Industries / Consumer goods / Jewelry
Cutting and setting stones
Cutting grinds a rough stone into a gem with flat, polished faces called facets, and setting fixes the finished stone in metal with claws called prongs, a metal rim called a bezel, or rows of tiny beads of metal (pavé).
7
Milestones achieved, from c. 1476 to 2024.
1
To watch for: announced, not yet confirmed outside.
3
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
c. 1476
A wheel polishes diamonds
The scaif is a spinning disc covered with diamond powder and olive oil, and it let cutters polish regular facets onto diamond, by the Diamond Museum in Bruges's description. Tradition credits Lodewijk van Berquen, a goldsmith of Bruges, with inventing it in 1476, but the museum calls this "an old myth", and the historian Jack Ogden writes that there is no evidence he ever existed; the story traces to Robert de Berquen (1615–1672), who claimed to descend from him. A polished diamond is recorded earlier, in a Burgundian court inventory of 1467, by the same account.
LessonA trade's founding story can rest on one family's legend, so names and dates need records behind them.
Credit
unknown; tradition names Lodewijk van Berquen
1886
A ring lifts the stone
Charles Lewis Tiffany introduced a ring with six prongs that hold a single diamond up above the band, where more light reaches it. Tiffany & Co. calls it "the first ring design in history to lift the diamond off the band"; that "first" is the company's own claim.
Credit
Charles Lewis Tiffany, Tiffany & Co
1908
Splitting the Cullinan
The Cullinan, found in 1905, weighed 3,106 carats in the rough. Joseph Asscher had new tools made for it in Amsterdam and spent weeks cutting a groove about 1 cm deep to start the split. On his first blow, on 10 February 1908, the knife broke and the stone held; a second attempt with heavier blades split it in two, four days later by Royal Asscher's own account. The pieces became nine large stones for the British Crown Jewels and 96 smaller ones, which Asscher kept as its fee, and polishing the largest took about six months.
LessonOn a very large stone, weeks of preparation go into a single cut, and the tools may have to be made for the job.
Credit
Joseph Asscher and his family's diamond firm in Amsterdam
1919
Brilliance and fire calculated
In his book Diamond Design, Marcel Tolkowsky used the optics of reflection and dispersion (light splitting into colors) to calculate the shape of a round brilliant that balances brilliance, the white light sent back to the eye, against fire, the flashes of color. His figures were a pavilion (lower part) angle of 40° 45', a crown (upper part) angle of about 34° 30' and a table, the large top facet, 53% of the stone's width.
LessonThe book's modern editor notes that the best crown angle and table size change with the pavilion angle, so his numbers are one good answer, not the only one.
Credit
Marcel Tolkowsky
1981
Lasers saw diamonds
A laser saw focuses a neodymium-doped YAG laser to a spot 25 microns across and can cut a diamond in any direction, not only along the few directions in which a bronze blade will saw. David Baker, a cutter in Columbus, Ohio, wrote in Gems & Gemology in 1981 that a 10-carat crystal took about eight hours, while a stone with a diamond crystal inside it could take days or weeks on a blade, and that about 50 laser saws were working worldwide by unofficial estimates. The laser lost about 5–7% of the weight, slightly more than a blade. Akiva Caspi dates the first laser sawing to the late 1970s.
Credit
makers of laser sawing systems; Baker names no single inventor
1997
Computers plan the cut
By 1997 cutters could use DiaExpert, a new system made by Sarin of Ramat Gan (now Sarine Technologies), which measures a rough stone from many angles, builds a 3D model and shows the marker, the worker who decides where to cut, what each choice would yield. Automatic polishing machines sense when a facet is finished and move to the next, so one trained operator could polish 16 diamonds at once. Akiva Caspi, an electronics engineer and a member of the Israel Diamond Institute from 1987 to 1997, wrote in Gems & Gemology that cutting had changed more since 1980 than in the 100 years before, and that a modern factory cost about 10 times as much to set up as a traditional one while the cost per stone fell.
LessonCaspi wrote that workers who could not adjust to the machines had to leave the industry, so automation changed who cuts as well as how.
Credit
Sarin (DiaExpert) and the makers of automatic polishing machines, as described by Akiva Caspi, who named Israel and Belgium as the most advanced centers
2024
Surat's cutters lose work
India's modern cutting industry began in the late 1950s, with Surat in Gujarat as its first center, and by 1996 India polished about 70% of the world's diamonds by weight and 35% by value with 600,000 to 800,000 workers, by Sevdermish, Miciak and Levinson's count in Gems & Gemology (1998). Scroll reported in April 2026 that the polishing sector employs an estimated 800,000 to 1,000,000 workers, and that the Surat Diamond Workers Union had counted more than 80 suicides among Surat's diamond workers since January 2024. Causes named there include sanctions on Russian rough, US tariffs of 26 to 50% from April 2025 (lifted in February 2026), the war in West Asia, and lab-grown stones, which many small units took up to stay open but which pay about Rs 12 to polish against Rs 15 to 60 for a natural stone.
LessonA center that cuts most of the world's stones depends on rough supply and demand that it does not control.
Credit
Palanpuri trading families built the Surat industry, by the 1998 account
Watch for
Automated planning of rough
Sarine says its Most Valuable Plan software, launched in 2024 and built on its Advisor planning software and its Galaxy scanners, which map the inclusions (flaws) inside a rough stone, gives "predominantly automated planning" for natural rough diamonds and saves time and cost.
