Guido van Rossum: The Rise and Fall of Python's Benevolent Dictator

He wrote Python over a 1989 Christmas break. By 2018 it ran AI and the web, and a fight over the walrus operator (PEP 572) drove him to resign as BDFL.

I. The Cold Room

Guido van Rossum was tired.

It wasn't the kind of tired a nap fixes. It was the kind that settles into the joints of a sixty-two-year-old man and stays there. Early on the morning of Thursday, July 12, 2018, at his home in Belmont, California, he began to type to the python-committers mailing list, the small closed room where the inner circle of his kingdom did its business.

Tak. Tak. Tak-tak.

He was the Benevolent Dictator For Life. The Benevolent Dictator For Life! The title began as a joke, scribbled into the notes of a 1995 meeting in Virginia. Over twenty-three years it hardened into something that was no joke at all. He had the final word on every comma, every keyword, every bit of syntax in a language that had quietly become one of the load-bearing walls of civilization. Python ran Instagram, the recommendation engines at Netflix, the playlists at Spotify. It was the mother tongue of artificial intelligence, of astronomers and geneticists and hedge-fund quants and twelve-year-olds making their first turtle draw a square. And every one of them, in some abstract and terrible sense, answered to him.

No scandal brought him down, no lawsuit, no hostile takeover. Two characters did, a colon and an equals sign:

:=

They called it the walrus. Tilt your head and you see the eyes and the tusks. For months it had been the cause of a civil war: hundreds upon hundreds of emails, a blizzard of Reddit comments, strangers questioning his judgment, his competence, his fitness to rule. The night before, he had formally accepted the proposal, PEP 572. He had won, and winning had cost him so much that he no longer wanted the prize.

He typed:

Now that PEP 572 is done, I don't ever want to have to fight so hard for a PEP and find that so many people despise my decisions.

Despise. He didn't write disagree with, or question. He wrote despise. That was the word for it.

I would like to remove myself entirely from the decision process.

He was Dutch, and he had never been a man for speeches, least of all dramatic ones. So the message that ended one of the longest reigns in open-source software came out almost offhand. He mentioned "that bus lurking around the corner," the programmer's grim joke about a creator stepping off the wrong curb. He noted that he wasn't getting younger. Then, with that dry deflection of his: (I'll spare you the list of medical issues.)

I am not going to appoint a successor.

So what are you all going to do? Create a democracy? Anarchy? A dictatorship? A federation?

And then the last line, the one people would quote for years:

I'm tired, and need a very long break.

He hit send. The King didn't abdicate in a scene of crashing thunder. He got up from the throne, left the crown on its cushion, and walked out of the room.

How does a holiday hobby, a toy to keep idle hands busy, become a weight that crushes the one who made it? To find out you have to go back to the real cold. Amsterdam, December, 1989.

II. Christmas in Amsterdam

Snow along an Amsterdam canal in winter Winter on Amsterdam's eastern canals. The kind of dark, still season that produced Python.

Picture the city in the last week of the 1980s. The canals have gone black and still, the gables lean over the water like tired old men, and darkness comes down by four in the afternoon. The Berlin Wall fell six weeks ago. History is happening everywhere, and on the eastern side of the city a thirty-three-year-old programmer is bored.

He works at the Centrum Wiskunde & Informatica, the CWI, the Dutch national research center for mathematics and computer science. Over the holidays the offices are closed. The young man with the high forehead and the spectacles has a computer at home and, as he'll later put it, not much else on his hands. He also has a grudge that's really more of a grief.

Entrance of the CWI in Amsterdam The Centrum Wiskunde & Informatica in Amsterdam, where ABC was built and Python began.

The grief is called ABC.

From 1982 to 1986 he had worked on the ABC programming language under Lambert Meertens and Steven Pemberton. ABC was beautiful, an academic's dream of programming for scientists, teachers and beginners, and its most radical idea was significant indentation: the visual shape of the code was its logical shape. No thicket of curly braces and semicolons. You indented, and the indentation meant something.

ABC was also a failure. It was too pure to survive, a sealed world that couldn't talk to the operating system, and the hackers who most needed a clean, high-level tool walked right past it.

