Bell Labs was four structural conditions, and it could not have kept them

Start with the anecdotes, because everyone does and because they are load-bearing in a way that is easy to miss.

Claude Shannon rode a unicycle down the Murray Hill corridors while juggling. He built Theseus, a maze-solving mechanical mouse, and a box whose only function was to reach out and switch itself off. Harold Black worked out the negative-feedback amplifier on his morning commute and sketched it on a newspaper aboard the Lackawanna Ferry. In 1943 Alan Turing, in the United States on cryptographic liaison, ate lunch in the same cafeteria as Shannon and compared notes.1 Two radio astronomers at Holmdel spent months convinced their antenna was dirty, scrubbed the pigeon droppings out of it, and found that the hiss they could not remove was the afterglow of the Big Bang.

Ken Thompson, alone for a stretch while his family travelled, wrote the first version of Unix on a cast-off PDP-7. Asked much later what he would change about it, he said he would spell creat with an e. That is sixty seconds of regret across an entire career, and it is what it looks like when the shape was right the first time.

These stories get told as evidence of a mood, and they are. But the mood has a mechanism, and it is more interesting than the mood.

What the anecdotes are actually evidence of

Take Black on the ferry. A man is thinking hard about a problem on his commute because the problem is well-posed and he has been given years to solve it. Take Shannon’s unicycle. A researcher is visibly, unembarrassedly playing in a corridor, which tells you nobody was counting his hours or asking what he shipped this quarter. Take Penzias and Wilson scrubbing the antenna. Two people are permitted to spend months eliminating a nuisance rather than publishing something, and the institution absorbs the cost.

The corridors themselves were the tell. Murray Hill’s hallways were built long on purpose, so that reaching lunch meant running a gauntlet of people who might ambush you with a question.1 That is not a vibe. That is a floor plan with an intention encoded in it, commissioned by someone who believed collision produces work and was willing to pay for the square footage.

Four legs

The culture is best read as four structural conditions holding each other up. Not a management philosophy, not a hiring bar, and emphatically not something a leadership team decided to have.

Patient money, and a known quantity of it. The Labs ran on a license fee amounting to roughly one percent of Bell System revenue. That is publicly funded basic research administered by a private monopoly, everyone involved understood it as such, and the obligation it implied was mostly felt rather than enforced. The relevant property is not the amount but the patience: money that does not ask what it bought this year buys a different kind of question.

One enormous real system, generating questions faster than anyone could answer them. The telephone network was treated as a single total physics problem, from the copper in the ground to the psychoacoustics of the human ear. Nobody had to invent a research agenda, which is why basic research there was never unmoored from anything. It also produced the institution’s signature move, which recurs across five decades and several disciplines: investigate the noise, discover a universe. Karl Jansky was sent to Holmdel to hunt down static on the transatlantic radiotelephone and founded radio astronomy instead. Penzias and Wilson, in the same place forty years later, did it again.

The whole stack under one roof, with an obligation to help. Physicists, chemists, metallurgists, circuit engineers, and a machine shop, in a building designed to make them collide. The norm that mattered most was closer to a duty than a courtesy: if a colleague came to your door with a question, you helped them. That is what makes “a thought before lunch and someone down the hall makes it out of germanium by Thursday” a description of a Tuesday rather than a legend.

Antitrust pressure that made openness the survival strategy. The 1956 consent decree settling the Justice Department’s suit required AT&T to license its patents broadly and to stay out of computing. The transistor was effectively given away. Because AT&T could not sell software, Unix went to universities for the price of the tape with source included, and BSD, the hacker culture, and eventually Linux all fall out of that restriction. The institution’s outsized influence on the world is substantially downstream of a legal prohibition on selling things.

The irony that makes the question hard

The four legs did not fall independently, and this is the part that makes every “why can’t we build another Bell Labs” conversation founder.

The same antitrust settlement that made the Labs generous is what eventually dismantled the monopoly that paid for them. Openness was a legal obligation before it was a virtue, and the legal apparatus that imposed it was the same apparatus that, in 1984, broke up the rent it was regulating. The conditions were self-undermining. Greatness and death issued from one source.

The timing is almost too neat. In 1984 the licence fee vanishes, roughly a third of the research staff leaves for Bellcore, and research begins having to justify itself against a competitive equipment business rather than a guaranteed monopoly. The same year, the Labs patent the Karmarkar algorithm and the mathematical community is scandalised, because patenting a linear-programming method would have been unthinkable when giving things away was the house strategy. And in 1985, one year after divestiture, C++ ships as a licensed commercial product, because AT&T could finally be in the software business.

Unix spread because it could not be sold. C++ spread because by then it could. Nothing about the quality of either explains the difference in how they travelled.

Complications worth keeping

A version of this essay that flattered its subject would leave things out, so here are the ones that matter most. The ascent was not smooth: the Labs shrank in the early 1930s, with cuts and shortened weeks. The golden-age story launders out that Bell Labs ran the Nike and Safeguard anti-ballistic-missile programmes during Vietnam, over real internal dissent. And the complaint that the place is not what it used to be starts around 1970 and never stops, which makes any particular decade’s version of it suspect. It is continuous, and it was said by people who were themselves in the middle of doing extraordinary work.

The Schön fraud, exposed in 2002, belongs here for a specific reason. A young physicist fabricated spectacular results at an impossible cadence, and the thing that should have caught him early was the senior colleague who wanders in and asks to see the raw data. That role had been hollowed out by cuts and by pressure to produce. The fraud reads better as evidence about the institution than about the individual.

Does anything have all four

Here is where the pondering has to stop rather than resolve, because I do not think the answer is obvious and I distrust the confident versions of it.

It is easy to point at the modern research labs and note what they have. Several have patient money at a scale that dwarfs a percent of Bell System revenue. Several sit inside a single enormous real system that generates hard questions honestly, which is more than most universities can say. A few have genuinely built for collision, though the corridor has been replaced by the campus and it is not clear the substitution works.

What none of them obviously has is the fourth leg, and the fourth leg is the strange one. Openness at Bell Labs was not a value the institution chose and could therefore un-choose. It was imposed from outside by a government that had it by the throat, and it lasted precisely as long as the legal circumstance that produced it. Modern labs publish when publishing serves them and stop when it does not, which is what a value looks like when nothing is holding it in place.

So the honest form of the question is not whether anyone could recreate the mood. It is whether the mood can exist without somebody being compelled. That is an uncomfortable thing to conclude, and I am not sure it is right. It might be that the compulsion was incidental and some other arrangement produces the same generosity. It might be that the whole thing was a happy accident of mid-century American antitrust policy that will not recur, and that the correct response is to enjoy what fell out of it rather than to hope for a repeat.

I notice I want the first answer to be true. That is not evidence for it.