Pittsburgh Year by Year: The Rise of Robotics

• Introduction
• Chapter 1 1979: Seeds in the Rust – The Founding of the Robotics Institute
• Chapter 2 1983: Three Mile Island and the Machine That Entered the Core
• Chapter 3 1984: The Terregator Rolls – Outdoor Autonomy Takes Its First Steps
• Chapter 4 1986: Navlab 1 and the Birth of the Self-Driving Dream
• Chapter 5 1989: Neural Networks Take the Wheel with ALVINN
• Chapter 6 1992: Dante's Descent – An Eight-Legged Descent into Mount Spurr
• Chapter 7 1995: No Hands Across America – Five Thousand Miles on Silicon
• Chapter 8 1997: RedZone and the Business of Hazardous Environments
• Chapter 9 1999: The National Robotics Engineering Center Opens Its Doors
• Chapter 10 2000: Nomad Finds Meteorites in the Antarctic Ice
• Chapter 11 2004: Heartbreak in the Mojave – The First DARPA Grand Challenge
• Chapter 12 2005: Sandstorm and H1ghlander – Redemption in the Desert
• Chapter 13 2007: Boss Rules Urban Traffic – Carnegie Mellon Takes the DARPA Crown
• Chapter 14 2009: The Google Exodus – Pittsburgh Minds Build Silicon Valley’s Moonshot
• Chapter 15 2011: Out of the Earth – Autonomous Haulage Revolutionizes Global Mining
• Chapter 16 2012: The Warehouse Awakening – Kiva Systems and the Logistics Gold Rush
• Chapter 17 2014: Robotics Row – The Strip District Sheds Steel for Sensors
• Chapter 18 2015: The Uber Heist – Forty Scientists Vanish from NREC Overnight
• Chapter 19 2016: The World’s First Robotaxi Fleet Takes the Monongahela
• Chapter 20 2017: The Billion-Dollar Bet – Argo AI and the Automaker Stampede
• Chapter 21 2018: Tragedy in Tempe – A Global Reckoning Felt in Western Pennsylvania
• Chapter 22 2019: Aurora Rises – Consolidating the Long Haul
• Chapter 23 2021: Astrobotic and the Lunar Ambition – From Pittsburgh Mills to the Moon
• Chapter 24 2022: The Collapse of Argo AI and the Winter of Hype
• Chapter 25 2024: Steel Grit, Silicon Soul – The Capital of Autonomous Tomorrow

To understand the twenty-first century, one must first look at the confluence of three rivers. For more than a century, the Monongahela and the Allegheny carried the coal, limestone, and raw ambition that forged the modern world, meeting at the Point to form the Ohio. The soot hung so thick over Pittsburgh that midday streetlamps burned against an artificial dusk, but beneath that sulfurous canopy pulsed the undisputed workshop of the globe. Pittsburgh steel built the Brooklyn Bridge, armored the Allied fleets of World War II, and framed the skylines of Chicago and New York. To be a Pittsburgher was to believe in brute force, metallurgical alchemy, and the enduring certainty of heavy manufacturing. Then, with a speed that felt like a catastrophic fracture, the furnaces went cold. The economic collapse of the late 1970s and early 1980s did not merely idle the mills; it cleaved the region’s identity from its bedrock, sending unemployment soaring, gutting neighborhoods, and scattering generations of working families into a diaspora of survival.

Yet, amid the ash of deindustrialization, an unlikely resurrection was already underway. In 1979, inside the brick-and-mortar halls of Carnegie Mellon University, a handful of visionary computer scientists and engineers made a radical wager: that the next industrial revolution would not belong to muscle, steam, or iron, but to machines that could see, think, and act on their own. While the popular culture of the era envisioned robotics as the chrome-plated fantasy of science fiction films, the pioneers in Pittsburgh set to work on far grittier frontiers. They built machines designed to roll into the radioactive ruins of melted nuclear cores, to crawl into the mouths of erupting subarctic volcanoes, and to sweep the blue ice of Antarctica for ancient meteorites. The ethos of the mill floor did not vanish; it simply migrated into code, actuators, and sensor arrays. Heavy steel found a new companion in silicon.

