Mervin Kelly, an executive who ran Bell Labs during the post-war era, made researchers walk to lunch. The dining room sat at the far end of a long corridor in Murray Hill, New Jersey, and the route was deliberate. A solid-state physicist heading for coffee would pass a metallurgist, a vacuum-tube engineer, a topologist working on switching theory, and a chemist staining slides. Kelly believed if you forced these people to bump into each other often enough, something useful would eventually fall out. During the decades following World War II, that building produced the transistor, the laser, the solar cell, information theory, the photovoltaic cell, the charge-coupled device, Unix, and the C programming language.
The transistor came first. In late December 1947, John Bardeen and Walter Brattain demonstrated a working point-contact transistor to Kelly and a handful of executives in a basement lab. William Shockley, technically their boss, was not in the room. He was furious about that for years.
The open-door rule
Kelly enforced a policy that researchers found exhausting and that historians later called the single most important management decision in twentieth-century industrial science. Every office door had to stay open. If a colleague walked in with a problem, you were expected to drop what you were doing and help. This structural choice accelerated knowledge sharing across disciplines. Kelly just called it courtesy.
The building itself was designed to enforce the rule. Murray Hill’s main wing was a single long corridor, with labs branching off on both sides. You could not get from your office to the cafeteria without walking past dozens of other offices. Claude Shannon, who published A Mathematical Theory of Communication in 1948 while working down that hallway, used to ride a unicycle through it while juggling. Nobody asked him to stop.
What the salaries looked like
A senior researcher at Bell Labs in the mid-1950s earned roughly $12,000 a year, which is about $145,000 in today’s money. That was less than a comparable job at IBM and far less than what a physicist could make consulting for the Pentagon. People came anyway because Kelly promised something more valuable than cash: you could spend five years on a problem that might not work, and nobody would fire you, and nobody would ask for a quarterly progress report.
The funding came from AT&T’s regulated monopoly on long-distance telephone service. Every American household paying a phone bill was, without knowing it, financing basic research into semiconductors. The 1956 consent decree with the Department of Justice required AT&T to license Bell Labs patents at reasonable rates to other American companies. The transistor patent was made widely available. Texas Instruments, Motorola, and Fairchild Semiconductor all built their early businesses on it. Silicon Valley exists because of a regulatory settlement signed in Washington.
Shannon’s office
Claude Shannon’s office at Murray Hill contained a chess-playing machine he had built himself, a mechanical mouse named Theseus that could solve a maze, and various unusual contraptions. He worked on whatever interested him. In 1948 what interested him was whether you could measure information mathematically, the way you measure mass or distance. The paper he produced that year invented the bit as a unit, defined channel capacity, and became the foundation of every digital communication system built since. He wrote most of it at night, alone, with the door closed. The open-door rule had exceptions for the people Kelly considered geniuses, and Shannon was the largest exception.
The laser was an accident of two people arguing
Charles Townes had developed the maser, an amplifier for microwave radiation, at Columbia in the early 1950s. He came to Bell Labs as a consultant. In the late 1950s he and his brother-in-law Arthur Schawlow, a Bell Labs physicist, started arguing over lunch about whether you could build a maser that worked at optical frequencies. Their joint paper, published in Physical Review in 1958, described what would become the laser. Schawlow filed the patent through Bell Labs. He later said the entire project would not have happened if he and Townes had worked at different companies, because no other industrial lab in America would have let two physicists spend a year on an idea that had no obvious telephone application.
The Unix story is stranger than the transistor story
In 1969, AT&T pulled out of a multi-company operating-system project called Multics. Ken Thompson, a Bell Labs computer scientist, found himself with a spare DEC PDP-7 minicomputer and nothing to do. His wife took their young son to California for several weeks to visit family. Thompson used the time of solitude to write an operating system, a shell, an editor, and an assembler. He called it Unics, partly as a joke. The name became Unix. Dennis Ritchie joined him, invented C to rewrite the system in a portable language, and by the early 1970s Unix was running on machines all over Bell Labs without management ever formally approving the project. Founders building infrastructure companies today still trace their lineage back to that PDP-7.
