Life is the universe developing a memory

Lee Cronin starts where a chemist has to start: a hot young planet, a short list of minerals, some carbon, some water. Earth formed roughly 4.6 billion years ago. Lex Fridman notes the timeline that still startles him — life has been here for something like 70 or 80 percent of that history, and the long wait was not for a first cell but for anything more complicated than bacteria. Cronin does not claim to know the recipe. He does claim the speed is the clue.

If life happens that quickly, he says, it is much more inevitable than origin-of-life lore usually allows. Frame the emergence of life on a planet as common as the formation of a star, and the universe looks different. Chemists sit at the center of the problem because we are made of chemistry, yet Cronin thinks selection had to run *before* biology was established. Selection built the framework; complexity, capability, and autonomy grew on top of it.

He does not want a definition that stops at a cell, a virus, or an on-off switch. "Life is the universe developing a memory." Physics, in this telling, is memory-free stuff: there are only so many ways to arrange fundamental particles. Life is stickiness. It is a coherent story you can trace backward through complexity. Selection is not one of the four fundamental forces, but Cronin is convinced it is a directing force in the universe, and that biology is a selection amplifier. Humanity and whatever autonomous machines come next will supercharge it again.

Chemistry, for Cronin, is what happens when electrons come together and form bonds. Talking about life elsewhere, he offers a line he likes to repeat: where there are bonds, there is hope. Bonds give you heterogeneity. They let you record memories. Atoms combine, those combinations take on identities, identities interact, and then you get causation, selection, and existence. Bonds come first. Life, intelligence, and consciousness can, in principle, arrive in any order after that — a Solaris-like plasma ocean, a neutron-star oscillator, something no human would count as alive.

The elegant path on Earth is still bonds, then selection, then life, then multicellularity and state machines. What looks like magic is often just a lowered energy barrier. In a soup of possible reactions, most never happen; a mineral or a bit of clay can sit between two molecules and promote one path over the others. Each successful step makes a molecule that can redirect the next. Cronin reaches for the smallest von Neumann constructor: not a full self-replicating machine in one leap, but the simplest molecule that helps build a more complicated one, until something in that chain can act back on the process that made it.

He thinks copying yourself is much easier than the mythology around the first replicator suggests. Replicating molecules are rare, but a glass of water holds on the order of a mole of molecules — roughly as many as there are stars — and most possibilities get smashed back into atoms. You only need one or two good copiers to take over a pool. In his own lab, molybdenum oxide plus acid and electrons grew gigantic wheels of 154 molybdenum atoms and dodecahedra of 132, all in the same pot. A random 12-atom species templated itself and, by accident, the ring; the ring helped make the 12-mer. That is an autocatalytic set: A makes B, B helps make A.

The structures look impossible in one step. They are not one step. Cronin reaches for the blind-watchmaker argument: a watch does not appear spontaneously; it is a lineage of cruder devices bootstrapped onto one another. Earth's ribosome is an incredibly complex object, a phase transition in replication, but it is Earth's answer, not a universal one. The molecules we have now are contingent on this planet's history. There are many solutions. That is the trap he thinks origin-of-life chemistry has fallen into.

Cronin had tweeted that origin-of-life research is a scam. He tells Fridman it was tongue-in-cheek, aimed at a method, not at the chemists. Plenty of excellent people are doing the work. The nudge is this: the field keeps assuming that if only we could make this RNA, this phosphodiester, this ATP, the origin would unlock. He likens it to a country that has never seen a car receiving its first Tesla, taking it apart, and declaring that doors, wheels, and steering columns are how cars emerged — while the real causal chain runs back through Henry Ford, the horse and carriage, and the wheel.

Obsession with the identities in current biology, he says, gives a false sense of security. Craig Venter's synthetic genome was a milestone, but Venter still had to boot it in an existing cell whose lineage goes back to LUCA. Synthetic biologists still cannot make a cell from scratch, because contingent information lives outside the genome. Cronin's own standard is sand to cells: grind rocks, add energy — heat, UV, lightning — cycle a closed world, and wait for a life form to crawl out of the test tube. He would bet that the genetic material would not be RNA, the proteins would not be ours, and the thing would be feeble. If it looked like Earth life, that would be a contamination warning, not a triumph.

