Digital stereotomy: when stone cutting meets computation

01

Stereotomy: geometry in service of stone

Stereotomy means the cutting of solids. In masonry it names the art of shaping individual blocks so that a vault, a stair, an arch over a skewed opening, or a dome resolves into pieces that can each be cut on the ground and then assembled true. The block a mason cuts is a three-dimensional answer to a three-dimensional question: what shape must this stone be so that its neighbors, and gravity, are satisfied? Get one face wrong and the error propagates through every stone that touches it.

The historical toolkit was drawing. Builders developed projection methods for unfolding complex solids onto flat surfaces, producing templates that transferred exact profiles to the block. Full-scale layouts were drawn on dedicated floors, and treatises on the geometry of stone cutting circulated in early-modern Europe, a documented tradition that later fed into formal descriptive geometry. The specific methods varied by region and era; the underlying move, reducing a solid to cuttable, checkable surfaces, has not changed at all.

02

The tracing floor was a computer

A tracing floor executed geometry step by step and stored the result as lines: run the construction, extract the template, apply it to stone. It is fair to call it computation done at architectural scale with cords and dividers, and it explains why masons could build forms nobody could then describe algebraically. The template is the crucial artifact. It freezes a geometric decision into a physical object that can be copied, checked, argued about, and reused, which is most of what a data file does.

Documentation of traditional masonry practice matters for more than nostalgia. Stewards of historic buildings rely on published preservation guidance to understand how such fabric was made and how to repair it in kind, and the U.S. National Park Service's Preservation Briefs are a standing example of that documented, practice-level record ². When a modern workshop reverse-engineers a damaged vault stone, it is rejoining a very long conversation, not starting one.

03

The digital chain today

A present-day stereotomy job typically runs: capture existing fabric or design intent by scanning or photogrammetry; build a parametric solid model in which each block is an object with faces, joints, and allowances; prove out the assembly digitally; rough each block by saw, CNC mill, wire, or robotic cell; then finish by hand. Additive processes join this chain when mineral printing produces formwork, trial parts, or study models, and the vocabulary standard ISO/ASTM 52900 gives the shared language for those process families ⁴.

Each stage has a distinct job. Scanning establishes what is true; modeling establishes what should be true; machining removes most of the difference quickly; the hand removes the rest accurately and leaves the surface the eye will actually read. Skipping the prove-out stage is the classic failure: blocks that are individually correct and collectively impossible. The machine did not make the geometry problem go away. It made the geometry problem explicit, which is better.

• Capture: scan or photograph the reference fabric
• Model: parametric blocks with joints and allowances
• Prove out: assemble digitally before cutting
• Rough: saw, CNC, wire, or robotic cell to near-net shape
• Finish: hand tooling to final surface and fit
• Record: what was cut, checked, and changed

04

Tolerance thinking joins mason and machinist

A machinist speaks of datums, allowances, and fits; a banker mason speaks of face marks, templates, and beds. These are the same ideas wearing different aprons. Both trades choose reference surfaces first, work everything else from those references, leave deliberate excess where later operations need material, and measure at agreed points rather than everywhere. Both also know the hardest question is not how to cut but when to stop cutting, because material removed from stone does not come back.

Tolerance is a budget, and the budget must match the joint. A bedding surface that will sit on soft lime mortar forgives more than a dry joint that must close over its whole face. Deciding where precision is spent, and where it would be wasted effort, is craft judgment supported by measurement, and it transfers directly between a machine shop and a masonry yard. The two cultures meet whenever a scanned historic profile becomes a toolpath and then, at the end, a hand-dressed face.

05

In service of heritage, under doctrine

Machine-assisted cutting earns its keep in restoration by producing replacement units faster and by matching recorded profiles closely, but heritage work does not relax its rules for speed. The Venice Charter requires that replacements integrate harmoniously with the whole while remaining distinguishable from the original, so that restoration does not falsify the document a building constitutes ³. A robot can rough a crocket; the decision about what that crocket should be remains an evidence question.

Conservation institutions frame this as science plus craft. The Getty Conservation Institute's work on the care of built heritage treats materials analysis, documentation, and practitioner skill as one system rather than rivals ¹, and preservation guidance for historic masonry stresses repair in kind and compatibility with the surviving fabric ². Digital stereotomy fits inside that frame comfortably: the computation serves the stone, and the stone serves the record. The order matters.

06

Open questions in the hybrid workshop

How far should automation reach? That is genuinely open. Machine roughing is established; fully automated finishing of heritage-grade surfaces is not, partly because the finished texture of hand tooling is itself part of what some buildings document. Training pipelines are open too: a workshop needs people fluent in both point clouds and pitching tools, and the supply of such people is a bottleneck that no standard resolves. Costs, small-shop access to machines, and sustainable stone sourcing are live economic questions.

What is not open is the direction of accountability. Whether a surface was cut by hand, machine, or both, someone must be able to say what the reference was, what tolerance applied, what was measured, and what deviated, in records that outlive the job ³. Computation makes those records easier to produce and easier to lose. Workshops that treat the digital file with the same seriousness as the template drawer will be the ones heritage clients trust.

• Established: machine roughing to near-net shape
• Open: automated finishing of heritage-grade surfaces
• Open: training people fluent in both crafts
• Open: machine access economics for small workshops
• Fixed: documentation and distinguishability duties