You’ve been taught that computers think in straight lines. Input, process, output. Step one, step two, step three. It’s clean, it’s comforting, and it’s fundamentally wrong.
Not wrong in a small way. Wrong in the way that Copernicus was right and everyone else was staring at the sky backwards.
In 2021, before the AI explosion swallowed every conversation, Stephen Wolfram’s physics project dropped a finding that should have cracked open how we think about computation itself. It involved something called a multiway Turing machine — and the discovery that computation might not be a line at all, but a braid.
Every computation you’ve ever run is just one thread pulled from a braid the universe already wove.
Let me explain why this matters, and why almost nobody is talking about it.
A classical Turing machine — the theoretical device that underpins all of computer science — reads a tape, follows rules, and moves step by step. It’s deterministic. Or, when we’re feeling adventurous, non-deterministic, which is just a fancy way of saying “it picks a path.” We’ve treated these as two flavors of the same machine.
But multiway Turing machines don’t pick a path. They take ALL the paths. Simultaneously. Every possible state transition branches outward, and the full computation is not any single thread but the entire branching structure — what Wolfram calls a multiway graph.
Here’s where it gets strange. When you map these branching computational histories, they form structures that look exactly like braids. The same braids that show up in knot theory. The same braids that describe particle interactions in topological quantum field theory. The same braids that appear in the mathematics of physical reality.
Computation doesn’t just describe the universe. It might BE the universe, braiding itself into existence.
If you’re feeling skeptical, good. I was too. The instinct is to say: “Sure, you can build an abstract model where all paths exist, but that’s just math. Real computers don’t work that way.”
But that’s the twist. The point isn’t that your laptop suddenly runs all paths at once. The point is that we’ve been confusing the implementation for the principle. Your laptop is a single-thread projection of something far more vast — like a 2D shadow of a 3D object, and we’ve spent eighty years studying the shadow.
Consider what this means for AI. Right now, the dominant paradigm is: train a model, run inference, get an output. Sequential. Linear. One path. But if computation is fundamentally multiway — if the natural state of a computational system is to branch and braid — then our current AI architectures are like trying to understand a river by studying a single water molecule’s trajectory.
The most powerful intelligence we know — human cognition — doesn’t compute linearly. It branches, it contradicts itself, it holds multiple thoughts in superposition, and then it collapses into a decision. Sound familiar?
This is the deeper symmetry that the multiway Turing machine reveals: branching isn’t a bug in computation. It’s not an optimization trick. It’s the native geometry of how information processes itself. Determinism and non-determinism aren’t opposites — they’re two views of the same braid, one seeing a single strand, the other seeing the whole.
Wolfram’s team didn’t just stumble onto a mathematical curiosity. They found a bridge between the abstract world of computation and the physical world of topology and geometry. The braid structure that emerges from multiway computation is the same structure that physicists use to model everything from quantum entanglement to the fabric of spacetime.
When the math of computation and the math of physics converge on the same shape, you’re not looking at a coincidence. You’re looking at a clue about what’s actually real.
For anyone building AI, studying intelligence, or simply trying to understand what computation IS — not what we’ve built on top of it, but what it is at its foundation — this should be a wake-up call. We’ve been optimizing within a paradigm that might be a special case of something much larger.
The next leap in AI won’t come from bigger transformers or more parameters. It’ll come from someone who takes the multiway structure seriously — who builds a system that doesn’t just pick one path through computation, but braids them all and lets the structure itself do the thinking.
The universe doesn’t compute in a straight line. Why should our machines?
FAQ
Q: Isn't this just theoretical math with no real-world application?
A: So was quantum mechanics in 1925. Multiway computation already describes physical systems through braid topology — the gap between 'interesting math' and 'engineering breakthrough' is closing fast, especially as AI architectures hit the limits of sequential thinking.
Q: What does this mean for people building AI today?
A: It means the current paradigm — train, infer, output — is a projection of a richer computational reality. The next architecture breakthrough likely won't come from scaling transformers but from systems that natively branch and braid computational paths, closer to how human cognition actually works.
Q: Is Wolfram just overselling his own framework?
A: Wolfram has a habit of being early and loud. But the braid-computation connection isn't his invention — it emerges naturally from the math. The structure speaks for itself. Whether his broader physics project succeeds, this specific insight stands on its own.