By 1989 he had moved on to Amoeba, a distributed operating system project at the CWI, whose team badly needed a scripting language. So, in the dark of that Christmas, he started writing one that would keep ABC's soul but open the windows to Unix and C.

He needed a name. He'd been reading the published scripts of Monty Python's Flying Circus, and he wanted something short, unique, a little irreverent. Python. That was the whole decision, a joke for people who knew the joke.

Stop here a moment, on the young man in the dark, because this is the last moment of perfect innocence in the story. No community, no mailing list, no corporation, no AI, no walrus. One man, a keyboard, a holiday, and the private pleasure of building a thing exactly the way he thinks it ought to be built. He'll spend the next three decades trying to get back to this room.

III. The Wager

The 1990s Python wordmark The early Python wordmark, designed by Just van Rossum, Guido's brother.

In February 1991, Python 0.9.0 was posted to alt.sources, one of the dusty Usenet bulletin boards of the early internet. No launch event, no press release. A file appeared, and a few strangers downloaded it.

Hidden inside that file was a wager, the most important bet in the whole story. He'd put it into words years later, in Python's style guide:

Code is read much more often than it is written.

In 1991 the macho culture of programming prized density. Perl hackers wrote one-liners that looked like a cat had walked across the keyboard, and they were proud of it. Cryptic was cool. Unreadable meant you were smart.

Van Rossum bet the other way. If a language made the clear path the easy path, the whole ecosystem would drift toward clarity like water running downhill. And he wrote it into the grammar. In Python, indentation wasn't a suggestion. It was law. Indent wrong and the program wouldn't run. The machine itself would make you write code a human could read.

Years later the core developer Tim Peters boiled the ethos down into nineteen lines of something close to poetry, The Zen of Python, which you can still summon from any interpreter by typing import this. Beautiful is better than ugly. Readability counts. And the line that would come back, nearly three decades later, as a weapon:

Zen of Python by Tim Peters

There should be one, and preferably only one, obvious way to do it.

Van Rossum didn't write those words. He didn't have to; they described what he'd already built. Python was idiosyncratic, approachable, a little goofy, and because anyone could pick it up and understand it, it spread from hand to hand and machine to machine like a friendly virus.

IV. The Dictator Who Wouldn't Shout

Windmills on the Overwaard polder at Kinderdijk The polder at Kinderdijk. Dry land is held there by agreement, and that's the Dutch art of consensus.

A crowd gathered around Python, a few dozen people, then hundreds, then thousands, sending bug fixes and feature requests and patches. Somebody had to decide which ones went in. In April 1995, at a meeting of early Python users at CNRI in Virginia, Ken Manheimer's notes named van Rossum "benevolent dictator"; Barry Warsaw is credited with adding "for life." Benevolent Dictator For Life. BDFL.

The joke part was the Benevolent and the For Life. The Dictator part was real. He had the final say on the design of the language. Full stop.

Other open-source founders came to be called BDFLs, but the most famous of them was Linus Torvalds, and the contrast between the two men tells you almost everything.

Torvalds, the creator of Linux, ruled his kernel like a warlord. He swore. He shamed. He sent emails that read like declarations of war, tearing into developers who'd submitted bad code, and he defended his right to do it. Friction forged steel. He liked the fight. (Two months after van Rossum quit, Torvalds would publicly apologize for his behavior and take a break of his own.)

Linus Torvalds at LinuxCon Europe Linus Torvalds, the other dictator for life, who thrived on the fight.

Van Rossum was nothing like that. He came from a country that had pulled itself out of the sea through cooperation. The Dutch call it polderen, after the polders, reclaimed lands kept dry only because farmers and towns and rival interests agreed, painstakingly, on who would maintain which dike. Government by consensus and patient compromise, because if anyone stops pumping, everyone drowns. That was his temperament. He didn't want to yell. He wanted to read the code, weigh the trade-offs, and make people feel heard.

In 1999, working by then at the Corporation for National Research Initiatives in Reston, Virginia, he wrote a funding proposal to DARPA, the Pentagon's research arm, with a title that said everything: Computer Programming for Everybody. Programming as a basic literacy, for high schoolers and hobbyists and ordinary people, not a priesthood. Python would be the vehicle.

