In a research post published on October 16, 2026, Anthropic states that Claude Mythos 5, working in a closed loop with an automated evaluator, designed circuits for eight computational primitives that improve on the best previously published designs, reducing both Toffoli-gate and qubit counts on seven of the eight, with a companion paper on arXiv.

● OPEN · settles 2026-10-21

Background

Quantum algorithms are usually costed in qubits and gates, and both are scarce, so the field spends much of its effort redesigning circuits to need less of each. Anthropic handed Claude Mythos 5 plain-language specifications for eight such circuits — basic routines drawn from chemistry, optimization, quantum simulation, number theory and numerical computing — plus an evaluator program that checked every proposal for correctness and scored it on qubits multiplied by gates. Working in that loop with no human input between rounds, Claude came back with designs that improve on the best published ones, in some cases by an order of magnitude or more; on seven of the eight tasks the new circuits cut both the Toffoli-gate count and the qubit count. The eighth, elliptic-curve point addition, took a single 53-hour run with minimal prompting. Its costly step, modular inversion, had resisted a practical quantum version because the classical method, divsteps, keeps a running record that a reversible circuit cannot simply throw away. Claude derived a way to read that record back out of the numbers themselves, proved the decoding rule, checked it in large-scale simulation and formalised it in Lean, producing the first reversible divsteps implementation that stores no branch record at all. One design in the resulting family uses 12% more qubits than the smallest previous circuit and 17 times fewer gates. On a second task, Claude rebuilt the decoder at the heart of decoded quantum interferometry around Berlekamp-Massey decoding instead of the extended Euclidean algorithm, reserved 19 qubits for saved values where a worst-case design would have set aside 665 — a figure it reached by running 403 million sample decodings and accepting a failure rate of about one in 10,000, inside the task's allowance — and cut the decoder's gate count by a factor of 12 against Google Quantum AI's construction. The work was led by Andi Gu under Anthropic's STEM Fellows Program with Alek Kemeny, Susanne Yelin of Harvard and Mikhail Lukin of Harvard and QuEra Computing, and a companion paper is on arXiv. None of the circuits has run on quantum hardware; the gains are in designs, checked by simulation and proof.

This market exists because launch announcements are easy to make and hard to grade. Press releases describe what a company intends to ship; they rarely come back later to say whether it shipped. AI Rank Bank pulls one checkable sentence out of the coverage — the sentence above — gives it a date, and lets readers price it in public. The launch itself stays with its publisher; what lives here is the claim, the crowd's call, and eventually the receipt.

How this claim is judged

Judging criteria, fixed when the market opened: Delivered if, at the check time, the post and the arXiv paper remain published reporting those results — eight tasks, seven with reductions in both Toffoli gates and qubits, the eighth yielding a new reversible modular-inversion architecture — and neither has been withdrawn or materially corrected. A retraction, or a correction showing the reported counts do not beat the previously published designs, does not count.

The source of record is the publisher's own page linked above. A claim about a company's product is judged against that company's public output — its product pages, documentation, changelogs and official posts — not against rumors or third-party summaries. Only yes-or-no outcomes count: partial credit is not a verdict, and a claim that needs a paragraph of excuses has already answered itself.

How readers called it

1 reader called this claim. 100% said it would deliver.

1 Will deliver · 0 Won't ship

Points reward being right against the crowd. A correct call pays 100 × (1 − the share of readers who agreed with you when you placed it), with a floor of 10: side with a 90% consensus and a win pays 10, stand alone and a win pays up to 100. Wrong calls pay nothing. The price is locked in when the call is placed, so early readers are paid for judging before the crowd forms, not for copying it.

How readers split

The split line starts with this market's first snapshots — a point is recorded whenever a reader calls the claim, and at least once an hour while it stays open. Check back after the crowd has moved and the line will be here.

Verdict

No verdict yet. This claim is checked on 2026-10-21 08:00 Eastern Time (12:00 UTC) against the source of record below. When the check happens, one of three stamps lands on this page permanently: delivered if the claim came true as stated, busted if it did not, or unproven if the date passes without public evidence either way. Unproven claims pay no points and do not count against any reader's hit rate.

About this market

AI Rank Bank is an AI news front page with a memory. Every launch on the homepage carries one claim like this one, with a check date. Readers call claims with points, build a public record, and climb the leaderboard. Companies build a record too: every settled claim lands on the company's bank page, where its delivery rate is simple arithmetic that nobody gets to edit afterwards. There is no money in the market — points buy bragging rights only, which keeps the incentives pointed at being right.