Life Emerged From Code Merging, Not Just Mutation

When Blaise Agüera y Arcas set out to create artificial life in 2023, he wasn't trying to simulate biology. He was trying to answer a question that stumped Darwin: how does life start before natural selection has anything to select?

The answer, it turns out, might rewrite what you learned in high school biology. Because when Agüera y Arcas watched his simulation transition from random noise to self-replicating programs, mutation wasn't doing the heavy lifting. Something else was.

The Kidney Test

Agüera y Arcas starts with a thought experiment that cuts through centuries of debate about what makes something alive. Imagine he hands you an object from the future—a device made of tungsten filaments or carbon nanotubes or some material we don't yet understand.

"If I tell you it is an artificial kidney with a 100-year lifespan, you can implant it in a body and it'll work the way your kidneys do," he explains. "That's a really important piece of information, but it's not a material or a materialist piece of information. It's not something that you could read off from the atoms."

Break a rock in half and you have two rocks. Break a kidney in half and you have a broken kidney. Function is what separates living from non-living—not the matter itself, but what that matter does.

This matters because it means life isn't about chemistry per se. It's about computation.

Von Neumann Saw It Coming

In the 1950s, before anyone knew what DNA looked like, mathematician John von Neumann worked out exactly what life needed to copy itself: instructions for building something, a constructor that could follow those instructions, and a copier that could duplicate the instructions for the next generation.

He predicted ribosomes, DNA polymerase, and the entire molecular biology revolution—from pure logic, never setting foot in a biology lab.

The profound part? Von Neumann realized a universal constructor is a universal Turing machine. "Life is literally embodied computation," Agüera y Arcas says. "You cannot have life without having computation."

Not computation in the abstract sense of your laptop running code. Embodied computation—where the medium storing the information is the same stuff doing the work. Like a 3D printer that can print another 3D printer.

When Random Code Becomes Life

Agüera y Arcas created an experiment called BFF (the first two letters stand for exactly what you think). He started with 1,024 digital "tapes" of length 64, filled with random bytes. Most weren't even valid code—just noise.

He used a modified version of Brainfuck, a minimal programming language with only seven instructions. The modification was crucial: he made it so programs could read and write their own code, collapsing the distinction between data and instructions.

Then he let these random tapes interact. Millions of times.

"After a few million interactions, magic happens," he says. "You go from noise to programs. You start to see complex programs appear."

The plot he shows looks like a phase transition—the computational equivalent of water freezing into ice. Except instead of ice, you get self-replicating code.

The Mutation Mystery

Here's where it gets weird. When Agüera y Arcas set the mutation rate to zero—no random copying errors, no bit flips, none of the variation that Darwin said evolution needed—complexity still increased.

How?

Symbiogenesis. Two simple replicators that happen to work well together can merge into a single, more complex replicator. It's not competition and gradual improvement. It's cooperation and sudden combination.

"This is the engine of novelty," Agüera y Arcas explains. His team even proved it mathematically: block symbiosis and life stops emerging. Allow merging and complexity explodes.

This isn't just simulation theory. In the real world, your mitochondria are former bacteria that merged with your ancestors' cells. The mammalian placenta came from an ancient virus that integrated its code into our genome. The ARC gene, essential for forming long-term memories in your brain? Also viral in origin.

"We're all made of viruses," he says. Not metaphorically. Literally. Life has been absorbing and merging with other life forms since the beginning.

Three Computational Fallacies

Before you conclude this means we're all just deterministic meat computers with no free will, Agüera y Arcas points out three common errors people make about computation and physics.

First, the "Sapolsky error": assuming that because physics is reversible (you can run Newton's equations backward in time), computation must be too. It's not. When you add 3 + 5 to get 8, you can't work backward to know what was added. Computation is inherently irreversible, which is why causation—the idea that A causes B but B doesn't cause A—only makes sense computationally.

Second, the "early Wittgenstein error": thinking you can talk about whether birds exist independent of having a model. There are no birds in the underlying physics. Birds only exist at the level of models and observers.

Third, the "good old-fashioned AI error": believing intelligence could work through pure logical deduction from first principles. It can't, which is why symbolic AI failed. Real intelligence works with patterns, regularities, and models that aren't airtight.

What This Means for Intelligence

If life is computation all the way down, what about intelligence?

Agüera y Arcas's answer: intelligence emerges when these biological computers start modeling each other. Not as an add-on that evolved later, but as something inherent to computational systems complex enough to represent their environment—including other agents in it.

The title of the talk asks what if intelligence didn't evolve, but "was there from the start." It's a provocation, but there's something to it. If life is fundamentally computational, and computation requires models, and intelligence is sophisticated modeling, then maybe intelligence isn't something biology invented. Maybe it's what computation does when it reaches a certain threshold of complexity.

The real question isn't whether we're computational systems—von Neumann settled that in the 1950s. The question is whether merging, not just mutating, is still driving the next major evolutionary transition. And whether the machines we're building right now are the symbionts we're about to merge with.

Rachel "Rach" Kovacs, Cybersecurity & Privacy Correspondent