When I ran my first worm simulation, the Santa Ana winds were already blowing hard. I’m not a hacker, but it was simple: you open a terminal shell, paste in commands from GitHub, and watch a cascade of symbols scroll across the screen—just like in the movies. As I scanned the code flying past my eyes for familiar words—neuron, synapse—a friend came by to take me out to dinner. “One second. I’m running a worm on my computer,” I shouted to him from the office.
The Korean restaurant was chaotic: the wind bent palm trees and sent shopping carts skidding across the parking lot. The atmosphere felt surreal, like a podcast playing at double speed. “What are you doing, cybercrime?” my friend asked. I tried to explain: “No, not a worm like Stuxnet. A real, living worm.”
By the time I got home, it was already dark, and the first sparks of wildfires were appearing in Altadena. On my laptop, a worm was waiting for me inside a volumetric pixel box. Pointed at both ends, it hung in a cloud of particles, strangely straight and motionless. Of course, it had never been alive. But to me, it felt deader than dead. “Bravo,” Stephen Larson told me when I wrote to him that evening. “You’ve reached the ‘hello world’ stage of the simulation.”
Larson is one of the founders of OpenWorm, an open-source project that has been trying since 2011 to build a computer simulation of the microscopic nematode Caenorhabditis elegans. Its goal is nothing less than a complete digital twin of a real worm, accurate down to the molecule. If OpenWorm succeeds, it will be the first virtual animal—and a representation of all our knowledge not only of C. elegans (one of the most well-studied animals in science), but also of how the brain interacts with the world to produce behavior. Participants in OpenWorm call this the “Holy Grail” of systems biology.
Unfortunately, they haven’t achieved that yet. The simulation on my laptop pulls data from experiments on real worms and translates it into the computational framework c302, which in turn drives the simulated muscles of C. elegans in a dynamic fluid environment—essentially simulating how the worm moves by wriggling through a thin layer of slime. Generating five seconds of this behavior takes about ten hours of computation.
A lot can happen in ten hours. Ash carried by the wind can drift from the foothills into a sleeping city. That night, following Larson’s advice, I adjusted the simulation’s time parameters to move one step beyond “hello world,” deeper into the uncanny valley of worm life. The next morning, I woke up to a strange orange haze. Barely opening my eyes, I checked my laptop—and my heart skipped a beat at two pieces of news: Los Angeles was burning. And my worm had started moving.
At this point, you might ask a perfectly reasonable question. My friend asked it at the restaurant. The question was: “Uh… why?” Out of all the diversity of our green world, and all the problems worth solving, why would anyone spend thirteen years trying to program the life of a microscopic worm?
In answering, I quoted one of physicist Richard Feynman’s most famous sayings: “What I cannot create, I do not understand.”
For most of its history, biology has been a reductionist science, guided by a simple principle: the best way to understand the overwhelming complexity of living organisms is to break them down into their component parts—organs, cells, proteins, molecules. But life is not a clockwork mechanism; it is a dynamic system, and unexpected properties emerge from the interactions of its parts. To truly understand life, it’s not enough to take it apart—you also have to be able to put it back together.
The nematode C. elegans is a tiny worm, about the thickness of a human hair, made up of fewer than a thousand cells. Only 302 of them are neurons—almost the minimum required for a brain. “I remember that after my first child was born, I was very proud when he finally learned to count to 302,” says neurobiologist Netta Cohen of the University of Leeds worm lab. But there’s nothing embarrassing about its small size, Cohen argues: despite having so little, C. elegans can do a lot. Unlike its parasitic relatives, it is a free-living organism. “It can reproduce, eat, forage, escape. It is born and develops, ages and dies—all within the scale of a millimeter,” she explains.
Worm enthusiasts will quickly tell you that research on C. elegans has led to at least four Nobel Prizes; it was the first animal to have its genome sequenced and its neural connections mapped. But there’s a difference between a wiring diagram and an instruction manual. “We know how everything is connected, but we don’t know the dynamics. You might think this is a perfect problem for a physicist, a computer scientist, or a mathematician,” Cohen says.
