It all started with a piece by Bryan Cantrill—creator of DTrace and co-founder of Oxide—"What Sun got wrong," dated September 20, 2026. A man who worked at Sun Microsystems almost until its death torched fifteen years of corporate nostalgia in one stroke: "Sun got bored with the mechanics of running a business." Technically, Sun was magnificent—SPARC, Solaris, Java, ZFS, NFS—and yet the company died. Cantrill recalls an incident from 2005: a startup that had grown on OpenSolaris and wanted to buy "a mountain of hardware" from Sun spent a month trying to reach the sales department. Meanwhile Dell was shipping servers via web form in the middle of the night within two weeks.
After a thesis like that, an engineering mind doesn't stop. If a magnificent product can die from lousy operations, that pattern should work in reverse too, and at other scales. And immediately you remember a story that for some reason almost no one remembers: the pizza startup Zume, which raised nearly half a billion dollars on the idea of "pizza as production line"—and quietly died in 2023 because cheese kept sliding off the pizza while it baked in a truck driving down a California highway.
That gives us a ready-made container story: product ≠ system. Sun made the best server in the world and couldn't sell it. Zume made the best pitch in the world and couldn't bake a pizza. KFC in 2009 baked everything ahead of time—and crashed at fulfillment. Three funerals, one disease. Let's figure out exactly how a system dies when it had everything going for it.
In 2015, two people with perfect résumés collided in Mountain View. Julia Collins—Stanford GSB graduate, analyst for Danny Meyer of Shake Shack, participant in the launch of the New York chain Mexicue. Alex Garden—at 23, author of the hit strategy game Homeworld, then director of Xbox Live at Microsoft and head of game studios at Zynga. Their company Zume Pizza promised to take a chunk of the American pizza market worth about $40 billion a year in the most Silicon Valley way possible: robotic production and smart delivery.
And you can't say it was an empty pitch. In 2016, Business Insider personally toured Zume's facility in Mountain View, and it looked like an engineer's dream come true. Dough fermented for two days for lightness and digestibility. Further down the line came sauce dispensers with friendly names Jojo and Pepe, a robot named Marta with spider-leg-like arms smeared sauce across the dough in seconds, and lifted by the robot Vincenzo (an ABB industrial manipulator), the pizza was lowered into an 800-degree Fahrenheit oven—about 427 Celsius. The ABB robots—the same ones that place parts on automotive assembly lines—cost $25–35 thousand apiece, and the integration of the entire line was patented. By 2018, the facility was producing, according to the company itself, about 370 pizzas per hour.
The cherry on top was logistics. The first pizza went out to a customer on April 1, 2016, the ten-thousandth in September, and following them came trucks with 56 ovens on board. If the address was more than twelve minutes away, the pizza was loaded into the oven without a box: a GPS trigger as the truck approached the house activated "finish baking" for the final three and a half minutes, and the driver pulled out a hot pie right at the curb. Boxes were made from sugarcane fiber, with clever ribs on the bottom channeling grease into a special well. Even cutting was perfected: their own self-washing slicer divided each pizza into exactly eight slices of 180 calories each.
Order data promised something even bolder: "we know what pizza you'll order while you're watching Monday Night Football, before you do"—and they loaded ovens before the call came in. Julia Collins framed this in a CNBC interview as a social mission: robots would take on dangerous drudgery, people would develop recipes, taste-test, and learn, employees got equity stakes and paid education. On paper, it all added up. In practice, unfortunately, no.
The first crack wasn't in the pitch but in physics. Robots assembled by engineers from the worlds of gaming and startups started behaving on extended runs like robots, not cooks: they shed metal shavings right into the cooking zone. This isn't an aesthetic detail—it's a sanitary inspection failure, the line between "restaurant" and "production facility unsuitable for certification." A former employee showed Business Insider journalists photos of shavings in the cooking zone and a formulation worth quoting verbatim: "they hired a lot of people who didn't understand what equipment you're supposed to use for food; many of the engineers weren't food equipment engineers."