It was a generous, democratic vision, and it had a hidden price. If you build a language for everybody, everybody comes. And everybody has opinions.

As Python grew, the process for changing it was formalized as the Python Enhancement Proposal, the PEP. Anyone could write one. Each was argued over in public on the mailing lists. It was open and, on paper, democratic.

In practice it was a funnel. All the anxiety, ambition and ego of tens of thousands of brilliant, stubborn engineers poured down it, and at the narrow end sat one man. He read the threads, hundreds of messages on a single question. He weighed technical merit against aesthetic purity, the beginner against the expert. He was the pressure valve for the whole community, and he wasn't built like Torvalds. He felt it.

V. Mondrian

Guido van Rossum speaking at Google I/O 2008 At Google I/O, 2008. Google paid him to spend half his time on his own language.

In December 2005, Google hired him.

This was Google at its most intoxicating, all lava lamps and free lunches, and it ran on his language. Its engineers' motto, made public by the Googler Alex Martelli in 2007, was Python where we can, C++ where we must. Google gave van Rossum half his working time to spend on Python itself. He was no longer a researcher with a hobby. He was a celebrity with a corporate badge, the subject of what one PyCon photographer would caption, without irony, the "obligatory Guido shot."

With the other half he built Mondrian, an internal code review tool named, being who he was, after a Dutch painter. It moved Google's review ritual onto the web, line by line, and taught a generation of engineers his deepest belief: human eyes reading code are the last defense against chaos.

Piet Mondrian, Composition II in Red, Blue, and Yellow (1930), namesake of Google's code review tool.

The irony would take years to ripen. He was building a culture of intense, line-by-line, public scrutiny, and that same culture, scaled up past the point where anyone knew each other, would one day be turned on him. On a small team, review is intimate: people who know and trust each other, catching each other's mistakes. Scale it to thousands of strangers on the internet and the intimacy goes sour. It becomes a tribunal.

VI. The Schism

Guido van Rossum at OSCON 2006 OSCON, 2006. This was the era of "Python 3000", before the long schism began.

The first real crack ran right down the middle of the kingdom.

By the mid-2000s, van Rossum could see that some of Python's early decisions, made in a hurry by a young man in a cold winter, were holding the language back. The worst was text. Python 2 muddled human-readable text and raw binary bytes. In 1991 that hadn't mattered. In a global internet of Chinese and Arabic and Hindi and emoji, it was a disaster.

The fix wasn't a patch. It was surgery, and it meant deliberately breaking backward compatibility. Python 3.0 arrived in December 2008, following the blueprint of PEP 3000. It was bold and it was right, and it meant that millions upon millions of lines of Python 2 code wouldn't run on it without being rewritten.

The migration was supposed to be brisk. It took a decade.

It was a grinding war of attrition. Big companies with mountains of technical debt refused to move. Red Hat Enterprise Linux and CentOS 7 clung to Python 2.7 for years. The volunteers who maintained the open-source libraries burned out making their code run on both versions at once. The community split into camps, the Twos and the Threes.

Van Rossum spent those years defending the decision to angry people, playing diplomat between factions of his own. Python 2's death date, first set for 2015, slipped five years to January 1, 2020, with a last-gasp release, 2.7.18, that April.

In the end the transition worked. But victory isn't the same as peace. The schism used up enormous amounts of the Dictator's political capital and stamina. He came out of it weary.

What he didn't know was that the real explosion hadn't even started.

VII. The Serpent Eats the World

Rows of servers in a data center The serpent's habitat: the server farms where Python became the mother tongue of machine learning.

Then, in the 2010s, the ground shifted under the whole of computing. Artificial intelligence and machine learning, which had lived in academic papers for decades, became almost overnight a corporate necessity. The people building them were mathematicians and physicists who thought in matrices and had no patience for Java's ceremony. They wanted a clean, readable wrapper around brutally complicated math.

They chose Python, for exactly the reasons van Rossum had bet on in 1991. NumPy and pandas turned it into a numerical powerhouse, and then it had a gravity of its own: researchers used it, so the tools were built for it, so more researchers came. When Google's TensorFlow and Meta's PyTorch arrived, and later the large language models of OpenAI and everyone else, they all spoke Python.