What Kelly actually believed
Mervin Kelly gave a speech in 1950 to the Royal Society in London where he explained his hiring philosophy. Historical accounts suggest he looked for people who could think independently and collaborate effectively with colleagues. He insisted on mixing theorists with experimentalists in the same hallways, not in separate buildings. He insisted on hiring metallurgists and chemists alongside physicists, because semiconductors required all three. He refused to organize the labs by project. He organized them by discipline, then forced the disciplines to collide in the corridor.
What Kelly built created an environment where people felt safe taking intellectual risks. Paul Santagata, head of industry at Google, told Harvard Business Review that Google’s two-year internal study on team performance found psychological safety — the belief that you will not be punished for mistakes — was the single strongest predictor of which teams produced breakthroughs. Bell Labs ran that experiment thirty years before anybody had a name for it.
The pay was not the point
John Bardeen, who shared the 1956 Nobel Prize in Physics for the transistor, left Bell Labs in 1951 for a professorship at the University of Illinois. He took a pay cut. He said the reason was that Shockley, by then promoted to research director, had been preventing him from working on superconductivity. Bardeen wanted to study superconductivity. So he left. At Illinois he won a second Nobel Prize in 1972 for the BCS theory of superconductivity. He remains the only person to have won the physics Nobel twice.
Shockley left in 1956 to start Shockley Semiconductor Laboratory in Mountain View, California. He brought eight researchers with him. Within two years, all eight had quit because Shockley was, by every surviving account, impossible to work for. They founded Fairchild Semiconductor. Two of them — Robert Noyce and Gordon Moore — later founded Intel. The entire semiconductor industry of Northern California descends from people fleeing the worst manager in Bell Labs history.
The cafeteria
The Murray Hill cafeteria had round tables on purpose. Rectangular tables let cliques form at the ends. Round tables forced strangers to sit next to each other. Kelly had read something about this in a management journal and applied it. Researchers who ate lunch alone were noted by their group leaders. Eating with the same three people every day was discouraged. The point of lunch was not lunch.
The Competing Values Framework would classify what Kelly built as a hybrid of clan and adhocracy — collaborative, mentorship-heavy, but also tolerant of risk and experiment. Most modern companies pick one. Kelly insisted on both, and made the architecture enforce it.
What broke the model
The 1984 breakup of AT&T into seven regional Bell operating companies cut the funding pipeline. Bell Labs was split, partially privatized, sold, restructured, and eventually absorbed into Nokia in 2016. The Murray Hill building still exists. It still has the long corridor. The doors are no longer required to be open.
The economics that made it possible — a regulated monopoly funneling consumer phone bills into open-ended basic research, with patents licensed away for free — have not existed in American industry since. Google, Microsoft, and Meta run large research labs, but they patent aggressively and the work is tied to product roadmaps. Research on organizational culture consistently finds that the cultures producing the most original work are the ones least focused on short-term returns. Kelly had a twenty-year runway. Modern research directors have a twenty-month one.
The number that still doesn’t make sense
Bell Labs employees won nine Nobel Prizes for work done at the labs. They won four Turing Awards. The transistor alone, by some economic estimates, generated more than two trillion dollars in downstream global GDP through the semiconductor industry it created. AT&T’s total cumulative investment in Bell Labs research from 1925 through 1984 was, in inflation-adjusted terms, roughly comparable to what Meta spends on capital expenditure in a single year now. The return on that money is incalculable because nobody has ever found a way to value what Unix is worth, or what information theory is worth, or what the laser is worth across medicine and communications and manufacturing combined.
The building is still there on Mountain Avenue in Murray Hill. The cafeteria is still at the end of the long hallway. If you walk it today, the corridor is quiet. The doors are mostly closed. Somewhere in the basement is the room where Bardeen and Brattain showed Kelly the transistor on a Tuesday afternoon four days before Christmas, with a strip of gold foil pressed against a sliver of germanium, and a small voice coming out of a speaker because the silicon was, for the first time in history, amplifying it.