The containment argument is blunt. Use chemistry that is not Earth-abundant — molybdenum, for example — and an escapee dies immediately because the environment cannot feed it. That is the opposite of gain-of-function work that walks a pathogen closer to humans. The terminator is not coming from a first lab-born cell any more than Boston Dynamics is about to birth Skynet; the nearer dangers are dumber and already here, from viral social networks to sloppy virology. Cronin still wants a new life form in his lifetime, to prove life is a general phenomenon, like gravity. Fridman adds that creating it may be cheaper, easier, and more conclusive than finding it.

Mars complicates the search. Mars was habitable for this kind of chemistry before Earth was, so Cronin thinks we might all be Martians: replicators ejected from Mars, landing here, skipping Earth's search. Similar life on Mars would be exciting and ambiguous. Completely different chemistry on Titan, Enceladus, or Europa would be the real story. He wants to classify planets the way we classify stars — dead, prebiotic, living, technological, perhaps a fifth for worlds that lived and died — instead of treating every methane or oxygen wiggle as a biomarker.

Sipping wine under the stars, Cronin has no census of civilizations. He also sees no reason to think life is as hard as we first supposed. If life is more general than our biology, the universe is full of it. The Fermi paradox, in that frame, is not empty sky. It is a translation problem. Our technologies may be so divergent, and universal facts so few, that we would not recognize the other party — not necessarily the aliens of *Arrival*, just something we are not equipped to see.

He will say it with confidence: biology as we find it on Earth does not exist anywhere else in the universe. The origin was contingent on the chemicals at hand. Reset Earth, and you get broad similarities, not elephants every time. Another rocky world in the habitable zone might look biological and still not be the same. Counting the bits in an elephant, and the causal chain back to LUCA, leaves huge scope for divergence. Looking for water, oxygen, methane, and amino acids is looking for our exam answers, not for selection.

Why no phone call, then, if someone is ahead of us? "It takes two to talk." The aliens might be there; if we cannot recognize them, they have little reason to chat. You would not travel twenty light-years to discuss calculus with a cat, even if you would still pet it. Cronin is not saying we are uninteresting. He is saying we have not yet qualified, because we do not understand causal chains, replication, or how the universe spat out life. We qualify as soon as we can decode the signal. JWST, he hopes, will start classifying exoplanets by color. The deeper fix is to make non-biological life on Earth, so we know what other solutions look like.

Fridman asks about UFO sightings, including the pilot reports. Cronin grants the chance — any decent scientist has to — and still thinks it unlikely, not impossible, that aliens are visiting now. The promised dumps of classified material disappointed him: poor footage, hearsay, instrument glitches. Everyone carries a high-resolution camera, and the pictures are still grainy. Fridman pushes back that incompetence and bad leadership mean a pocket smartphone is not enough to catch rare events. Cronin restates the claim more carefully: he has not seen enough replicated data to take the files seriously.

What did interest him was the pandemic. Locked down, people let their imaginations run, and humanity opened its mind to aliens and the mystery of the universe. He will not call those thoughts stupid. The yearning is real, and it drives him harder toward artificial life and toward helping NASA look. Anomalies matter in chemistry too: phlogiston, the ether, cold fusion. Some collapsed; some, he thinks, still hide real electrochemical phenomena we lack the design to catch. The challenge is not to dismiss an anomaly because you cannot yet replicate it.

That stance comes with a biography. Cronin was put in the learning-difficulties class and told he was stupid, so when academia said he was wrong he told them to join the queue. He pictures science as a Japanese house of paper walls. He walks through them — chemist into mathematics, into computer science — then has to win the people in the next room with their own standards. Two superpowers, he says: a useful stupidity that lets him start again, and a curiosity bigger than his ego. A hypothesis he cannot test with an experiment or a thought experiment is, in the end, without merit.

For nearly a decade at origin-of-life and artificial-life meetings, Cronin felt everyone was dancing around what life *is*. Assembly theory is his hammer. Take an object, tap it gently, and count the unsymmetrical fragments. The more distinct parts, especially if the object is not a one-off but abundant, the more likely it was built by an evolutionary or informational process. A unique intricate thing can still be a fluke. Ten thousand copies of the same intricate thing, the number you typically need for a mass-spectrometer signal, is not a fluke.

