Quantum error correction is perhaps the most significant barrier to practical quantum computers. Modern qubits are extremely sensitive to external noise, and without reliable error correction mechanisms, scaling computations remains out of reach. Recent advances from Microsoft Quantum and IBM Research show that the industry is approaching a solution to this fundamental problem.
Microsoft and Quantinuum: 800-fold Error Reduction
The collaboration between Microsoft Quantum and ion-based quantum processor developer Quantinuum has yielded impressive results. In an experiment on the Quantinuum processor, two optimized error correction schemes were tested: a 12-qubit code inspired by the Knill scheme and a 16-qubit tesseract color code. The former encodes two logical qubits, the latter four.
The results, published in the journal Nature, demonstrate a reduction in logical error rates by a factor of 11 to 800 compared to comparable physical circuits. Specifically, when preparing a Bell state, the error rate dropped from 0.8% for the physical circuit to 0.001%. This means that modern quantum devices are already capable of using fault tolerance to significantly suppress errors in non-trivial circuits. Repeated correction showed a result 51 times lower than the physical baseline, while preparing a 12-qubit cat state yielded a 22-fold improvement.
IBM Research: AI-driven Search for New Codes
IBM Research took a different approach, applying large language models (LLMs) to search for new error correction codes. The system, based on the open-source library OpenEvolve, analyzed bivariate bicycle codes—a type that IBM plans to use in its fault-tolerant quantum computers.
During initial runs, the system proposed 465 candidates. The most notable was the [[288,50,8]] code with 50 logical qubits—a significant improvement over the previous record of 16 for this family. A compact [[72,4,8]] code with 72 physical qubits and variants [[288,16,12]] and [[360,12,≤24]] were also discovered.
However, as IBM emphasizes, the practical applicability of these codes requires further verification. Some may be comparable to the [[144,12,12]] gross code, which the company plans to use in its future fault-tolerant computer, IBM Quantum Starling, slated for 2029.
Expert Opinion: These results are not merely academic achievements. They signal a transition from theoretical exploration to real engineering solutions. An 800-fold error reduction and the use of AI to discover new codes are steps that could make quantum computing a practical tool for cryptography, materials modeling, and financial analysis in the coming years. For the crypto industry, this is a reminder: the era of post-quantum security is not far off.