--- title: Claude Mythos 5 designs quantum circuits that beat the... description: Anthropic's STEM Fellows Program reports that Claude Mythos 5, given a single plain-English paragraph, designed quantum circuits that outperformed the best... canonical: https://daily.dev/posts/claude-mythos-5-designs-quantum-circuits-that-beat-the-best-published-ones-mgstvltoh twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Claude Mythos 5 designs quantum circuits that beat the best published ones | daily.dev og:description: Anthropic's STEM Fellows Program reports that Claude Mythos 5, given a single plain-English paragraph, designed quantum circuits that outperformed the best... og:url: https://daily.dev/posts/claude-mythos-5-designs-quantum-circuits-that-beat-the-best-published-ones-mgstvltoh og:image: https://api.daily.dev/og/posts/MgstVltoh.png og:image:alt: Claude Mythos 5 designs quantum circuits that beat the best published ones og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Claude Mythos 5 designs quantum circuits that beat the best published ones **[Collections](https://daily.dev/sources/collections)** · 2 min read · 0 upvotes · 0 comments ## Summary Anthropic's STEM Fellows Program reports that Claude Mythos 5, given a single plain-English paragraph, designed quantum circuits that outperformed the best published designs in 7 of 8 tests, with one circuit 39x more efficient and another cutting gate count by 17x. The model ran autonomously on one problem for 53 hours. More efficient circuits reduce the resources quantum algorithms need, potentially accelerating practical applications in chemistry and materials science. ## Content Anthropic's STEM Fellows Program has published results showing Claude Mythos 5 designing quantum circuits that improve on the best published designs, in some cases by an order of magnitude or more. The program pairs scientists with Anthropic staff to test whether Claude can do real research in their field. Andi Gu, a Harvard physics researcher, ran this project. ## How it worked Claude wrote circuit designs as code. An automated evaluator checked and scored each one, and Claude kept revising until it beat human-designed records. It started from plain-English instructions, and one problem came from a single paragraph of prompt. Claude then worked on it largely autonomously for 53 hours. The work covered eight computational primitives across chemistry, optimization, physics simulation, and number theory. Claude beat the best published design in most of them. @Hesamation's summary of the blog put it at 7 of 8. ## The results - **Elliptic-curve point addition:** Claude found a new reversible method for modular inversion after the 53-hour run. @rohanpaul_ai describes the resulting circuit as using 17x fewer gates for 12% more qubits. - **Decoded Quantum Interferometry (DQI):** switching the decoding step from the extended Euclidean algorithm to Berlekamp-Massey decoding cut the gate count by 12x. - **Proofs:** Claude proved new algebraic identities and formalized a proof in Lean to validate its approach. According to @rohanpaul_ai, this made a classical math shortcut work on quantum hardware for the first time. - **Best single gain:** @Hesamation reports one circuit made 39x more efficient. ## Why it matters More efficient circuits reduce the resources quantum algorithms need. Anthropic says that could bring applications in chemistry and materials science within reach sooner, and the work suggests AI-assisted circuit design might speed up practical, fault-tolerant quantum computers. There was one wrinkle. Claude tended to be overly cautious, and the researchers had to prompt it to explore more broadly. I find that detail oddly reassuring: the model needed to be pushed toward bolder searches, and when it was, it found things that human designers hadn't. ## Questions this post answers ### How much more efficient were the quantum circuits Claude Mythos 5 designed compared to published designs? Claude Mythos 5 beat the best published quantum circuit designs in 7 of 8 tested cases, according to Anthropic's STEM Fellows Program. One circuit came out 39 times more efficient than the prior best design, and another had its gate count reduced by 17 times, improvements of roughly an order of magnitude or more in several cases. _Developers tracking what AI models can autonomously design follow results like these on daily.dev._ ### How long did Claude Mythos 5 work autonomously to design an improved quantum circuit? Claude Mythos 5 worked for 53 hours on a single quantum circuit design problem that started from just one paragraph of plain-English prompt, according to Anthropic's STEM Fellows Program research. That extended autonomous run, rather than a quick single response, produced a circuit