This book is the chronicle of that transformation, told not as an inevitable triumph, but as a hard-fought, messy human drama unfolding year by annual milestone. The story of robotics is often told from the manicured campuses of Silicon Valley, where venture capital flows like water and software scales frictionless across the cloud. But true autonomy—machines moving through physical space, navigating mud, snow, unpredictable traffic, and harsh terrain—demands an engagement with physical reality that pure software companies often fail to appreciate. Pittsburgh understood the physical world because it had spent two centuries wrestling with it. The city’s winding, steep-graded streets, treacherous bridges, and punishing winters formed the ultimate crucible. If a machine could navigate Pittsburgh, it could navigate anywhere on Earth—or beyond it.

Beneath the technical breakthroughs, however, lies an intensely human story of obsession, loyalty, rivalry, and betrayal. The narrative tracked in these pages reveals how academic purists became reluctant entrepreneurs, how military moonshots funded in the wake of desert competitions ignited a multi-billion-dollar corporate gold rush, and how global tech giants descended upon Western Pennsylvania to strip-mine its intellectual capital. The legendary raid that saw an entire research laboratory hollowed out overnight for a rideshare titan is as much a part of this history as the elegant mathematics of early neural networks. Readers will witness the dazzling heights of the autonomous vehicle boom—when venture capital poured into the Strip District at unprecedented scale—alongside the sobering corrections, boardroom reckonings, and the tragic reminders of what happens when machine intelligence meets human vulnerability.

Pittsburgh Year by Year invites you into the machine shops, laboratory basements, and corporate glass corridors where the future was hammered into shape. From the early ponderous roll of the Terregator across a campus lawn to autonomous haulers reshaping massive open-pit mines, and from driverless cabs charting city bridges to private landers destined for the surface of the Moon, this is the story of how a wounded city reengineered itself into the indispensable capital of autonomy. The mills of the Mon Valley may have fallen silent, but the spirit that made them great never left the soil; it taught the machines how to think.

The air over Pittsburgh in the late autumn of 1979 was clearer than it had been in a century, but nobody was celebrating the view. For generations, the grit in the atmosphere had served as the local economic barometer. If your collar turned gray by lunchtime and the streetlights flickered on at two in the afternoon to pierce the orange haze over the Monongahela River, it meant the open-hearth furnaces at Homestead, Edgar Thomson, and the Jones & Laughlin works were gorging on iron ore, coke, and limestone. It meant shifts were running around the clock, paychecks were clearing, and the taverns along East Carson Street were packed two deep with men washing sulfur from their throats with draft Iron City beer.

By 1979, the sky was clearing because the fires were going out. Global steel markets had shifted with brutal swiftness, undercut by more modern, efficient facilities in West Germany and Japan, while corporate headquarters downtown scrambled to divest from heavy metals and shift their capital into financial services and real estate. Mill gates were locking shut across the Monongahela and Allegheny valleys. Layoff notices, once temporary inconveniences tied to the ebb and flow of union contract negotiations, became permanent pink slips. A civic panic settled over Allegheny County. Entire towns built around single mills faced immediate, existential obsolescence. The region was bleeding jobs, population, and pride, leaving millions of tons of cold structural steel standing like industrial tombstones along the riverbanks.

Two miles up the hill from the river basin, in the academic enclave of Oakland, sat Carnegie Mellon University. Founded in 1900 by Andrew Carnegie as a vocational school for the children of his millworkers, the institution had spent eight decades operating alongside the city’s manufacturing spine. But CMU possessed a different sort of capital. In the basement of Science Hall—a monolithic concrete bunker built in the brutalist style and later renamed Wean Hall—sat one of the finest collections of computer science intellects on the planet. Men like Allen Newell and Herbert Simon, who had helped invent the field of artificial intelligence in the 1950s, roamed the corridors debating heuristic search algorithms, cognitive architectures, and symbolic reasoning. Simon had just won the Nobel Memorial Prize in Economic Sciences the previous year, yet the university itself was not insulated from the misery below. The city’s tax base was crumbling, regional prestige was evaporating, and there was a pervasive, suffocating sense that Western Pennsylvania was quietly being written out of the modern world.

