings.
One construction graph handles both geometric understanding and physics-aware reasoning, and the underlying building model stays the same throughout.
Loads, flow participation and conservation residuals computed from the same connected model.
- Thermal and energy modeling. Reads nodes as capacities, boundaries and flows.
- Unintended leak prediction. Surfaces hidden links across structural, thermal and MEP zones.
- Conservation-law monitoring. Detects when energy, air or mass balance breaks.
Shape, position, adjacency and the relationships a conventional graph can observe.
- Object and role identification. Assigns useful roles to walls, beams, systems and openings.
- Connectivity and compliance. Infers missing links and checks paths against building rules.
- Geometric clash detection. Flags conflicts before they become field conditions.
Attributed graph
Message passing
Vector per node
Readout → output
A four-tier reference framework, not a single benchmark.
Spacial checks each prediction against reference evidence that gets harder to match at every step: analytic answers, then independent solvers, then measured buildings, and finally real field records.
Analytical solutions
Closed-form references confirm the model reproduces the governing equations correctly and reveal any numerical instability.
Independent solvers
We compare results against established engineering software that had no role in producing the model’s training data.
Measured field data
We compare predictions with data from instrumented buildings, where what the building does counts for more than agreement with a simulation.
Field defect records
Review comments, RFIs and change orders reveal which errors would have been caught before construction.
Three rules hold for every result.
The tiers set what a prediction is checked against. These rules set how the check is run, so a figure can’t look better than the model really is.
No leakage between train and test
Evaluation splits at the project level, grouping near-duplicate plans and keeping the geographic, jurisdictional and author separation a real deployment needs.
Criteria are registered before the run
Tolerance bands are locked before testing, and any result that misses its criterion is reported as a failure.
The tail is reported, not just the mean
Averages don’t describe every prediction. Maximum and high-percentile error are reported beside central estimates, so safety-relevant outliers are visible and can be escalated for engineering review.
Request and reach out.
Questions about the model, the evidence or the white paper? Talk to the team.
Go inside
the model.
Explore the architecture behind building intelligence: element attributes, physical behavior and connected systems.
For teams exploring the technical foundations of the Construction World Model.
Request the white paper
Go deeper into the model.
Explore how geometry, physics and building systems connect in Spacial’s world model.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.