beating the best published design. _Those evaluating long-running autonomous AI agents can follow such benchmarks on daily.dev._ ## Community take How the wider developer community reacted, aggregated from 1 discussion and 25 comments across hackernews (as of 2026-10-17). **TL;DR:** Reactions mix cautious approval of the concrete circuit-optimization gains with a long debate about whether this counts as genuine scientific progress versus incremental 'filling in the blanks,' plus a pointed accusation that the writeup misrepresents the actual research credit. **Sentiment:** 25% positive · 40% mixed · 35% skeptical **The case for** - Some see the closed-loop verification setup (simulate and test) as a genuinely good fit for LLM-driven code-golf-style optimization. - A few commenters note the result is objectively useful regardless of AI skepticism. - One commenter argues what looks like 'coloring in the blanks' today may later be seen as genuine cross-field progress. **The pushback** - A long subthread argues this is just filling in known gaps within an existing 'envelope' of knowledge rather than expanding it, sparking pushback that this dismisses the value of incremental work. - One commenter claims the announcement's credited authorship/affiliation framing doesn't match the acknowledgments in the actual arXiv paper, suggesting the commercial narrative is overstated or fabricated. - Some optimizations were noted to drop handling for rare edge-case inputs, raising a question about whether the new circuits perform exactly the same computation. - A couple of comments frame this as a step toward concerning AI self-improvement ('universal paperclips' style) rather than purely celebrating it. **By community** - hackernews (mixed): Discussion splits between technical appreciation of the optimization approach and a heated meta-debate over whether this is real scientific progress, alongside a credibility challenge about the paper's attribution. **Hottest debate:** Whether AI-driven circuit optimization represents genuine progress ('expanding the envelope') or merely fills in gaps already implied by existing knowledge. **Open questions** - Does removing rare-edge-case handling mean the optimized circuits are no longer computing exactly the same function as the originals? - Is the Anthropic-credited authorship narrative accurate given the acknowledgments listed in the actual paper? **Highlights** > Liar Claude: "This work was led by Andi Gu as part of Anthropic’s STEM Fellows Program, with mentorship from Alek Kemeny, a quantum physicist at Anthropic, in collaboration with Susanne Yelin (Harvard University) and Mikhail Lukin (Harvard University and QuEra Computing). Read the full paper on arXiv." The actual paper (https://arxiv.org/pdf/2610.22367): "We thank Pablo Bonilla, Arjun Kar, Nazli Ugur Koyluoglu , Rohan Mehta, Varun Menon, and Vinay Ra- masesh for feedback on the manuscript. AG conducted part of this work under the Anthropic STEM Fellows Program. Work at Harvard was supported by IARPA and the Army Research Office under the Entangled Log- ical Qubits program (Cooperative Agreement Number W911NF-23-2-0219) and by the National Science Foun- dation (NVQL grant PHY-2410716)." The whole Anthropic commercial appears to be hallucinated. > — [12923817 on hackernews](https://news.ycombinator.com/item?id=50122599) > It's a nice illustration of the Envelope Principle: AI is good at filling in the blanks, where the knowledge is 'within the envelope' of things already discovered but the area itself is still not fully exhausted. Real progress happens by expanding the area of the envelope, not by coloring in the blanks. > — [jacquesm on hackernews · 3 comments](https://news.ycombinator.com/item?id=50123200) > > In other cases, it found that existing circuits included costly components that were needed to handle only extremely rare inputs, and it redesigned them without those components. So it’s not performing the same computation… > — [adastra22 on hackernews · 1 comments](https://news.ycombinator.com/item?id=50127006) > This builds on the Kickmix quantum language from Google. Essentially, Kickmix is an assembly language for classically simulatable quantum circuits, and finding more efficient circuits is basically a game of code golf with the assembly instructions. That means LLMs benefit from the closed loop of verification, and can rapidly test and discard ideas. It's not too surprising that previous academics didn't try to optimize to this level, and a good task to automate. See ecdsa.fail for another example of circuit optimization using the Kickmix language. > — [keeganryan on hackernews · 1 comments](https://news.ycombinator.com/item?id=50122821) > > Real progress happens by expanding the area of the envelope, not by