TL;DR
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
Researchers from Anthropic and Harvard used Claude to design quantum circuits for eight computational tasks. Seven designs reduced both estimated costly operations and qubit use against the study’s baselines, but the results are computational checks and resource estimates, not demonstrations on a quantum computer.
Anthropic and Harvard researchers say they used Claude to design quantum circuits with lower estimated resource requirements across seven of eight computational tasks, including a decoder design requiring about 12 times fewer Toffoli gates than its comparison circuit. The results, posted as a preprint, come from computational checks and resource estimates—not calculations run on a working quantum computer—so their practical value remains to be tested in complete algorithms and hardware.
The study examined eight computational building blocks used in areas including chemistry, optimization, physics simulation and number theory. Researchers report that seven designs reduced both costly quantum operations and qubit requirements relative to the study’s baselines. The eighth was an arithmetic design that offered different trade-offs between gate count and the number of qubits needed.
For one decoding circuit, the paper estimates a reduction from 5.1 million Toffoli gates to about 440,000. Its estimated peak qubit requirement fell from 1,938 to 1,837. Toffoli gates are a costly operation in many proposed error-corrected quantum-computing designs, so the count is one way to compare the amount of work a circuit may require. These figures describe circuit estimates, not measured performance on a quantum device.
The team used a process in which Claude could propose and revise circuits with programming tools and an evaluator that checked designs and scored their estimated resource use. Anthropic’s accompanying blog post identifies the model as Claude Mythos 5 and says its elliptic-curve arithmetic work took 53 hours. The post says human involvement consisted of initial instructions and occasional prompts to broaden the search.
Lower Costs for Quantum Routines
Quantum algorithms are built from smaller routines for operations such as arithmetic, data access and numerical calculations. Reducing the resources those routines require could lower the demands placed on a larger computation, a relevant goal as researchers work on both better algorithms and more capable quantum machines.
The reported changes address more than deleting redundant operations. Quantum circuits often need to preserve information so their operations can be reversed; temporary calculations may need to be stored or recomputed. Those choices create a trade-off between qubits and gates. The researchers describe designs that reconstruct some information when needed, let different computational paths share arithmetic, or avoid resources needed only for rare inputs. Whether these savings carry through to useful end-to-end applications is not yet established.
quantum computing simulator software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How the Circuit Designs Save Resources
A quantum circuit is a sequence of operations on qubits, the units used to store quantum information. Unlike ordinary software, quantum computations generally must be reversible. Temporary data cannot simply be overwritten; circuits may need to clear intermediate work through a process called uncomputation. Storing more intermediate results can require additional qubits, while recomputing or clearing them can add gates.
One highlighted result concerns a decoder used in decoded quantum interferometry, an algorithmic approach that can make better answers more likely to appear when a quantum state is measured. The decoder is part of an optimization routine; in this setting, it is not a circuit for correcting faults in the quantum hardware. The study describes replacing an approach based on the extended Euclidean algorithm with one based on Berlekamp–Massey, then limiting storage by reconstructing information rather than keeping a record of every intermediate step.
The paper says a storage need that could reach 665 qubits in the worst case was reduced to an allocation of 19 qubits, with the design accepting failure on rare inputs beyond that capacity. The source report also describes a second stored entry as needed with a probability of about one in 10,000. Those details illustrate that a lower resource estimate can depend on the allowed probability of failure, not just on circuit structure.
“Claude Mythos 5”
— Anthropic, in its accompanying blog post
As an affiliate, we earn on qualifying purchases.
Hardware Performance Still Untested
The reported work has not demonstrated these circuits on a working quantum computer. The available results are based on computational checks and estimates, so it remains unclear whether the designs can be integrated into complete algorithms and implemented efficiently on real machines.
The resource figures also depend on the study’s chosen comparison designs and assumptions. For some approaches, the reported savings involve trade-offs, including tolerating a small probability of failure for rare inputs. The source material does not establish how the circuits would perform under the noise, error-correction overhead and engineering limits of a particular quantum device, nor whether the estimated savings would translate into a practical speed or cost advantage.
The work is described as an arXiv preprint, rather than a peer-reviewed result in the supplied report. Independent evaluation and further testing would help establish how broadly the methods apply and whether other researchers can reproduce the designs and estimates.
As an affiliate, we earn on qualifying purchases.
Testing Beyond Resource Estimates
The next steps are to evaluate the proposed circuits in the context of complete algorithms, check how their resource estimates change under realistic hardware and error-correction assumptions, and test whether the designs can be implemented on quantum systems. The preprint and accompanying company account provide no timetable for such demonstrations.
Further scrutiny can also clarify the comparison baselines, the conditions under which the circuits allow rare failures, and how much human guidance was required across all eight tasks. Until those questions are answered, the study is evidence that AI-assisted search can produce candidate circuit designs with lower estimated resource needs—not evidence that quantum applications have already become cheaper or faster in practice.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did Claude help design?
Researchers from Anthropic and Harvard used Claude to develop quantum circuits for eight computational tasks. They report lower estimated gate and qubit requirements for seven designs compared with the study’s baselines.
Were the circuits run on a quantum computer?
No hardware demonstration is reported. The results are based on computational checks and resource estimates; performance on a working quantum device remains unconfirmed.
How large was the reported reduction in the decoder example?
The paper estimates about 440,000 Toffoli gates, compared with 5.1 million in the comparison design. The estimated peak qubit requirement fell from 1,938 to 1,837.
Does the result prove quantum computing will become more practical?
No. It suggests that AI-assisted design may find circuits with lower estimated resource demands. Whether those savings hold in complete algorithms and on real quantum hardware has not yet been shown.
Source: rss
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