The assembly index is the minimum number of steps to rebuild the object if you may reuse parts you have already made — the memory of the universe, written as a causal chain. The word *abracadabra* takes seven steps that way, fewer than its letter count. Cronin and a student ran the algorithm on about 20 million molecules from a database, then checked it in the lab. Mass spectrometry, infrared, and NMR tracked the index, almost one-to-one once bond energies were allowed for. Asymmetry is the point: the less symmetry, the more information, the weirder the object.

Then they sniffed the world: seawater, Antarctica, fossils, outer space, plus NASA-blinded samples. Biology produced molecules above a threshold of about 15. Abiotic material stayed low. On Earth, at least, natural processes do not seem to make molecules that need more than 15 steps. That is a scale of life, and of technology — some terrestrial molecules, he says, are beyond even biology and required microprocessors. NASA's Dragonfly, a nuclear-powered drone with a mass spectrometer, is due to fly Titan. Cronin wants that data read for assembly. You do not need cells, water, or RNA. You need a scale and a way to count fragments.

The same logic applies to an iPhone or a 747 if you find it in abundance. Assembly is a function of an object's complexity and its copy number. Compared with Shannon entropy or Kolmogorov complexity, the pitch is measurability: complexity has usually needed a human to label things or a program to compare them. Assembly is close to logical depth, Cronin tells information theorists, but you can do it in an instrument. It gets NP-hard for huge objects; for molecules of modest weight it is good enough to tell chance from causation.

Look around a table and the missing quantity becomes obvious. Two mugs, two water bottles, copies everywhere. In a combinatorial explosion the universe should be an unrepeatable mess; it is not, because once you make a thing you can use it again. Reusability is memory. Cronin wants three quantities on the books: energy, entropy, and assembly. Entropy tells you what you have lost. Assembly, built from the bottom up, keeps the causal chain. The complexity of the cup in his hand is not only in the object. It is in the lineage of people making cups, the abstraction of a cup in someone's head, and the next slightly different cup.

That is also a claim about causation, which many physicists treat as emergent and absent at the bottom. Cronin thinks that cannot be right, because life needs molecules to bump, replicators to make polymers, cause and effect at the molecular scale. If causation only appears in the macro world, life never starts. Leibniz, he suggests, was groping toward the same idea when he gave objects a "soul": not mysticism, but the conserved history without which this water bottle would not exist. You are looking at every water bottle humanity has conceived.

He lists three things in the physicist's belief system that make him grumpy: a magical low-entropy order at the beginning, a second law that has to be added by hand, and emergent time. If time is fundamental — a global clock you cannot step out of — the asymmetry is built in, the second law falls out of more states becoming available, and you do not need the past hypothesis. "Luckily I am the Regius chair," he jokes; it would take an act of Parliament to fire him. He is happy to be wrong. He just wants fewer assumptions.

The payoff is a universe that gets more novel at every step. It was not capable of consciousness on day one, because it did not have enough states. Today it does. Dark energy, he wonders aloud, might just be time; he has no evidence, only a direction. Assembly theory still works if time is emergent, he admits, but then he loses a satisfying account of how causation and these objects move through space. Sean Carroll can keep emergent time if he likes. Cronin would rather walk through that wall and see what happens.

Cronin once tweeted that assembly theory explains why Nietzsche understood limited freedom rather than radical freedom. Sartre treated the imagination as unbounded. Nietzsche saw the constraints. The childhood illustration is *Dragon's Lair*, the laserdisc arcade game: gorgeous animation, no generation, just a stick flicked at the right instant to play the next clip. You can finish in twelve minutes if you never miss. That is not interaction. It is a script. Assembly theory says the same of a life. A poppy can invent alkaloids; it cannot invent an OLED molecule. The causal chain that produced you limits the future, and also lets you do amazing things inside those limits.

"If everywhere you are in cages, if you can move around in that cage a little bit, you're free." He will not follow Sam Harris into thinking himself out of existence. Will, for Cronin, is a set of nearly equally weighted decisions, not energy minimization, shaped by the model in your head, other people, and happenstance. Some days he pretends he has no free will at all. The trick makes him more radical: if the choice is already made, he can ask the question he dared not ask and do the thing he dared not do, and the accessible state space gets larger. Decision-making, he suspects, is yet another face of selection — counterfactuals held in working memory, in an intrinsically open universe.

