AI is changing the world. Don't stay behind. Clear summaries, community insight, delivered without the noise. Subscribe to never miss a beat.

Designing more efficient quantum circuits with Claude

anthropic.com

October 16, 2026

18 min read

🔥🔥🔥🔥🔥

48/100

Summary

Anthropic says Claude Mythos 5 autonomously optimized eight quantum-circuit primitives after receiving plain-language specifications and feedback from an automated evaluator. The system generated circuit designs for chemistry, optimization, quantum simulation, number theory, and numerical computing that Anthropic says used fewer qubits and gates than previously published designs. The work was led by Andi Gu through Anthropic’s STEM Fellows Program with collaborators including Harvard’s Susanne Yelin and Mikhail Lukin. In a 53-hour run, Claude developed a reversible modular-inversion method for elliptic-curve point addition. It derived a way to recover divsteps branch information from the numerical state rather than storing the full history. Anthropic says one resulting design uses 12% more qubits than the smallest earlier circuit while requiring 17 times fewer gates, expanding the available qubit-runtime trade-offs. Claude also redesigned the Reed–Solomon decoding step in decoded quantum interferometry, replacing an extended-Euclidean construction with Berlekamp–Massey decoding. A simulator tested 403 million sample decodings and found that reserving 19 qubits for one exceptional value was sufficient for nearly all runs; about one in 10,000 runs exceeded that capacity and failed within the allowed error rate. Anthropic says the decoder uses 12 times fewer gates than Google’s prior construction.

Key Takeaways

Anthropic says Claude Mythos 5 improved quantum circuits across eight computational primitives through an automated propose, evaluate, and revise loop without human input between rounds.
Claude’s reversible modular-inversion design avoided storing divsteps branch history by decoding it from the computation’s numerical state.
Anthropic says one elliptic-curve circuit design traded 12% more qubits for 17 times fewer gates than the smallest previous design.
Claude’s Reed–Solomon decoder for decoded quantum interferometry used Berlekamp–Massey decoding and, according to Anthropic, reduced gate count by a factor of 12 versus Google’s previous construction.
Quantum circuits must follow fixed reversible operations, making conditional branching and disposal of intermediate data costly in qubits and gates.

What the discussion said

The thread treated the result less as a mysterious quantum breakthrough than as a particularly well-shaped optimization loop for a language model. Several commenters noted that the target circuits are expressed in a compact, classically simulatable instruction language, so candidate rewrites can be checked automatically and bad ideas discarded quickly. In that framing, large reductions in operation counts are useful and credible: Claude reportedly simplified eight simulated circuits, sometimes by recomputing temporary values and sometimes by substituting a stronger algorithmic approach. The praise was qualified rather than breathless. One influential view held that this is exactly the kind of bounded search-and-verify problem where current models excel: exploiting a known design space more thoroughly than humans had time to do. That makes the automation valuable, but not evidence that the model is creating fundamentally new quantum-computing knowledge. A more jokingly alarmed branch imagined AI improving quantum hardware to accelerate itself, while a reply argued advanced general AI is likelier to arrive before practical quantum machines matter. There was also a sharp credibility complaint about how prominently the organization described its role compared with the paper's more limited acknowledgement of fellowship support. Even a self-described AI skeptic conceded that lowering circuit costs is hard to object to on the merits.

Where opinion split

The central dispute is whether the work demonstrates meaningful AI-led scientific progress or merely exhaustive optimization inside a pre-existing formal sandbox. Supporters see verified circuit compression as a practical win with real computational value; skeptics argue that searching and filling gaps in an established space is not the same as expanding the frontier of quantum science.

Read original article

Community Sentiment

Positive

Positives

Automated verification turns circuit optimization into a tight propose-test-reject loop, giving language models a rare setting where useful rewrites can be validated instead of merely sounding plausible.
Cutting simulated quantum-circuit operations by as much as 17-fold could make expensive computations materially cheaper, especially when the model discovers redundancy or swaps in a better algorithm.
The result wins over at least some AI skeptics because reducing concrete computational waste is a plainly beneficial application rather than a vague demo.

Concerns

The achievement looks like high-powered search within an already mapped instruction language, not evidence that the model has uncovered new quantum principles or expanded scientific understanding.
Claims that circuit optimization could help AI bootstrap itself through quantum computing were dismissed as premature because usable quantum hardware remains far behind near-term AI progress.
The organization's presentation drew skepticism for implying stronger institutional ownership than the paper's acknowledgements appeared to support, weakening trust in the research framing.