The man who decided to confront this decline with a machine was Richard Cyert, the president of Carnegie Mellon. Cyert was an economist by training, possessing a sharp nose for institutional survival and a pragmatic streak that favored balance sheets over academic sentimentality. He looked at CMU’s extraordinary Department of Computer Science and saw an engine of brilliance entirely detached from the physical reality around it. Newell, Simon, and the charismatic department head, Raj Reddy, were generating brilliant theoretical papers on knowledge systems, speech recognition, and game theory, but they were doing it largely inside mainframe computer memories. Outside the laboratory windows, the physical foundation of the American century was rusting into scrap. Cyert believed that if the university were to remain relevant, it needed to bridge the widening chasm between pure computation and physical industry.

Cyert began proselytizing for an audacious, slightly absurd venture: an independent research institute devoted exclusively to robotics. At the time, the word robot occupied an awkward space in the cultural and technical lexicon. To the general public, it suggested either the tin-plated comic relief of the newly released Star Wars film or the sinister automatons of pulp science fiction. To mainstream computer scientists, robotics was considered an untidy, low-status sibling to pure artificial intelligence. Pure AI lived in the clean, Platonic realm of formal logic, chessboards, and mathematical theorems. Robotics, by contrast, required getting one's hands dirty with gears, hydraulics, unpredictable electric currents, grease, and noisy analog sensors. Robots had a terrible habit of dropping things, burning out their motors, tipping over, and colliding with walls that the software insisted were not there.

Cyert found his operational visionary in Raj Reddy. Born in a rural village in Andhra Pradesh, India, Reddy had come to the United States to earn his doctorate under John McCarthy at Stanford before joining the Carnegie Mellon faculty in 1969. Reddy was a pioneer in machine speech understanding, a discipline that forced him early on to confront the chaotic messiness of the analog world. Unlike text, human speech was slurred, muffled, accented, and context-dependent. Reddy understood that the physical world was fundamentally analog, stubborn, and non-deterministic. He shared Cyert’s conviction that computing power was about to experience an exponential explosion, and that applying that power solely to screens and punch cards was a waste of human potential.

Reddy had a sprawling, panoramic ambition. He did not merely want to study industrial arms that screwed bolts into car doors; he envisioned autonomous machines capable of operating in spaces too dangerous, distant, or complex for humans to endure. He imagined robots harvesting crops, repairing deep-sea pipelines, cleaning hazardous waste, and navigating alien planets. To bring such machines to life, however, Reddy and Cyert needed money, and local banks and state agencies were wholly occupied with triaging the economic wreckage of the steel collapse.

Enter Thomas Murrin, the president of Westinghouse Electric Corporation’s Public Systems Company. Westinghouse was a cornerstone of industrial Pittsburgh, a historic titan that had electrified the nation in the days of George Westinghouse and Nikola Tesla. But like Big Steel, the manufacturing giant was feeling the cold breath of global competition, particularly from Japanese conglomerates that were rapidly modernizing their assembly lines using early programmable logic controllers and factory automation. Murrin was an executive with an engineering soul who recognized that American manufacturing was suffering from terminal inertia. The assembly methods inside Westinghouse plants were not fundamentally different from those used during the Korean War. Murrin had made trips to Japan and returned with an urgent, near-apocalyptic message for his fellow executives: American industry would either automate or evaporate.

When Cyert and Reddy approached Murrin with the proposal to build a dedicated robotics institute on the CMU campus, Murrin did not just sign off on a modest corporate grant; he threw the weight of Westinghouse behind the project. In the late summer of 1979, Westinghouse agreed to provide a seed commitment of roughly three million dollars over several years—a staggering sum for academic computer science at the time. Cyert added institutional support, and on paper, the deal was struck.

The announcement, made public in early 1979, was met with instant, furious skepticism from organized labor. The mill workers of the Mon Valley were already watching their livelihoods slip away; now, their hometown university, funded in part by one of the region’s largest employers, was openly establishing a center dedicated to inventing machines that could replace human hands on the assembly floor. Local union representatives publicly denounced the initiative. Editorial pages in the Pittsburgh Press and the Pittsburgh Post-Gazette ran anxious letters asking why money was being funneled into artificial men when real men were lining up at food pantries in Clairton and McKeesport.