And they’ve tried to solve it. The first simulation of C. elegans was created by Sydney Brenner, who turned a humble compost-heap worm into a scientific superstar in his landmark 1986 paper, The Structure of the Nervous System of the Nematode Caenorhabditis elegans, reverently known among researchers as “The Worm Brain.” Brenner’s team spent thirteen years in a Cambridge lab, painstakingly slicing worms and photographing them under an electron microscope. Using a first-generation minicomputer with punch cards, they turned their data into a rudimentary map of the worm’s nervous system.
Since then, every decade or two, computer scientists have tried to build on and expand Brenner’s work. But biology has a way of humbling them. In 2003, computer scientist David Harel called the simulation of C. elegans a “grand challenge” of biology, overdue for a crucial shift “from analysis to synthesis.” While he was right, he managed to model little more than the worm’s vulva.
Netta Cohen has spent much of her twenty-year career publishing breakthrough computational models that capture the worm’s sinusoidal motion as it moves through environments of varying viscosity. But backward movement remains a completely separate, unsolved problem. And don’t even ask how the worm moves up and down—let alone why it does so. All our behavioral data comes from observing worms in flat Petri dishes with agar. In nature, they may behave very differently. “Why not?” Cohen laughs. “It’s biology.”
When OpenWorm announced its ambitions in 2011, Stephen Larson, an engineer who had “found religion” in open source, believed that assembling a group of passionate computational researchers would lead to breakthroughs in biology and simulation. Thirteen years later, enthusiasm has given way to humility. “This project might be like a cathedral. If I can’t finish it, at least others can continue building it.”
Perhaps it’s burnout: leading a nonprofit open-source project can exhaust even the most dedicated idealist. Or perhaps it’s the deceptive simplicity of the C. elegans brain, still resistant to analysis. Or maybe it’s simply a matter of timing.
OpenWorm does not conduct its own experiments. Its volunteers extract information from the C. elegans literature, integrating whatever data they can find into the simulation. This makes them dependent on research labs like Cohen’s, which are often slow to produce data that is truly useful for computational models. But over the past decade, experimentalists have improved microscopes and genetic techniques, generating richer and higher-quality recordings of the worm’s brain. At the same time, machine learning tools have emerged to process this data, and computing power continues to grow rapidly. Because of this convergence, Larson remains hopeful. “When you live in a time of near-exponential technological growth, things that once seemed crazy become possible,” he says.
I asked Netta Cohen, who serves on OpenWorm’s scientific advisory board, whether this is truly feasible. “We should start by asking: what is ‘this’? What are we trying to achieve?” Cohen is one of 37 co-authors of a recent paper proposing a new roadmap: using genetic imaging techniques to activate each of the worm’s 302 neurons one by one and measure their effects on all the others. Repeating these experiments hundreds of thousands of times in parallel could generate enough data to give computer scientists something to work with—information that might even allow a full reverse engineering of the worm.
It’s an extraordinarily ambitious proposal, requiring unprecedented collaboration among about twenty worm research labs. Neurobiologist Gal Haspel of the New Jersey Institute of Technology, the paper’s lead author, estimates it could take up to ten years, tens of millions of dollars, and 100,000–200,000 living worms. The project would generate more data on C. elegans than has been collected in all of scientific history. And what would reverse engineers ultimately get? “A lot of people and computers will recreate only what this tiny animal can do,” Haspel says.
But he’s being a bit ironic: he compares the project to NASA’s moon program. Such an effort would push technology forward, forcing engineers to build better tools and scientists to collaborate more closely. Haspel sees worm simulation as a pathway to a new kind of science driven by automation, big data, and machine learning. And although the end product might just be a worm—in some sense, the most complex Tamagotchi ever—it could be a stepping stone toward understanding more complex nervous systems, and eventually, the human brain.
Last summer, a crypto developer posted an animated GIF of a virtual C. elegans wriggling in an app window on X. The animation was generated using the same code I ran on my laptop, freely available on OpenWorm’s GitHub. “If a worm matrix can run on my M1 Mac, what are the odds that we’re in base reality?” he wrote, suggesting that perhaps we are the worms—running on some cosmic MacBook in a higher layer of reality. The post went viral; of course, Elon Musk liked it.