Achieved whena cutting firm or researcher outside Sarine publishes the polished weight and value from its plans against experienced markers' plans on the same set of stones.
Claimed by
Sarine Technologies, June 2026
Open
01
Open rough-planning benchmark
A public set of 3D scans of real rough diamonds, each with the cutting plan an open-source planner proposes, the plan an experienced marker or commercial planner chose, and the polished stones that came out. Lab-grown rough makes this affordable, and an open baseline planner lets anyone compare methods on the same stones.
The briefCredit · unclaimed02
Measured strength of settings
A bench test, and a public data set from it, of how well common settings hold their stones: the force needed to push a stone out, and how many stones loosen after a fixed number of standard knocks, for four-prong, six-prong, bezel and pavé settings made by several bench jewelers.
The briefCredit · unclaimed03
Filmed robot setting trials
An open, low-cost robot cell that picks small round stones from a tray and places them table-up into drilled seats in a wax or metal plate, the first step of wax setting and pavé, with every attempt filmed and tallied against a hand setter.
The briefCredit · unclaimed
Open 01 · the brief
Open rough-planning benchmark
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public set of 3D scans of real rough diamonds, each with the cutting plan an open-source planner proposes, the plan an experienced marker or commercial planner chose, and the polished stones that came out. Lab-grown rough makes this affordable, and an open baseline planner lets anyone compare methods on the same stones.
Why it is still open
Commercial planners such as Sarine's Advisor and Most Valuable Plan are closed and publish no yields measured against people. The closest public work found is a 2021 University of Massachusetts class project that trained a model on synthetic rough shapes made in Blender, not on scans. Searches of GitHub, Hugging Face, Zenodo and Crossref found no public scans of real rough with polished outcomes.
Built with what exists
Lab-grown rough, which is now cheap, since Edahn Golan put wholesale lab-grown prices near $100 per carat at the end of 2025; a turntable with a macro lens and open photogrammetry software such as Meshroom, or a desktop structured-light scanner; Python with trimesh and SciPy for the planner; a model such as Claude Fable 5.2 to write and check the optimization code (see the Note "Three working devices"); and a cutting workshop in Surat or a hobby cutter to polish the stones.
First step
Buy ten small lab-grown rough stones, scan each, and check every scan by multiplying its volume by diamond's density (3.52 g/cm³) and comparing with the stone's weight on a carat scale.
Done looks like
At least 200 scanned rough stones released under an open license with their plans and polished weights, and a baseline planner whose predicted polished weight is within 5% of the actual for at least 80% of the stones, compared on the same stones with the human or commercial plan.
Open 02 · the brief
Measured strength of settings
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A bench test, and a public data set from it, of how well common settings hold their stones: the force needed to push a stone out, and how many stones loosen after a fixed number of standard knocks, for four-prong, six-prong, bezel and pavé settings made by several bench jewelers.
Why it is still open
GIA's Jewelry Manufacturing Arts instructors publish benchmarks for prong settings, such as a prong angle of 70 to 80 degrees, prong height of 75 to 85% of table height and prong contact over 33 to 50% of the crown, but cite no tests behind them. The only test found is a 2014 soak test by Esslinger, a jewelry supply company, which put set rings in bleach and hot-tub chemicals, stopped at the first failure and gave no sample count. Crossref and GitHub searches found no published push-out or impact data for jewelry settings.
Built with what exists
Identical cast sterling silver or brass mountings; cubic zirconia or low-cost lab-grown stones; a digital force gauge on a motorized test stand with a 3D-printed fixture; a simple drop or tumble rig; a USB microscope, with a vision model such as Claude Fable 5.2 to measure prong tips from the photos (see the Note "Three working devices").
First step
Order 20 identical four-prong silver mountings, have two jewelers set them, and push five stones out on a rented force gauge.
Done looks like
At least 30 settings of each of four styles tested, with push-out force and the share of stones loosened after 1,000 standard knocks published with photos, plus a test of whether settings that meet GIA's benchmarks hold their stones better than those that do not.
Open 03 · the brief
Filmed robot setting trials
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open, low-cost robot cell that picks small round stones from a tray and places them table-up into drilled seats in a wax or metal plate, the first step of wax setting and pavé, with every attempt filmed and tallied against a hand setter.
Why it is still open
Setting machines exist but none has published an outside tally. Giovanni Daems of Kessel, Belgium, developed an automatic diamond setting machine in 2018 that, by the account of the jeweler Baunat, sets stones 15 to 20 times faster than by hand. The jewelry designer Christian Tse told JCK in August 2023 that his setting robot sets 0.6 mm melee (very small diamonds), with no accuracy figures given. GitHub and Crossref searches found no open setting robot or published success rate.
Built with what exists
A low-cost arm such as the SO-101 recorded and trained with LeRobot (see the Note "No download"); a small vacuum pick-up nozzle and a USB microscope camera; an open robot model such as NVIDIA GR00T 2 (see the Note "Five demonstrations per task") or Runway Praxis-1 (see the Note "Open robot model weights"); cubic zirconia stones of 2 and 3 mm.
First step
Drill a 10-by-10 grid of seats in a wax plate and record 50 teleoperated placements with the vacuum nozzle to use as training demonstrations.
Done looks like
200 filmed attempts at each stone size, with the share of stones seated table-up and level, as judged from microscope photos by two people who did not watch the runs, and the time per stone against a hand setter's.