Reddy and Cyert found themselves on the defensive, forced to articulate a counter-intuitive argument that felt cold to those living through the immediate trauma of unemployment. Their pitch was simple, if harsh: the jobs were not leaving because of automation; the jobs were leaving because of a refusal to automate. If Western Pennsylvania did not build the technologies that made factories hyper-efficient, overseas competitors would build them anyway, and the domestic market would collapse entirely. Furthermore, Reddy insisted that the scope of robotics extended far beyond the factory floor. The goal was to remove human beings from dull, dirty, and lethal work altogether—to take the worker out of the coal seam, out of the toxic chemical vat, and out of the sweltering heat of the forge.

With the funding secured, Reddy was named the founding director of the Carnegie Mellon Robotics Institute. He had an empty slate, a handful of borrowed offices in Science Hall, and a formidable challenge: he needed people who actually knew how to build physical hardware that could interact with the real world. Academic computer scientists in 1979 spent their days interacting with DEC PDP-10 mainframes and VAX minicomputers, punching commands into green-screen VT100 terminals. They wrote code that existed purely in silicon and phosphors. Very few of them possessed the skills to machine an aluminum axle, solder a noisy H-bridge motor controller, or calibrate an optical rangefinder.

To build an institute from scratch, Reddy needed an entirely new breed of researcher. He began scouring the globe for people who lived at the messy intersection of computer science, mechanical engineering, applied physics, and sheer, stubborn fabrication. One of his first and most critical recruits was Takeo Kanade.

Kanade was a young, soft-spoken, and remarkably brilliant associate professor of information science at Kyoto University. In the early 1970s, Kanade had written a doctoral thesis that effectively helped launch the field of computer vision, developing an automated program that could analyze human facial features from black-and-white photographs. It was a staggering achievement for an era when computers had mere kilobytes of memory and digital image sensors were primitive, expensive curiosities. Kanade was a master of the mathematical foundations of vision, understanding how light bouncing off a three-dimensional surface could be reconstructed into geometric forms through edge detection, stereoscopy, and perspective geometry.

Reddy invited Kanade to Pittsburgh in 1979 as a visiting scientist. What Kanade found when he arrived was a city shrouded in damp river fog and an academic department operating in a state of chaotic, infectious creative energy. In traditional Japanese academic departments of the era, hierarchy was rigid, progress was measured, and senior professors wielded absolute, conservative control over research topics. At Carnegie Mellon, Raj Reddy handed young researchers keys to empty rooms, promised them financial resources, and told them to build things that had never existed before. Kanade, who originally planned to stay for a single academic year, found himself utterly hooked by the lack of institutional boundaries. He would end up staying for the rest of his life.

While Kanade brought computational elegance and rigorous vision algorithms to the fledgling institute, the hardware side remained a gaping hole. Building robotic mechanisms required someone who understood structural loads, metal fatigue, kinematics, and the brutal physics of movement. It required someone who was unafraid of brute mechanical force.

That person arrived in late 1979 in the form of a towering, abrasive, and relentlessly driven Vietnam veteran named William L. Whittaker, known universally as "Red."

Red Whittaker was not an academic theorist. The son of a chemist and a fiercely determined mother, Whittaker had grown up on a farm in Pennsylvania, wrestling machinery, breaking horses, and running heavy equipment. He had enlisted in the United States Marine Corps, serving during the Vietnam War, an experience that permanently stripped him of any tolerance for excuses, administrative delay, or intellectual complacency. After leaving the military, Whittaker had arrived at Carnegie Mellon to complete his graduate studies in civil engineering, focusing on structural dynamics and soil mechanics.

Whittaker was large, barrel-chested, and sported a shock of unruly red hair. He spoke in clipped, gravelly, declarative sentences, frequently employing military terminology to describe academic engineering. To Whittaker, research was not a polite philosophical debate conducted at symposiums; it was a campaign to conquer physical ground. He viewed mechanical design with the visceral passion of a blacksmith and the precision of an aerospace engineer. He had little patience for cleanroom purists who spent years simulating robotic movement on a computer screen without ever cutting a piece of metal. Simulation, in Whittaker’s view, was a form of intellectual cowardice. The physical universe was filled with dirt, friction, mechanical slop, thermal expansion, and unexpected torque—variables that software models routinely ignored. If a machine did not move under its own power through actual dirt and rock, it simply did not count.