The problem runs deeper than "wrong metal in the gears." Food production is a regulated environment with a thousand requirements, from biocompatibility of materials to washdown procedures, and here the interface between robotics and HACCP standards turned out to be that very "business mechanics" that Sun couldn't find a salesperson for and Zume couldn't find an engineer for. Notice: the strategy ("automate pizza") was brilliant. The kitchen itself, meaning the company's actual operating system, wasn't even assembled at the level of sanitary standards.
You can see it in other parts of the stack too. A Business Insider taste tester in 2016 ate a $15 "El Camino" pizza and left a prophetic verdict: toppings were excellent—thin crispy pepperoni, juicy mushrooms, cheese stretched like taffy—but the crust sagged under its own weight, and the dough's flavor got lost under the toppings. A perfect metaphor for the whole company: everything glossy on top, load-bearing structure underbaked.
When robotic production stalled, Garden came up with act two: bake on the road. The flagship truck with its own name "Marta" cost over a million dollars—and, according to testimony from at least five former employees, lacked the technology to adequately bake pizza. Inside—broken glass, weak cooling, boxes on the floor. Delivery crews regularly returned to bake pizza at the warehouse and delivered it in personal cars. Garden personally ordered the QA team to keep all eleven trucks operational "until something major happened." Something happened: at one catering event, ovens in two trucks failed simultaneously, and a hundred guests were fed pizza finished in a toaster.
And here's the most important detail of the whole story, in my view. In June 2023, Bloomberg summed up an eight-year saga with one sentence: Zume "struggled to keep melted cheese from sliding off pizzas that were being baked in transit." Sounds funny, but behind this joke lies strict physics. An oven on a rocking truck means variable accelerations, tilts, and vibrations constantly churning the molten mass; liquid cheese, unlike dough, doesn't hold its shape and flows to the edge with every turn. The problem of "finish baking in motion" is a problem of heat transfer plus fluid dynamics plus mounting mechanics all at once. Solving it with a half-billion-dollar budget is, in principle, possible. But Zume spent that half billion not on it but on scaling the concept—and simultaneously on buzz about pizza that "bakes en route," even as employees seriously discussed filing an FTC complaint for misleading advertising.
The ending came in two acts. January 2020—360 people laid off, about half the staff, robots and trucks shut down; SoftBank refused in December 2019 to lead another funding round. May 2023, The Information reported that the company was insolvent and restructuring firm Sherwood Partners was selling its assets. Total—about $500 million raised over eight years, zero operating restaurants. Employees among themselves called the company "Juiceranos"—a hybrid of Juicero and Theranos, two iconic Silicon Valley seppuku stories for a pretty idea.
So you don't get the impression that death by operations is a privilege of startups with robots, let's mix in a chain whose kitchens and processes had existed for decades. On May 5, 2009, Oprah Winfrey announced live on air a free meal of new "Kentucky Grilled Chicken": two pieces, two sides, a biscuit, coupon downloadable at Oprah.com for twenty-four hours. The offer combined two explosive reagents—"free" and "healthy food from Oprah," and in the depths of the recession this mixture proved nuclear-equivalent.
Result: 10.5 million coupons downloaded, food worth $42 million, lines "out the door and around the block." The website crashed, restaurants unprepared for such flow sometimes just refused to accept coupons. Thursday evening KFC surrendered: coupons no longer valid, replaced by a "rain checks" program, a promise to give out your meal later. The bottom line, which Roger Eaton, KFC US president, stated on Oprah's show via Skype: 4 million meals distributed, another 6.5 million—owed to customers.
The difference with Zume is where exactly it broke. Here kitchens worked perfectly: chicken fried, biscuits baked. What broke was the last mile—the mechanics of fulfillment and communicating the promise. PR veteran Alan Marcus gave the formulation: "If you make a commitment and don't fulfill it, you lose trust, and that ripples through the whole company." It's the same Cantrill statement, seen from the other end: doesn't matter that you have the best product on the market—if the system for delivering it to people isn't designed for actual load, customers remember the failure.
