ARE THE DAYS OF HUMAN GEM DESIGN OVER?

In 1902, at Carl Zeiss in Germany, tracing one single ray of light through one spherical glass lens surface by hand took printed 7-decimal logarithm tables and about ten minutes of human arithmetic. Because math was so slow, even a legendary lens designer like Paul Rudolph (who designed the Zeiss Tessar) could only trace thirty or forty rays over six months. Everything else was bridged by "master intuition" and trial and error in the glass shop.

When mainframe computers arrived around 1960 and physicists introduced Damped Least Squares optimization, the old European lens masters scoffed. A silicon adding machine, they said, would never have a human master's "feel for glass." Within ten years, manual lens design was dead. Today, not a single camera lens on earth is designed by a human guessing radii with a logarithm table.

A faceted gemstone is literally a small, single-element polyhedral lens. It runs on the exact same Snell's Law, the same Fresnel polarization equations, and the same wavelength dispersion as a camera lens.

Yet in 2026, most gemstone diagrams are still made the 1938 way. A human sits down at a 1990s Windows CAD program, nudges three pavilion angles with a mouse until a 2D brightness meter ticks up at zero degrees of tilt, names the result Mystic Dragon Starburst VII, and guards the three-page PDF like a state secret.

Which raises a fair question: are the days of a human designing a gemstone cut by hand over?

Every Other Craft Already Had This Argument

Whenever a traditional craft insists its geometry is too artistic for math, history runs the same experiment:

• America's Cup Yachts (1983): From 1851 to 1983, the New York Yacht Club won the America's Cup for 132 straight years. Yacht designers were an aristocratic priesthood who carved cedar half-models by eye and rubbed their thumbs along the hull to "feel a fast boat." Then Dutch aerospace physicist Joop Slooff ran a fighter-jet vortex-lattice computer code for the Australian challenger, flipped the keel upside down, and bolted wings to the tip. The traditionalists called Australia II an ugly gimmick and tried to ban it before Race 1. It smoked the American boat and ended the 132-year streak on the spot.
• NASA's "Drunken Paperclip" Antenna (2006): For a century, radio engineers built antennas out of straight rods and neat helices, not because Maxwell's equations prefer straight lines, but because straight lines are all a human can solve on paper. When NASA Ames wired an evolutionary optimizer to an electromagnetic simulator for the ST5 spacecraft, it bred a crooked, four-branched wire that looked like a stepped-on coat hanger. It hit 93% efficiency, beat the human contractor team's best antenna across the board, and flew in orbit.
• AlphaZero (2017) and AlphaFold2 (2020): DeepMind gave AlphaZero nothing but the bare rules of chess, with zero human opening books. It needed four hours to surpass 1,500 years of grandmaster theory. Three years later, AlphaFold2 solved 3D atomic protein structures to the width of a single carbon atom (1.6 Angstroms) in minutes, a job that used to take a human PhD student five years at a synchrotron.

The 2,496-year Arithmetic Problem

Traditional gem cutters sometimes hear that our catalog has around 100,000 optimized designs and assume it must be computer-generated spam.

Do the multiplication. When you sit down at a faceting machine with a piece of rough, how many basic variables define the cut you need?

• 20 outline shapes (round, oval, cushion, emerald, octagon, hexagon, trilliant, pear, marquise, shield, kite, heart...)
• 8 refractive indices (from quartz at 1.54 and sapphire at 1.76 up to diamond at 2.42 and moissanite at 2.67; an angle that blazes in diamond leaks like a window in quartz)
• 39 crown architectures (flat tables, step cuts, crisscuts, honeycombs, Archimedean and Catalan duals, Penrose tilings, chessboards, Origami folds...)
• 5 length-to-width ratios (a stubby 1.15 oval and a slender 1.65 oval need completely different pavilion keels so the ends do not bow-tie)
• 4 facet-count tiers (from quick 32-facet weekend cuts to 120-facet showpieces)

Multiply those five numbers together: 20 shapes × 8 refractive indices × 39 crown styles × 5 aspect ratios × 4 complexity tiers = 124,800 distinct cutting problems.

100,000 designs is barely one single diagram per niche, before you even count index gears (64, 72, 80, 96, 120) or deep versus shallow rough.

A meticulous human designer working by hand in legacy CAD takes about a week to build, check meetpoints, and tune one diagram for one refractive index and one fixed length-to-width ratio. At one diagram per week, fifty weeks a year, filling out that 124,800-cell grid once by hand would take 2,496 years. If Euclid had installed GemCAD in Alexandria in 300 BC and worked without a single weekend off until this morning, he would still be twelve thousand diagrams short.

Why A Human At A Slider Is Flying Blind

Even inside a single cell of that grid, a human tweaking sliders by hand is fighting with one hand tied behind their back:

1. Everything Moves at Once: Nudge a pavilion angle by 0.3 degrees on a 73-facet stone and you shift the girdle meetpoint, alter the size of the break facets, and change the exit angles of light through thirty-nine crown facets across every viewing angle simultaneously. No human brain can track nineteen 3D refractive ray trees across red, green, and violet light in working memory at once.
2. The 0-Degree ASET Myth: Traditionalists love judging contrast by counting blue ink in a static 0-degree ASET circle. Across 4,799 stones in our 3D Sobel contrast study, static 0-degree ASET Blue correlated with real face-up contrast across tilt at negative 0.174.
3. The Mechanical Cliff: A computer screen has zero mechanical play; a real faceting machine has mast flex, dop wax creep, and human fingers. When we simulated 8.33 million cuts with random half-degree angle jitter across 83,328 designs, we found that many hand-tuned "competition" diagrams sit perched on a razor-thin optical spike: miss the pavilion angle by 0.3 degrees at the bench and the center blows out. An optimizer tests random angle perturbations automatically and parks the design on a wide, forgiving plateau.

So What Is Left For Humans?

None of this makes the human gem cutter obsolete. A GPU cannot hold an irregular pebble of Montana sapphire up to a lamp, spot a feather inclusion, orient a royal-blue color zone into the culet, or feel when a stubborn corundum parting plane is dragging on a tin lap. Nor can a GPU decide which look you want your stone to wear.

But what about sitting at a mouse for a week trying to guess three tangent angles by hand?

When Zeiss retired its logarithm tables in 1963, photography did not die; photographers just got sharper lenses. Whether manual gem design is still a living craft, or the modern equivalent of looking up 7-decimal logarithms by candlelight, is up to you. Browse the 191 stones in our Curated Collection or spin a stone in ProFacet Studio and judge for yourself.