Whittaker had been watching the creation of the Robotics Institute from across campus in the Department of Civil Engineering. He saw immediately that the computer scientists upstairs were missing the structural grit required to build real machines. When Reddy opened the doors to the institute, Whittaker barged through them. He recognized that the heavy fabrication skills native to Western Pennsylvania—skills that the declining steel industry was casting aside—were precisely the ingredients needed to build large-scale robotic systems. The machinists, welders, and patternmakers who were losing their jobs in the Mon Valley possessed an intimate, generational knowledge of metal, hydraulic pressures, and mechanical tolerances. Whittaker saw no contradiction between Pittsburgh's industrial past and its robotic future; he saw an unbroken evolutionary line.

The third pillar of this early trifecta was Hans Moravec. Where Kanade was quiet and mathematically precise, and Whittaker was a hard-charging mechanical commander, Moravec was the resident eccentric genius, a prophet of machine intelligence whose visions bordered on the metaphysical. Moravec came to CMU from the Stanford Artificial Intelligence Laboratory (SAIL), where he had spent years working under John McCarthy. At Stanford, Moravec had built a legendary contraption known as the Stanford Cart. The Cart was a spindly, four-wheeled aluminum vehicle equipped with a swiveling television camera that inched along through hallways and parking lots.

The Cart was agonizingly slow. It would roll forward by a few feet, stop, swivel its camera across several discrete positions to capture a stereo baseline, broadcast the video signals over an analog radio link to a remote mainframe computer, wait several minutes while the machine chugged through stereo-correlation algorithms to identify obstacles, plan a short path, and then lurch forward another few feet. It moved at the speed of a dying snail, often taking five hours to cross an obstacle-strewn room. Yet, it was one of the very first machines on Earth to use visual processing to map an unknown physical environment and navigate through it without human intervention.

Moravec had grown frustrated by the bureaucratic limitations and shifting research funding at Stanford. In late 1979, Reddy lured him to Pittsburgh. Moravec arrived with an unshakable, radical thesis: that machines would not only achieve human-level intelligence, but would eventually supersede biological life entirely. He viewed robotics not as an industrial tool, but as the emergence of a new kingdom of life. His immediate technical goal, however, was far more practical: he wanted to build a vehicle that improved upon the clumsy architecture of the Stanford Cart. He wanted a machine that could move continuously, mapping its world using high-speed sonar and vision, transforming the chaotic stream of incoming sensory data into an internal, mathematical grid of the surrounding space.

By late 1979, the basement of Science Hall had transformed from a quiet computer room into a raucous, semi-feral workshop. The Department of Computer Science was accustomed to the rhythmic clatter of line printers and the steady hum of air conditioning units designed to keep mainframes cool. Now, the concrete basement corridors echoed with the whine of metal lathes, the acrid stench of soldering flux and burnt epoxy, the sharp sparks of angle grinders, and the roar of pneumatic lines.

The researchers who gathered in that basement formed an eclectic, tribal collective. Long-haired software hackers who had grown up on UNIX and the early ARPANET worked shoulder-to-shoulder with grease-stained civil engineers who owned their own torque wrenches. The dress code was flannel shirts, heavy steel-toed work boots, and jeans permanently stained with hydraulic fluid. Sleep was treated as an optional nuisance. Graduate students would write vision code for thirty hours straight, crash on a pile of foam packing material beneath an electronics workbench for four hours, wake up to a pot of bitter, scorched coffee, and resume debugging motor drivers.

There was little formal administrative structure. Raj Reddy managed the institute with a light, permissive touch, operating on the principle that exceptional researchers should be given funding, space, and absolute autonomy, then shielded from the bureaucratic interference of the central university administration. If a team needed an expensive optical encoder or a specialized hydraulic pump, the institutional paperwork was dealt with retroactively. What mattered was velocity.

