54 Qubits Braided and Fused: Non-Abelian Codes Demonstrate a Universal Gate Set on Quantinuum Hardware

The Cost Accounting First

In today's mainstream fault-tolerant quantum computing architectures, the vast majority of physical qubits never get to compute. They are spent on error correction: spreading one logical bit's information across many physical qubits so local noise cannot destroy it.

This scheme preserves information—but leaves a hole. Once protected, the information still cannot support the full set of operations required for universal computation. The standard engineering patch is called magic state distillation: purifying a batch of noisy magic states until they become usable. In standard error-correcting codes, this is among the most expensive processes and consumes a large fraction of the physical qubits.

A paper published July 15, 2026 in *Nature* (Nature 655, 591-597, 2026) offers a different way to plug this hole.

Two Roads, and One Dead End

Topologically ordered phases offer two routes to quantum memory:

• Ground-state subspace encoding (represented by the toric code) — but this does not natively support a universal gate set.
• Encoding in anyon excitations (topological quantum computation) — gates are implemented by braiding non-Abelian anyons around one another.

The problem: the simplest non-Abelian generalizations of the toric code have been proven not to achieve universality through braiding alone.

What this paper does is elevate fusion to a computational primitive. Two anyons are brought together, and the merged state is measured. Braiding *plus* fusion turns these minimal non-Abelian topologically ordered states into a universal resource.

Why the Switch to S3

Previously, the same group of researchers worked with D4—the rotations and reflections that leave a square invariant. In 2024, Verresen's team demonstrated this non-Abelian order on quantum hardware for the first time, on Quantinuum's trapped-ion machine, proving such exotic particles could be created and manipulated.

Verresen's own assessment: that result did not prove these emergent forces were sufficient for quantum computing—the created universe wasn't powerful enough.

S3 is the rotations and mirror flips that leave an equilateral triangle invariant—the smallest non-Abelian group. The team entangled 54 qubits on Quantinuum's H2 processor, prepared the ground state of the S3 quantum double, and encoded logical information in the global fusion space of the non-Abelian anyons.

A benefit of storing information here: it is spread across many entangled qubits, located nowhere in particular, naturally shielding it from small perturbations that would flip ordinary qubits.

Incidentally, what is encoded here is a topological qutrit—three levels, not two.

The Demonstration: Making a Magic State

Universal gate sets are hard to demonstrate directly, so the paper picked a concrete target: topologically preparing a magic state.

The pipeline: prepare the S3 quantum double ground state on 54 qubits → encode logical information in the global fusion space → apply braiding operations and fusion-and-measurement → universal topological gate set plus readout → topologically prepare a magic state—bypassing distillation and cultivation.

Henrik Dreyer, Director and Head of Science at Quantinuum's Munich office, put it plainly: non-Abelian codes are a dark horse in the quantum error correction race, and this work delivers the first universal gate set within non-Abelian codes—suggesting fault-tolerant computation can in principle proceed without magic state distillation or cultivation, the most expensive operations in standard quantum error-correcting codes.

Verresen's framing is more vivid: if you treat these emergent particles as a version of quarks, store information inside them and move it around, any quantum computation becomes possible.

The theoretical roots go back to 2003, when Carlos Mochon, then a student of John Preskill at Caltech, proposed the underlying concept. Turning the paper concept into a hardware-executable experiment took more than two decades of theory and experiment.

The Weight of the Paper

The author list includes Harvard graduate students Chiu Fan Bowen Lo and Anasuya Lyons; Quantinuum's Dan Gresh, Michael Mills, Peter E. Siegfried, Maxwell D. Urmey, Nathanan Tantivasadakarn, and Dreyer; Harvard's Ashvin Vishwanath; Ruben Verresen of the University of Chicago's Pritzker School of Molecular Engineering; and Mohsin Iqbal. Data and numerical simulation code are available on Zenodo (doi:10.5281/zenodo.18054264).

Lo and Lyons shared a reflection: seeing something imagined during their PhD realized in the lab is deeply satisfying—something made possible only by recent advances in quantum hardware.

How Far Should You Trust It

Three boundaries:

1. This is a proof of principle. 54 physical qubits, one ground state, a set of gates and one readout—several orders of magnitude away from a machine running practical algorithms. 2. It proves the path works in principle; it does not prove it is cheaper in engineering terms than the mainstream surface-code-plus-distillation route. A real comparison requires post-scaling fidelity, gate times, and preparation overheads—numbers that don't exist yet. 3. A magic state was prepared topologically, but a full architectural argument stands between that and large-scale fault-tolerant computing without distillation. Dreyer's statement retains the qualifier "in principle."

Why It's Worth Watching

The tightest constraint in quantum computing today is not whether more qubits can be built, but how many of them survive to do actual work. Any route that cuts error-correction overhead changes the ledger of the whole machine.

Non-Abelian codes have long sat at the edge of the track, with mainstream attention on surface codes. This paper moves them from "theoretically interesting" to "universal gate set demonstrated on hardware." It does not yet earn the right to rewrite the roadmap—but it adds a branch to that roadmap that no one had seriously drawn before.

Things to Watch

• First steps of scaling: how preparation fidelity degrades beyond 54 qubits.
• Head-to-head cost comparison with surface codes: physical qubits and time per logical operation on both routes.
• Engineering cost of fusion as a primitive: whether measurement and post-processing overheads are fully accounted for.
• Other non-Abelian groups: S3 is the smallest—would larger groups yield cheaper gates?
• Independent replication: can other hardware platforms reproduce this S3 preparation?

References

1. Chiu Fan Bowen Lo, Anasuya Lyons, Dan Gresh, Michael Mills, Peter E. Siegfried, Maxwell D. Urmey, Nathanan Tantivasadakarn, Henrik Dreyer, Ashvin Vishwanath, Ruben Verresen, Mohsin Iqbal, *Universal gates from braiding and fusing anyons on quantum hardware*, Nature 655, 591-597 (2026), 2026-07-15 2. University of Chicago Pritzker School of Molecular Engineering press release, September 2026 3. Quantinuum official announcement and Quantum News coverage, July 2026 4. ScienceNet, "Universal gates from braiding and fusing anyons on quantum hardware," 2026-07-19