Fridman and Cronin had been arguing offline about whether cellular automata exist outside the human mind. Cronin loves them; he also thinks people marvel at the wrong layer. Before that fight he tells a lab story. He asked his group for two chemical robots that would play chess with dye chemistry; they said no. They built Hex instead, John Nash's game, communicating moves on Twitter. Each robot ran reactions from a shared list. The bluer the dye, the better its chance to place a hex. Cleaning the pipes was optional. Both robots learned not to clean, because dirty pipes went blue faster. Twitter banned the accounts. Unintended consequences of a fixed objective, long before anyone said reward hacking.

Stephen Wolfram, whose work Cronin admires, told him assembly theory was nonsense. Cronin's counter is about boundary conditions. Run Conway's Game of Life on a computer and you have already poured in operations, error correction, a display, a von Neumann machine sitting on a Turing machine. "A grid is not simple." It takes extraordinary machinery to bring a grid to life. Von Neumann and Feynman, drawing automata on paper, never saw the elaborate structures TikTok can loop for hours. The richness is partly the operating system underneath.

So his group built a physical Game of Life: 49 stirred cells of Belousov–Zhabotinsky reaction, each oscillating red to blue, little chemical bits coupled as an analog computer. Structures appeared that the silicon CA would not have produced. The oscillator leaves a wake; wakes interfere; the leftover noise of one round trails into the next. Silicon states are clean. Chemistry is not. The paper, he says, sat in purgatory for a long time. The same box can be shaken as a Monte Carlo solver for problems like the traveling salesman. He wonders, without wanting to quote a size, whether a large enough chemical grid could reach a kind of chemical supremacy on optimization.

Do cellular automata exist in physical reality? Probably not, until minds construct them — on paper, in memory, in a lab. Physics, chemistry, and biology are the same discipline at different depths of memory. Physics is low-memory stuff, so patterns recur and look like laws. Chemistry is more contingent. Biology more so. Technology more so still. The less memory you need, the more universal the regularity. That is all he wants to take from the CA romance: local rules are real, and the machine that runs them is never free.

Chemputation, Cronin's coinage, is not simulating chemistry. It is using a state machine to make any molecule in the lab, reproducibly. Chemists carry tacit knowledge that never makes it into a paper: what "reflux two hours" actually means, how fast to stir, how pure the solvent was. Electronic lab notebooks he calls a data cemetery. About eight years before this conversation he tried to turn his group into robots with a programming language for walking around the bench. They kicked him out. The next day he offered to fund robots. That went better.

Any molecule, abstractly, is four operations in a loop: reaction, workup, separation, purification. Treat that as a chemical Turing machine. After an architecture conference and some 3D-printed test tubes, his group designed pumps, valves, and electronics in-house, each device with an IP address, power and data on one Ethernet cable into a router. Batch, not flow, so states can start and stop. Cleaning is built in, because electrons do not leave a mess and chemicals do. The first three encoded molecules were a sleeping drug, the anti-seizure compound rufinamide, and Viagra. Organic chemists then complained they were becoming bored programmers.

The answer was kai-DL, χDL, named by a student and a postdoc: first an XML-like script, now a Turing-complete language with loops, conditionals, and variables. A chemist types "add A to B, reflux"; a linter checks it; the graph of the hardware compiles the program — *chempilation* — onto any compatible rig. They have used natural-language processing to read papers into that state machine. An OpenAI Codex-inspired summer workshop produced a chemical analog Cronin calls Chemex: it proposes code, and a human must correct it. AI that writes and then runs chemistry without a person in the loop is, at this stage, unsafe. A mistranslated "reflux" is not a crash. It can be a fire.

The machines clog and fail, like ENIAC's exploding valves. They are also getting better. The lab now has about 15 of them, and some molecules the group would rather make on the computer than by hand. An anti-influenza compound, Arbidol, is six steps: a week of continuous human labor, or load the reagents, press go, and drink coffee. Cronin lists why chemistry needs a universal language — reliability, interoperability, collaboration, less ambiguity, lower cost, more safety, molecular customization, and publication of *executable* chemical code. A cancer candidate that takes three years and a finished PhD to repeat could become a download. That is the printing press, and also a supply chain for drugs that currently go out of print.

Fridman asks whether a reliable universal chemputer excites or terrifies him. The upside is a Moore's law of molecules: faster basic science, cheaper personalized drugs, manufacturing that is no longer trapped in a few factories. Cronin finds it outrageous that not all of humanity has access to medicine. Clinical-grade computers in towns around the world could make the next-best therapy on demand. A genome assistant that warns you of cancer in seven years, and a personal machine that cooks the molecule, is a dream he is willing to say out loud. The near-term version is a pharmacy, not a kitchen counter.