When I told Patrick Gleeson, OpenWorm’s director and a neurobiologist at University College London, about the “worm matrix,” he winced. “Some people join the project because they’re drawn to philosophical speculation. And that’s fine. But my priorities are, first and foremost, biological.”
If Larson is Steve Jobs, Gleeson is Wozniak: he’s less interested in building a super-worm than in creating a platform that integrates many small, separate models of C. elegans biology. Computational modeling is common in biology; it’s a low-cost way to encode and test theories as “thought experiments” before moving to Petri dishes. Typically, models focus on small aspects of an organism—say, a handful of neurons controlling movement. When it comes to modeling, “we don’t need the map to equal the territory—that would get in the way of our goals,” explains Eduardo Izquierdo, a worm modeler at the Rose-Hulman Institute of Technology. “We’re looking for something that helps with careful analysis.”
No one mistakes a biological model for a living organism. But a full-scale simulation opens up an entirely new class of questions. Using Izquierdo’s definition, the map would be as accurate as the territory—and that invites new ways of thinking about the nature of that territory, not to mention life itself. A model helps scientists answer questions; a simulation creates them. For example: what is the difference between a virtual worm and its living counterpart, if they are identical down to the molecule?
Larson believes that building a fully accurate simulation would not overturn our understanding of life, but expand it. “If we say a living organism is only a system of physical molecules existing on a planet, then a computer model without physical molecules cannot be alive. But if we expand the definition of life to include information, then perhaps ‘life’ could apply to a simulated organism. Then the question becomes: does it even matter?”
I think it does. Life is information—but also something more, something we feel most strongly when it’s gone. In that sense, Feynman’s statement might need revision: understanding doesn’t come from creation alone. Only by trying to recreate life do we begin to understand how irreplaceable it is.
I am seeing this firsthand now, surrounded by destruction. The air is poisoned; flakes of white ash have seeped into every corner of my home. Living on the edge of an evacuation zone, smelling smoke, I distract myself by running more proto-simulations of C. elegans. Watching them, I can’t help but marvel at how easy it is to destroy life—and how hard it is to create it. It takes a single spark to burn down centuries of living systems overnight; but to come close to creating a virtual worm takes decades—and perhaps that work will never be finished.
The Korean restaurant was chaotic: the wind bent palm trees and sent shopping carts skidding across the parking lot. The atmosphere felt surreal, like a podcast playing at double speed. “What are you doing, cybercrime?” my friend asked. I tried to explain: “No, not a worm like Stuxnet. A real, living worm.”
By the time I got home, it was already dark, and the first sparks of wildfires were appearing in Altadena. On my laptop, a worm was waiting for me inside a volumetric pixel box. Pointed at both ends, it hung in a cloud of particles, strangely straight and motionless. Of course, it had never been alive. But to me, it felt deader than dead. “Bravo,” Stephen Larson told me when I wrote to him that evening. “You’ve reached the ‘hello world’ stage of the simulation.”
Larson is one of the founders of OpenWorm, an open-source project that has been trying since 2011 to build a computer simulation of the microscopic nematode Caenorhabditis elegans. Its goal is nothing less than a complete digital twin of a real worm, accurate down to the molecule. If OpenWorm succeeds, it will be the first virtual animal—and a representation of all our knowledge not only of C. elegans (one of the most well-studied animals in science), but also of how the brain interacts with the world to produce behavior. Participants in OpenWorm call this the “Holy Grail” of systems biology.
Unfortunately, they haven’t achieved that yet. The simulation on my laptop pulls data from experiments on real worms and translates it into the computational framework c302, which in turn drives the simulated muscles of C. elegans in a dynamic fluid environment—essentially simulating how the worm moves by wriggling through a thin layer of slime. Generating five seconds of this behavior takes about ten hours of computation.
A lot can happen in ten hours. Ash carried by the wind can drift from the foothills into a sleeping city. That night, following Larson’s advice, I adjusted the simulation’s time parameters to move one step beyond “hello world,” deeper into the uncanny valley of worm life. The next morning, I woke up to a strange orange haze. Barely opening my eyes, I checked my laptop—and my heart skipped a beat at two pieces of news: Los Angeles was burning. And my worm had started moving.