Look at the size of it! By 2018, nearly 39 percent of developers in Stack Overflow's survey used Python, the fastest-growing major language in the world, and behind them poured biologists, astronomers, accountants and teenagers who had never thought of themselves as programmers at all.

The language had become the scaffolding of the AI revolution and a load-bearing pillar of the global economy. Every machine-learning lab, every startup chasing the next model, every billion-dollar bet on artificial intelligence, bets that by 2026 would swell into the trillions, now ran through a language whose final arbiter was one tired man reading his email. A small tweak under the hood could move a Fortune 500 company's server bills by millions. Every decision landed on him.

He lived it from the inside, too. At Dropbox, which he joined in 2013, he spent years helping migrate more than five million lines of code from Python 2 to Python 3, in the trenches of the very war he had started. Meanwhile the mailing lists filled with a new generation who saw him not as a peer but as a gatekeeper between them and the feature they needed right now. The intimacy was gone. It only needed a spark.

VIII. Tusks

A walrus on Svalbard The walrus. Tilt your head at := and there are the eyes and the tusks.

The spark was a walrus.

In February 2018 a developer named Chris Angelico, joined by Tim Peters and van Rossum himself, drafted PEP 572: Assignment Expressions.

On paper it was about as dull as language design gets. Python had always kept a strict wall between statements, which do something, like store a value in a variable, and expressions, which evaluate to something. PEP 572 proposed a small door in that wall: a new operator, :=, that would let you assign a value and use it at the same moment, inside an expression. It would save computing expensive things twice and untangle clumsy loops.

Take the old, familiar ritual of reading a file in chunks:

while True: chunk = file.read(8192) if not chunk: break process(chunk)

With the walrus, it collapses into this:

while chunk := file.read(8192): process(chunk)

To the purists none of that mattered, because to them PEP 572 was heresy.

They had the scripture to prove it. There should be one, and preferably only one, obvious way to do it. Python already had a way to assign a variable. Now there would be two. To the faithful this was the road to Perl, to show-off one-liners that would baffle beginners. Readability counts! Readability counts! And here was the Dictator himself, they said, selling it out.

What should have been a dry argument about parsing rules turned, almost immediately, into something feral.

The volume of email was staggering. Thread after thread on python-dev and python-ideas, hundreds of messages, every one a little hotter than the last. People dug in. They accused the authors of smuggling C-style syntax into a language that had been designed specifically to reject it. And the poison of modern internet discourse seeped into the technical debate, the way it seeps into everything. People were no longer just disagreeing with the proposal. They were attacking the people: questioning their competence, their motives, their intelligence. And by extension, unavoidably, they were attacking the man at the narrow end of the funnel.

He looked at the evidence, the way he always did. In huge real-world codebases, including Dropbox's, programmers were already doing awkward and risky things to save a line. Used with care, the walrus would make real code cleaner. He believed in it.

On July 9 he announced his intention to accept. On the evening of Wednesday, July 11, 2018, he formally accepted PEP 572 for Python 3.8.

By then the backlash had been building for months, and his decision only fed it. He had ruined the language. People announced they'd ban the walrus from their projects. People talked about forking Python.

This was a man who had spent nearly three decades building a community on collaboration and mutual respect, who came from a culture of the polder and the patient meeting, who had believed that if you made things readable people would understand each other. He was hit with a wave of contempt so large that he couldn't do anything with it. He couldn't answer it or reason with it. It was too much to read and too much to absorb.

Being the final arbiter, the one person who has to soak up the collective fury of thousands of engineers over a colon and an equals sign, had finally changed something between him and the thing he loved. The language that had been his joy had become the source of his misery.

He had won the walrus war and lost the thing he'd been fighting for.

The next morning, roughly fifteen hours after he accepted the proposal, he opened the python-committers list, and you know the rest.

IX. Transfer of Power

Guido van Rossum on stage at PyCon 2013 Van Rossum on stage at PyCon, 2013.

The subject line was almost comically bureaucratic: Transfer of power.

The shock spread across the technology world within hours. Guido steps down. Python's BDFL quits. Reddit threads filled with hundreds of comments. Podcasts rushed out emergency episodes. Developers on the far side of the planet woke to news that felt, absurdly, like the death of a pope, or a parent leaving home.