The initial projects that took shape in that first year reflected the broad, wild spectrum of the institute's ambitions. At one end of the spectrum was factory automation, the core requirement of the Westinghouse grant. Westinghouse wanted intelligent assembly stations, advanced tactile sensors, and robotic arms that could inspect precision components without human error. To address this, teams began experimenting with sensor arrays that could measure minute forces on a robotic gripper, attempting to give machines a human-like sense of touch. They realized early on that an industrial arm was essentially blind and numb; if a part was misaligned by a fraction of a millimeter, a standard robotic manipulator would simply smash into it, stripping gears and snapping bolts. The CMU researchers set about outfitting industrial arms with early microprocessors and optical sensors, allowing them to adapt their paths in real time to errors in their workspace.

At the other end of the spectrum were the dreamers of autonomy. Hans Moravec had immediately begun sketching designs for his next-generation mobile platform, which he named the CMU Rover. Unlike the boxy, clumsy Stanford Cart, the Rover was envisioned as a marvel of omnidirectional movement. Moravec designed it with a cylindrical body and three steered wheel assemblies that could move and rotate independently, allowing the platform to slide sideways, rotate in place, and glide through tight interior spaces with ballerina-like grace. It was to be packed with an array of ultrasonic range sensors and an elaborate video camera assembly mounted on a high-precision track. It was vastly overly complex, technically extravagant, and structurally fragile, but it represented an uncompromising leap toward true environmental perception.

Whittaker, meanwhile, looked at Moravec’s delicate, polished brass-and-aluminum indoor designs and sneered. Whittaker had no interest in robots that required polished linoleum floors, air-conditioned rooms, and umbilical cables tethered to off-board mainframes. He believed that the real world lived outdoors, where there were no flat planes, no predictable light sources, and no clean electricity. Whittaker wanted machines with thick steel plates, deep-lugged rubber tires, and internal combustion engines. He began scouring the campus and local junkyards for components, drafting students from the civil and mechanical engineering departments into a culture that felt less like an academic seminar and more like a military boot camp. He took students out to the slag heaps that towered along the Monongahela River—vast, artificial mountains of glowing-hot refuse dumped by the blast furnaces over a century—and made them study how vehicles struggled to traverse the shifting, unstable slopes.

The technical hurdles the team faced in 1979 are difficult to overstate from the perspective of modern computation. Today, a microscopic chip costing a few dollars possesses millions of times more processing power than the entire computing inventory of Carnegie Mellon University in that inaugural year. In 1979, the digital image was an enormous computational burden. A single, grainy, black-and-white image of 256 by 256 pixels, with eight bits of grayscale depth, occupied 64 kilobytes of memory. A DEC VAX-11/780 minicomputer—a machine the size of a double-door industrial refrigerator that cost hundreds of thousands of dollars—could take several seconds or even minutes to run a rudimentary Sobel edge-detection algorithm across just one of these frames.

To give a robot the ability to see and react to its environment in real time was mathematically almost impossible with the existing hardware. A mobile robot rolling along at a brisk walking speed of three miles per hour covers more than four feet every second. If your perception system takes thirty seconds to process a single visual snapshot, your robot is effectively driving blind into the dark for 120 feet between glimpses.

This computational bottleneck dictated the entire philosophy of the Robotics Institute’s early work. Kanade and his students could not rely on raw processing power to brute-force visual problems. They had to be mathematically ingenious. They developed elegant algorithms that discarded ninety-nine percent of the visual information in an image, hunting only for the starkest geometric boundaries, the sharpest contrasts, and the most reliable features. They wrestled with lighting changes; a cloud passing over the sun completely altered the pixel values of a outdoor scene, causing an edge that the computer had identified as a wall to vanish instantly, or turning a harmless shadow cast by a tree into an impenetrable, terrifying virtual barrier.

The sensors themselves were painfully temperamental. Ultrasonic range sensors, originally harvested from consumer Polaroid autofocus cameras, were notoriously erratic. They functioned by emitting a high-frequency click and timing how long it took for the echo to bounce back from an object. But if the sound wave struck a wall at an oblique angle, it would bounce away like a billiard ball, never returning to the sensor, leaving the robot entirely convinced that the solid wall in front of it was empty space. If the sound struck soft clothing or curtains, the pulse was absorbed entirely. If two robots operating on the same frequency entered the same room, their sonar systems blinded and cross-talked each other into digital madness.