The nightmare he actually uses is the Germanwings crash: a co-pilot sets descent to zero and the autopilot flies into the Alps. Software needs situational awareness — computer says no. For kai-DL, computer science already has the tools. Encrypt programs, issue an FDA-like license for a single instance of use, make reverse-engineering as hard as breaking the key. Recreational drugs will force a social argument as costs fall; he wants that argument, not a shrug that lumps THC for epilepsy together with terrorism. Chemify, the company, has to talk with the people who pay for the tech about longer lives versus one-off misuse.

A draft paper on autonomous exploration of nanomaterials shows the same loop in physical space. A robot squirts gold and a reducing agent, reads the plasmon color, and runs a genetic algorithm plus MAP-Elites through exploration and then optimization. In three days and about a thousand experiments it searches a space he puts at 10 to the 23 possible reactions, five generations of sharper particles, then outputs code that another chemputer can run. The particles are dumb; the robot has the mind. Cronin's kill switch is unchanged: search objects that do not exist in the environment. Even a self-replicating nanomaterial, if it got out, would starve. That is the opposite of walking a dangerous organism toward human biology.

He would rather seed dead worlds than brute-force human bodies onto Mars. Elon Musk wants people off Earth; Cronin asks whether the payload should be a chemputer, or origin-of-life algorithms that embed culture and evolve life suited to Mars. Two biospheres on a tidally locked planet, never sharing chemistry until they become intelligent enough to notice each other, is a fantasy he likes. The moral obligation, if we never find life, is origin-of-life bombs — and then forgetting we did it, and coming back to ask where they came from. He knows how that sounds. He still thinks the universe needs more life.

Fridman worries that the inner life — hoping, fearing, loving — could die if we cannot pass it on. Cronin's robot lawnmower, which his kids named and gendered, would count as conscious under a crude reading of integrated information theory. He thinks assembly theory is the better measure, and that consciousness evolved over a population, not in a flash. The best way to make it, he says unashamedly, is chemical. A brain has more possible configurations than there are atoms in the universe, and it can switch among them. A Core i10 has tens of billions of transistors that you cannot reconfigure that way. Silicon AGI, if it comes, will still sit on a human causal chain. It cannot be separate from us.

The existential story he wants people to drop is the superintelligence that hates us. Fridman had already put it well: worry about dumb algorithms, written by humans, driving us insane at Twitter scale. Cronin agrees. He uses the platform to provoke thought, not outrage, because outrage makes people take sides and taking sides is a terrible state. Superintelligence that can show up at a party — wit, humor, emotional intelligence, the whole dance — would enrich life. Scaled stupidity is the old danger in a new amplifier. He is oddly calm about climate and heat death. If time is fundamental, the universe constructs; you do not have to give it back.

A dinner with Joscha Bach produced a tweet about Turing machines, wakes of non-existence, and automata that exist by default. Cronin thinks they misunderstand each other because Bach floats in language while Cronin builds concepts from the bench. The universe is not a Turing machine; biology is not either, because Turing machines do not evolve. Constructors came before abstractors. Alan Turing is a late gift of a universe that had to make biology and intelligence first. Cronin's third point is that something may sit below Church–Turing: the constructor that builds the computer that then notices it can abstract. The universe, he suspects, had less computational power in the past than it has now. Intelligence is not conserved. Neither is complexity. New books are being written, not just rearranged on the shelf.

What is the meaning of it? It depends on the day. We make our own meaning among ourselves; that layer needs no cosmology. The universe wants to explore every configuration it is allowed, and we do not see a combinatorial explosion — we see memory, constructive and destructive interference. Maybe the point is to make as many objects as possible. He does not buy heat death. With Sarah Walker at Arizona State he argues about how creative and how determined the universe is; they are converging on the admission that we do not yet know the rules that let the universe contemplate itself, so asking for the meaning first is premature. If an assumption cannot be tested by experiment or thought experiment, it is without merit. Fridman closes with Rick Sanchez: to live is to risk it all, or you are just randomly assembled molecules drifting wherever the universe blows you. Cronin would add that molecules which keep a memory are no longer only molecules.