While engineers argued about how to bake pizza in motion, mathematicians since the 1960s had calmly been solving the question of how to divide it fairly. In 1967, Mathematics Magazine published a challenge problem: prove that if you cut a round pizza with straight cuts through an arbitrary point—not through the center!—into sectors of equal angle, and two eaters alternately take sectors every other one, then the portions come out equal. The first solution came from Michael Goldberg in 1968, and an elegant "proof without words" from Larry Carter and Stan Wagon in 1994: they cut sectors into small pieces so that each piece from an odd sector had an exactly congruent piece in an even one.
The funny edges of this picture are very precise. The condition "number of sectors divisible by four" is strict: Don Coppersmith showed that with four sectors or another non-divisible number there's no equality. And Rick Mabry and Paul Deiermann in 2009 clarified who exactly wins with unequal cutting: if there are six mod eight sectors and no cut goes through the center, then the eater who got pieces containing the cut point gets more pizza but—elegant compensation—less crust. There's even a game version: if cuts are arbitrary and eaters take turns taking pieces adjacent to eaten ones, the first player guarantees himself exactly 4/9 of the pizza, and no more even against a perfectly playing opponent (solution—Cibulka and colleagues, 2010).
This branch seems to me the best illustration of the whole Zume story. Cutting pizza is child's play? No: strict geometry with half a century of literature, from puzzle-problem to theorem in three dimensions (there, by the ham sandwich theorem, there exists a plane that simultaneously bisects base, sauce, and cheese). Baking pizza in a moving truck is a far more complex problem, but it too could have been solved as an engineering task. Zume instead solved it as a PR problem. The company's own patented slicer, by the way, divided pizza into exactly eight slices—the only place in the entire process where mathematics worked flawlessly.
This story has an almost identical twin that survived. Wonder Group by Marc Lore—the same one who sold Jet.com to Walmart for $3.3 billion—started exactly the same way: a fleet of trucks with mobile kitchens that parked at customers' homes in New Jersey and New York suburbs and cooked food on site. The idea of "cooking outside restaurant kitchens" mesmerized everyone equally.
Then the paths diverge. In January 2023, Wonder "pulled the plug": trucks closed, hundreds of employees laid off, model rewritten for stationary kitchens plus purchased rights to dishes from José Andrés, Bobby Flay, and Michael Symon—mix dishes from four restaurants in one order. This was followed by a $350 million round at a $3.5 billion valuation, purchase of Blue Apron for $103 million, and a hundred million from Nestlé, which is starting a partnership with... pizza and pasta under Wonder kitchen equipment. In business school analyses (The Spoon, 2023), this turnaround was formulated as four lessons, the main one being honest admission: "building your own fleet with built-in ovens is expensive madness," because you're pouring hundreds of millions into what other companies already know how to do, for marginally fresher pizza.
And here the puzzle closes. Wonder survived not because its product was better. It survived because it agreed in time that mobile kitchens don't exist as a working system—and moved capital to where operational mechanics are understood. Zume held onto the concept until the end, and the concept consumed it along with the money. Wonder, essentially, did what Sun didn't do: admitted that its "business mechanics" don't lie where it's proudly drawn on slides.
Three stories, one law. Sun Microsystems with the best engineers on the planet died because it couldn't sell its own server to a customer standing with money in hand. Zume with half a billion in the bank died because its "production revolution" didn't survive its first encounter with sanitary inspection and the physics of melted cheese. KFC with functioning kitchens became a negativity generator for a week because meal fulfillment wasn't designed for 10.5 million coupons. Cantrill's formula—"a company that got bored with the mechanics of running a business won't survive no matter how successful its strategy"—scales without remainder: substitute "cooking food" or "distributing coupons" for "running a business," and you get the obituary of any of these companies.
What I especially like about this construction is its symmetry. Operations failure not only kills a strong product but can save a weak one: Stellar Pizza, assembled by SpaceX alumni, honestly aims not at Neapolitan craft but at competing with Domino's—humbler ambition, but actually existing system. Honesty about where your mechanics actually work turned out more important than pitch genius. And if you're ever vetting a promising startup (or your own plan) for viability, don't ask how good the idea is. Ask who will wash the equipment between shifts—and what will happen to the cheese on a turn.