Æther FacetDesign Bench Start Demo

The Æther Advantage

Your rough. Your vision.
More possibilities.

What if you could design a gem specifically for the stone in your hand—not simply choose a familiar cut and make the rough fit it?

Æther Facet is being built to combine individual rough geometry, simulated physical optics and your creative judgment. The aim is to find exceptional designs that might otherwise remain unexplored.

A new design process

A better question than “Which standard cut fits?”

Start with the rough and the aesthetic result you want. Let computation widen the possibilities while you keep control of the tradeoffs.

01 / GeometryResearch demonstrated · workflow in development

A cut for this exact stone.

Design around the shape, proportions and constraints of an individual piece of rough rather than fitting a stock diagram to it. Rough-aware optimization has been explored in GemML research; automatic photogrammetry is not yet connected to the public demo.

02 / Creative controlSeven metrics now · more planned

Optimize for your definition of beauty.

Brilliance, fire, scintillation, contrast, retained volume, entropy and fractal dimension are the seven current objective categories. Approximately 24 additional metrics are planned. Set your priorities, compare alternatives and explore new optical patterns—not just the familiar ones.

03 / Crystal physicsFuture optical models

Make anisotropy part of the design.

Uniaxial and biaxial gems have direction-dependent refractive behavior. Future models aim to account for those differences in choosing crystal orientation and facets, rather than approximating every material with a single isotropic refractive index.

04 / Light and colorFuture optical models

Go beyond refractive index alone.

Spectral absorption, color, pleochroism and difficult orientations—including a dark or closed c-axis—will become part of the design problem. Additional optical behavior may be incorporated as the physical model develops.

05 / Real-world roughFuture defect-aware models

Work around what makes a stone imperfect.

Inclusions, fractures, and blemishes affect what can be cut. Future models will aim to account for these features and explore whether a different orientation or geometry can preserve useful material without compromising the result.

06 / Human + computationDesign philosophy

More possibilities. Your final judgment.

The model searches. You decide which candidates are worth pursuing. Each choice is intended to guide the next round as preference learning is integrated. Your aesthetic judgment is the point—not something to automate away.

Read our AI Philosophy →

Preliminary GemML research · earlier models

The search has already uncovered promising possibilities.

Earlier rough-matched simulations found designs with stronger optical scores, better finished yield, or deliberate tradeoffs between the two. Here are selected examples—not typical or guaranteed outcomes.

One October 14 benchmark · same rough, two designs

More simulated brilliance and more finished stone.

The optimized design improved both outcomes relative to the fitted library design for that particular rough.

GemML v4 rough-joint, run 20261014_155716, validation round 000285.

+20.09%Simulated brilliance0.7229 → 0.8681
+10.95%Finished volume45.42% → 50.40% of rough

Balanced-profile averages · November 11

Fire & scintillation

+15.4%Simulated fire
+23.0%Simulated scintillation

Reported averages from a separate balanced-profile set; not additional gains on the October 14 cut.

Optical-first example · November 7

Make a deliberate tradeoff.

+12.84%Simulated brilliance
−0.97%Finished volume

One simulated moissanite case traded a small amount of weight for a substantially higher brilliance score.

Beyond a single score

What could this mean for the value of your rough?

A better-looking stone may be more desirable even when its finished weight is lower. In other cases, a design may improve optical appearance while retaining more material. Which matters more depends on the stone, buyer and intended jewelry.

In simplified terms, finished gross selling value depends on finished carat weight × realized price per carat. The research challenge is to explore both without assuming that one must always be maximized.

The opportunity: a more compelling cut—and, when useful, more finished carats—could make an individual piece of rough substantially more valuable. The actual economic outcome must be established in the market, not inferred from simulation percentages.

Development roadmap

Built in stages, with the research still moving forward.

The public interface demonstrates the direction of the product. The new machine-learning model and its scientific evaluation are still being developed and connected to WordPress.

Now / demonstration

See how the Studio will work

Interactive rough previews, material-property inputs, seven optimization priorities, sample candidates, saved gem records and session organization. Earlier research provides preliminary rough-aware isotropic optimization results.

Planned / under development

Expand the physical and creative search

Automated photogrammetry, new candidate generation and preference learning, calibrated evaluation, cutting diagrams, defect-aware geometry, anisotropic crystal optics, absorption, pleochroism and approximately 24 additional metrics.

Creativity with a larger design space

Better possibilities begin with your rough and your priorities.

Explore the current demo, or learn how Æther Facet is being designed to support rather than replace the creative faceter.