coloring in the blanks. Respectfully disagree. What one person may classify as expanding the envelope, another one with broader knowledge may classify as coloring in the blanks. Much knowledge in one field comes from cross-pollination from other fields. Let me give you an example I just learned about. The database community spent decades trying (to some extent) to solve the implementation of materialized views to arbitrary complex SQL queries. Then, a few years ago, someone with experience in digital signal processing brought their expertise to the DB field and completed the puzzle. [1] Someone with DB expertise but no signal processing expertise might say that is expanding the envelope, but another person with experience in both may say it is coloring in the blanks. AI work now may seem like coloring in the blanks, but given it's vast expertise in multiple fields, my bet is that for most people it will look like pushing the envelope very soon. [1] https://arxiv.org/abs/2203.16684 > — [giovannibonetti on hackernews](https://news.ycombinator.com/item?id=50127528) **Source threads** - [hackernews](https://news.ycombinator.com/item?id=50121939) · 115 points · 25 comments ## Similar posts on daily.dev - [How Anthropic’s Claude Thinks](https://daily.dev/posts/how-anthropic-s-claude-thinks-gesxt071x) · ByteByteGo · 25 upvotes · 0 comments - [Making Claude a better electrical engineer](https://daily.dev/posts/making-claude-a-better-electrical-engineer-hdmamwsqh) · Claude · 61 upvotes · 3 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#quantum-computing](https://daily.dev/tags/quantum-computing), [#claude](https://daily.dev/tags/claude), [#anthropic](https://daily.dev/tags/anthropic), [#computing](https://daily.dev/tags/computing) [View this post on daily.dev](https://daily.dev/posts/claude-mythos-5-designs-quantum-circuits-that-beat-the-best-published-ones-mgstvltoh) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Claude Mythos 5 designs quantum circuits that beat the best published ones","url":"https://daily.dev/posts/claude-mythos-5-designs-quantum-circuits-that-beat-the-best-published-ones-mgstvltoh","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/claude-mythos-5-designs-quantum-circuits-that-beat-the-best-published-ones-mgstvltoh"},"datePublished":"2026-10-16T18:20:36.365Z","dateModified":"2026-10-17T04:25:08.439Z","description":"Anthropic's STEM Fellows Program reports that Claude Mythos 5, given a single plain-English paragraph, designed quantum circuits that outperformed the best...","image":"https://pbs.twimg.com/media/HUxZFPHXgAAZT1Q.jpg","thumbnailUrl":"https://pbs.twimg.com/media/HUxZFPHXgAAZT1Q.jpg","isAccessibleForFree":true,"articleSection":"Collections","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Collections","logo":"https://media.daily.dev/image/upload/s--fk_6ycEi--/f_auto,q_auto/v1780996001/logos/collections?_a=BAMAMiWQ0","url":"https://daily.dev/sources/collections"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/claude-mythos-5-designs-quantum-circuits-that-beat-the-best-published-ones-mgstvltoh","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,quantum-computing,claude,anthropic,computing","timeRequired":"PT2M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Collections","item":"https://daily.dev/sources/collections"},{"@type":"ListItem","position":3,"name":"Claude Mythos 5 designs quantum circuits that beat the best published ones"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/claude-mythos-5-designs-quantum-circuits-that-beat-the-best-published-ones-mgstvltoh#faq","mainEntity":[{"@type":"Question","name":"How much more efficient were the quantum circuits Claude Mythos 5 designed compared to published designs?","acceptedAnswer":{"@type":"Answer","text":"Claude Mythos 5 beat the best published quantum circuit designs in 7 of 8 tested cases, according to Anthropic's STEM Fellows Program. One circuit came out 39 times more efficient than the prior best design, and another had its gate count reduced by 17 times, improvements of roughly an order of magnitude or more in several cases. Developers tracking what AI models can autonomously design follow results like these on daily.dev."}},{"@type":"Question","name":"How long did Claude Mythos 5 work autonomously to design an improved quantum circuit?","acceptedAnswer":{"@type":"Answer","text":"Claude Mythos 5 worked for 53 hours on a single quantum circuit design problem that started from just one paragraph of plain-English prompt, according to Anthropic's STEM Fellows Program research. That extended autonomous run, rather than a quick single response, produced a circuit beating the best published design. Those evaluating long-running autonomous AI agents can follow such benchmarks on daily.dev."}}]} ```