At this point, you might ask a perfectly reasonable question. My friend asked it at the restaurant. The question was: “Uh… why?” Out of all the diversity of our green world, and all the problems worth solving, why would anyone spend thirteen years trying to program the life of a microscopic worm?
In answering, I quoted one of physicist Richard Feynman’s most famous sayings: “What I cannot create, I do not understand.”
For most of its history, biology has been a reductionist science, guided by a simple principle: the best way to understand the overwhelming complexity of living organisms is to break them down into their component parts—organs, cells, proteins, molecules. But life is not a clockwork mechanism; it is a dynamic system, and unexpected properties emerge from the interactions of its parts. To truly understand life, it’s not enough to take it apart—you also have to be able to put it back together.
The nematode C. elegans is a tiny worm, about the thickness of a human hair, made up of fewer than a thousand cells. Only 302 of them are neurons—almost the minimum required for a brain. “I remember that after my first child was born, I was very proud when he finally learned to count to 302,” says neurobiologist Netta Cohen of the University of Leeds worm lab. But there’s nothing embarrassing about its small size, Cohen argues: despite having so little, C. elegans can do a lot. Unlike its parasitic relatives, it is a free-living organism. “It can reproduce, eat, forage, escape. It is born and develops, ages and dies—all within the scale of a millimeter,” she explains.
Worm enthusiasts will quickly tell you that research on C. elegans has led to at least four Nobel Prizes; it was the first animal to have its genome sequenced and its neural connections mapped. But there’s a difference between a wiring diagram and an instruction manual. “We know how everything is connected, but we don’t know the dynamics. You might think this is a perfect problem for a physicist, a computer scientist, or a mathematician,” Cohen says.
And they’ve tried to solve it. The first simulation of C. elegans was created by Sydney Brenner, who turned a humble compost-heap worm into a scientific superstar in his landmark 1986 paper, The Structure of the Nervous System of the Nematode Caenorhabditis elegans, reverently known among researchers as “The Worm Brain.” Brenner’s team spent thirteen years in a Cambridge lab, painstakingly slicing worms and photographing them under an electron microscope. Using a first-generation minicomputer with punch cards, they turned their data into a rudimentary map of the worm’s nervous system.
Since then, every decade or two, computer scientists have tried to build on and expand Brenner’s work. But biology has a way of humbling them. In 2003, computer scientist David Harel called the simulation of C. elegans a “grand challenge” of biology, overdue for a crucial shift “from analysis to synthesis.” While he was right, he managed to model little more than the worm’s vulva.
Netta Cohen has spent much of her twenty-year career publishing breakthrough computational models that capture the worm’s sinusoidal motion as it moves through environments of varying viscosity. But backward movement remains a completely separate, unsolved problem. And don’t even ask how the worm moves up and down—let alone why it does so. All our behavioral data comes from observing worms in flat Petri dishes with agar. In nature, they may behave very differently. “Why not?” Cohen laughs. “It’s biology.”
When OpenWorm announced its ambitions in 2011, Stephen Larson, an engineer who had “found religion” in open source, believed that assembling a group of passionate computational researchers would lead to breakthroughs in biology and simulation. Thirteen years later, enthusiasm has given way to humility. “This project might be like a cathedral. If I can’t finish it, at least others can continue building it.”
Perhaps it’s burnout: leading a nonprofit open-source project can exhaust even the most dedicated idealist. Or perhaps it’s the deceptive simplicity of the C. elegans brain, still resistant to analysis. Or maybe it’s simply a matter of timing.
OpenWorm does not conduct its own experiments. Its volunteers extract information from the C. elegans literature, integrating whatever data they can find into the simulation. This makes them dependent on research labs like Cohen’s, which are often slow to produce data that is truly useful for computational models. But over the past decade, experimentalists have improved microscopes and genetic techniques, generating richer and higher-quality recordings of the worm’s brain. At the same time, machine learning tools have emerged to process this data, and computing power continues to grow rapidly. Because of this convergence, Larson remains hopeful. “When you live in a time of near-exponential technological growth, things that once seemed crazy become possible,” he says.