Look at what he didn't do. He didn't name an heir or hand the crown to a trusted lieutenant. He didn't present a constitution. He did something stranger and, in its way, braver. He threw the question back at them.

Create a democracy? Anarchy? A dictatorship? A federation?

I'll still be here, but I'm trying to let you all figure something out for yourselves.

He would stay on as an ordinary core developer, he wrote. He'd be around to mentor. But the burden of being the last word on everything was over.

The King wasn't overthrown. Nobody stormed the palace. He just stood up, set the crown down on the cushion, and walked out of the room, and behind him the court stood frozen, staring at the empty throne.

X. The Republic

Guido van Rossum with attendees at PyCon 2012 Van Rossum with the photographer, Mahmoud Hashemi, and Kurt Rose at PyCon 2012. The community that would one day govern itself.

What came next was something close to existential dread.

How does a dictatorship invent democracy? How do you decide how to make decisions when your only mechanism for deciding has just walked out the door? The core developers were brilliant engineers, not political philosophers, and they found themselves in a constitutional convention they'd never asked for.

They had good reason to be afraid. Open-source governance is a fragile thing, held together by volunteer labor and goodwill. When a founder leaves, the vacuum is usually filled in one of two ugly ways: a corporation moves in and captures the project, or it shatters into warring forks. The history of free software is littered with the remains of projects that didn't survive their creator's departure.

The models on offer each had a flaw. There was the benevolent dictator, the old Python and the current Linux, where everything hung on one person's health, sanity, and patience, which was exactly the problem they were trying to solve. There was the corporate team, the model of Google's Go, where a company's priorities could always quietly override the community's. There was the meritocratic voting of the Apache projects, which tended to be too heavy, too bureaucratic, too slow for a living language. And there was the elected council, the model of projects like Django, which needed elections, voter lists and rules, and carried the constant risk of gridlock.

Here's the elegant twist. By refusing to name a successor, van Rossum had forced his community to find, in the words of their own scripture, one obvious way, and so they would own it. Seven competing governance proposals were written. In December 2018 the core developers voted, and the winner, by Nathaniel J. Smith and Donald Stufft, became PEP 13, arguably the most important document in Python's history that contains no code at all.

PEP 13 created a five-person Steering Council, elected by the core team after each major release. The authors had studied the failure modes and built in defenses against them. To keep any one company from capturing the language, no more than two council members could work for the same employer; if a third was elected, they were disqualified. Elections used a carefully designed multi-winner voting system meant to produce broad consensus rather than factional victory. And there was a fail-safe, the vote of no confidence: if the council lost the community's trust, the core team could remove a single member or dissolve the whole council and force a new election.

Most telling, the council's mandate was written in van Rossum's image: to seek consensus, to act as a court of final appeal, not a throne, and to use its formal power as little as possible. The Republic would be governed the way the King had always wanted to govern. This time, though, five people would share the weight.

When the first council was elected in early 2019, one of the five names on it was Guido van Rossum. The core developers wanted him there, as a citizen and not a king, and he served one term. Then he stepped back again and left his seat to others.

The real test came with PEP 634: Structural Pattern Matching, a whole new match/case construct for Python 3.10 that took apart complex data structures in a single stroke. It was a far bigger change than the walrus, with far more ways to go wrong.

Again opinion split. Again the threads ran hot. Again people said the language was being ruined.

This time the weight didn't fall on one man. The Steering Council took the pressure and in early 2021 approved the proposal, acknowledging the reservations openly. Nobody burned out. Nobody walked away. The language had survived its creator's leaving, and it was stronger for it.

XI. Retirement Is Boring

Plaza on the Microsoft Redmond campus The Microsoft campus in Redmond, where the retired King came back to work as an engineer.

In October 2019, van Rossum left Dropbox and retired. He was sixty-three. He had seen the Python 3 migration through and laid down the crown. He had earned a quiet life.

It lasted about a year.

Retirement, it turned out, was boring. Then came the pandemic and the empty days of 2020, with nothing to push against. He missed the code. He missed having a team. He missed the satisfaction of a hard technical problem.

In November 2020 he stunned the tech world by coming out of retirement to join Microsoft, in its Developer Division, as a Distinguished Engineer.