Hardware failures were an hourly occurrence. Electric motors drawing heavy stall currents would melt through primitive, breadboarded transistor circuits, filling the basement with the foul, unmistakable odor of fried silicon. Lead-acid batteries, heavy and dangerous, leaked corrosive acid onto student workbenches and gave off volatile hydrogen gas that required researchers to crack open the frosted basement windows to the cold autumn air.

Yet, despite the crude equipment and the staggering technical barriers, there was a visceral sense of manifest destiny inside the institute. The researchers knew they were the first people on Earth to grapple with these problems in a unified, dedicated institution. Before 1979, robotics research was fragmented: an isolated mechanical engineer in Tokyo worked on the linkage kinematics of an arm; a mathematician in Edinburgh worked on path-planning graph algorithms; a vision researcher at MIT wrote code to recognize geometric blocks in an idealized white box. At Carnegie Mellon, Reddy had forced all these disciplines into a single room and locked the door. The vision people were forced to watch their precious code fail because a mechanical axle rattled out of alignment; the mechanical engineers were forced to understand why their gear ratios were useless if the path-planning software choked on a memory limit.

Outside the walls of CMU, the city of Pittsburgh was continuing its descent into historic economic agony. In late 1979, the United States Steel Corporation announced the permanent closure of dozens of facilities nationwide, including massive cuts in the Monongahela Valley. Thousands more workers found their access to the mills cut off by chain-link fences and padlock chains. Across the region, real estate values plummeted. Foreclosure signs sprouted across neighborhoods like Swissvale, Braddock, and Homestead. High school graduating classes gathered for their final ceremonies and immediately boarded buses headed for Texas, California, and Florida, fleeing an industrial graveyard.

The contrast between the dying empire of heavy manufacturing and the embryonic world inside Science Hall was jarring. Visitors who descended the stairs into the basement found themselves stepping from a city that appeared to be running out of historical time into an enclave where the future was being constructed with frantic urgency. The metal chips flying off the lathes in Whittaker’s shop were coming off the very same grades of carbon and structural steel that the idled mills downstream had forged. But instead of being destined for a bridge beam, a railroad tie, or an automotive chassis, that steel was being turned down into drive shafts for automated rovers, welded into gantries for intelligent manipulators, and machined into mounting brackets for custom optical assemblies.

The local community remained deeply wary. When local television news crews occasionally visited the institute to capture b-roll of primitive mechanical arms moving geometric blocks or to film Moravec’s experimental Rover spinning its wheels on a test bench, the coverage was framed with a mixture of bemusement and dread. Anchors spoke of "mechanical men" being engineered in Oakland, asking aloud whether these machines were the nail in the coffin of the American working man.

Raj Reddy, Tom Murrin, and Richard Cyert never wavered in their response. They argued that an economic era had ended, irreversibly and without sentiment. The world of manual, repetitive, back-breaking labor was being rewritten by geopolitical forces that no regional policy or union contract could hold back. The only path forward for Pittsburgh was to reinvent itself as the intellect behind the machine. If the muscle of the industrial era was being outsourced or obsolete, the brain had to be engineered here, on the bluffs above the Monongahela.

As 1979 drew to a close, the Robotics Institute was no longer just an ambitious concept or an executive press release. It had physical reality. It had secured its multi-million-dollar war chest from Westinghouse. It had drawn together an unprecedented, volatile, and brilliant triumvirate in Kanade, Whittaker, and Moravec, overseen by the diplomatic and strategic vision of Reddy. It had established machine shops, computing nodes, and fabrication bays. Most importantly, it had established an ethos: a rejection of pure, disembodied theory in favor of functional, physical reality.

The machines built in that basement in 1979 were crude, slow, and perpetually broken. They possessed a tiny fraction of the intelligence of a common insect. They bumped into walls, lost their bearings, misread their sensors, and occasionally burst into flame on the test bench. But they were moving. In a city that was watching its centuries-old economic engine grind to a dead halt, the gears in Science Hall were beginning to turn. The seeds were planted directly into the rust, and beneath the frozen surface of a shattered industrial capital, something entirely new, autonomous, and relentless was taking root.