I asked Netta Cohen, who serves on OpenWorm’s scientific advisory board, whether this is truly feasible. “We should start by asking: what is ‘this’? What are we trying to achieve?” Cohen is one of 37 co-authors of a recent paper proposing a new roadmap: using genetic imaging techniques to activate each of the worm’s 302 neurons one by one and measure their effects on all the others. Repeating these experiments hundreds of thousands of times in parallel could generate enough data to give computer scientists something to work with—information that might even allow a full reverse engineering of the worm.
It’s an extraordinarily ambitious proposal, requiring unprecedented collaboration among about twenty worm research labs. Neurobiologist Gal Haspel of the New Jersey Institute of Technology, the paper’s lead author, estimates it could take up to ten years, tens of millions of dollars, and 100,000–200,000 living worms. The project would generate more data on C. elegans than has been collected in all of scientific history. And what would reverse engineers ultimately get? “A lot of people and computers will recreate only what this tiny animal can do,” Haspel says.
But he’s being a bit ironic: he compares the project to NASA’s moon program. Such an effort would push technology forward, forcing engineers to build better tools and scientists to collaborate more closely. Haspel sees worm simulation as a pathway to a new kind of science driven by automation, big data, and machine learning. And although the end product might just be a worm—in some sense, the most complex Tamagotchi ever—it could be a stepping stone toward understanding more complex nervous systems, and eventually, the human brain.
Last summer, a crypto developer posted an animated GIF of a virtual C. elegans wriggling in an app window on X. The animation was generated using the same code I ran on my laptop, freely available on OpenWorm’s GitHub. “If a worm matrix can run on my M1 Mac, what are the odds that we’re in base reality?” he wrote, suggesting that perhaps we are the worms—running on some cosmic MacBook in a higher layer of reality. The post went viral; of course, Elon Musk liked it.
When I told Patrick Gleeson, OpenWorm’s director and a neurobiologist at University College London, about the “worm matrix,” he winced. “Some people join the project because they’re drawn to philosophical speculation. And that’s fine. But my priorities are, first and foremost, biological.”
If Larson is Steve Jobs, Gleeson is Wozniak: he’s less interested in building a super-worm than in creating a platform that integrates many small, separate models of C. elegans biology. Computational modeling is common in biology; it’s a low-cost way to encode and test theories as “thought experiments” before moving to Petri dishes. Typically, models focus on small aspects of an organism—say, a handful of neurons controlling movement. When it comes to modeling, “we don’t need the map to equal the territory—that would get in the way of our goals,” explains Eduardo Izquierdo, a worm modeler at the Rose-Hulman Institute of Technology. “We’re looking for something that helps with careful analysis.”
No one mistakes a biological model for a living organism. But a full-scale simulation opens up an entirely new class of questions. Using Izquierdo’s definition, the map would be as accurate as the territory—and that invites new ways of thinking about the nature of that territory, not to mention life itself. A model helps scientists answer questions; a simulation creates them. For example: what is the difference between a virtual worm and its living counterpart, if they are identical down to the molecule?
Larson believes that building a fully accurate simulation would not overturn our understanding of life, but expand it. “If we say a living organism is only a system of physical molecules existing on a planet, then a computer model without physical molecules cannot be alive. But if we expand the definition of life to include information, then perhaps ‘life’ could apply to a simulated organism. Then the question becomes: does it even matter?”
I think it does. Life is information—but also something more, something we feel most strongly when it’s gone. In that sense, Feynman’s statement might need revision: understanding doesn’t come from creation alone. Only by trying to recreate life do we begin to understand how irreplaceable it is.
I am seeing this firsthand now, surrounded by destruction. The air is poisoned; flakes of white ash have seeped into every corner of my home. Living on the edge of an evacuation zone, smelling smoke, I distract myself by running more proto-simulations of C. elegans. Watching them, I can’t help but marvel at how easy it is to destroy life—and how hard it is to create it. It takes a single spark to burn down centuries of living systems overnight; but to come close to creating a virtual worm takes decades—and perhaps that work will never be finished.