Microsoft! The irony was hard to miss. For decades Microsoft had been the great dark empire of the open-source imagination, the company whose leadership had once called Linux a "cancer." Now the creator of one of the most important open-source languages ever was drawing a salary in Redmond. But the Microsoft of Satya Nadella wasn't the Microsoft of the 1990s. It had become, improbably, an open-source powerhouse. And it offered van Rossum something he hadn't really had since that Christmas in 1989: the freedom to just be a programmer.

He put together a small team with the core developer Mark Shannon and went after the oldest complaint about Python: it's slow. No mailing-list wars, no final say, just a hard problem. Their Faster CPython project taught the interpreter to rewrite its own hot code paths as it ran, and Python 3.11, released in 2022, ran about 25 percent faster on average. For a man who had once absorbed a whole community's fury, it was the work he had always wanted: a benchmark, not a referendum.

Corporate patronage cuts both ways, though. In May 2025, in company-wide layoffs, Microsoft ended its support for Faster CPython and let go of Mark Shannon along with fellow core developers Eric Snow and Irit Katriel. The work went on in the open-source community. A year later, in May 2026, van Rossum retired from Microsoft. This time, the language didn't need him to stay.

XII. The Lock

There was one dragon left, the oldest and biggest: the Global Interpreter Lock, the GIL.

Since its earliest days, CPython had used a single lock that let only one thread run Python code at a time. It kept things simple, and it meant Python couldn't really use the multi-core processors that had become universal. It was a fossil from the young man's cold winter, still lodged in the heart of the machine.

In 2023, the Steering Council took up PEP 703, a proposal by the Meta engineer Sam Gross to make the GIL optional. It was open-heart surgery on the language, with the risk of breaking decades of extensions the whole ecosystem depended on, a far more dangerous change than the walrus.

Think about what that proposal would have done to the old order. If it had landed on the BDFL's desk in 2018, alongside the walrus and on top of everything else, the pressure of managing the fallout might well have broken him outright.

In 2023 there was no single desk to land on. The council accepted PEP 703 as a body and laid out a careful, multi-year road. An experimental free-threaded build appeared in Python 3.13. In Python 3.14, in 2025, it became an officially supported option.

The GIL work belonged to others, to Gross and the council and a long list of contributors, and that was exactly the point. Van Rossum spent those years as an engineer, not a besieged monarch, tuning the interpreter and arguing over implementation details as an equal, while a decision bigger than the walrus was made without him.

He had found his way back to the room in Amsterdam.

XIII. The Liberation of Limits

In the end, this is a story about the limits of a single human being.

We love the lone genius in technology. We build myths around founders who started in a dorm room or a garage or over a bored Christmas holiday, and we assume the spark of creation comes with the armor needed to rule an empire. We expect the founder to be in command, unbothered, turning any amount of pressure into more power.

But software doesn't have a nervous system. People do.

Python scaled without limit, copying itself across servers and laptops and research clusters until it was the foundation of the AI age. Its creator couldn't scale. There was one of him, with one inbox, one finite supply of patience, and one body that wasn't getting younger. He had built a language on human readability, and it had grown a community that was, in the end, emotionally unreadable.

By walking away, Guido van Rossum won a rare and quiet victory. He escaped the founder's trap, the one that catches creators who can't let go and end up strangling the thing they made, or being strangled by it. He forced his community to grow into a self-governing republic that could carry out pattern matching, a faster interpreter and the slow dismantling of the GIL without tearing itself apart. He made sure Python would outlive him.

The young man in the dark Amsterdam winter of 1989 wanted to build something readable that would let people talk to machines, and to each other, more clearly. He did that. Then it grew into a monster, and the monster nearly ate him. When the fury over a walrus pushed him to the edge, he walked away from his crown, and in doing so he saved his kingdom. More importantly, he saved himself.

Guido van Rossum was tired. Then, at last, he rested. And then, because he'd always been a programmer before he was a king, he went back to writing code, until in 2026, at seventy, he retired once more. This time he went on his own terms.

Author's note: This profile is narrative nonfiction. Interior moments and scene-setting are imagined, drawn from the public record. Quotations from the "Transfer of power" email are from van Rossum's message to the python-committers list on July 12, 2018.