According to Woofun AI, Anthropic published a study on AI optimization of quantum circuits. This progress directly triggered the market to re-evaluate the security of Bitcoin and traditional financial infrastructure. The study used the Claude model to reconstruct the elliptic curve point addition quantum circuit in 53 hours, significantly reducing computational costs.
This technological breakthrough not only touched on the core defense mechanism of Bitcoin signatures, but also forced the banking industry to speed up the review of the security of its encrypted connections, marking that post-quantum migration has moved from theoretical discussion to an urgent stage of engineering practice.
In terms of technical details, Claude Mythos 5 has fundamentally optimized the divsteps algorithm through in-depth mathematical deduction, greatly reducing the consumption of the Toffoli gate. The most expensive part of the point addition operation in quantum computation is module reversal. Traditional computers use the divsteps algorithm very efficiently, but in quantum circuits, since computation must be reversible, each step of branch selection requires storing records for subsequent cleaning, which takes up a large number of qubits. Claude discovered that there was no need to explicitly store these records because the remaining pairs of numbers after each step of calculation contained enough information in themselves to reverse the previous branch path.
Although all of the rules initially tried on 256-bit numbers failed, the final solution achieved a breakthrough through the derivation of algebraic equalities and the use of Lean to formally prove the core nature of divsteps. The design consumes about 12% more qubits than the previous most qubit-saving solution, but the number of Toffoli gates was reduced to about one-seventeenth. The same automated iterative approach was applied to 7 other tasks, and one of the biggest improvements reduced the backlog of qubits multiplied by the number of gates by 39 times. As a comparison, an estimate in the field of quantum chemistry dropped from 50 trillion Toffoli gates to 340 million, and it took 8 years for multiple research teams to complete the relay, and this optimization only took 53 hours of continuous operation.
However, this technological breakthrough has clear limitations and requires a comprehensive evaluation in combination with a complete attack scheme and different hardware constraints. First, the optimization targets only a single module. The entire process from inputting the Bitcoin public key to recovering the private key also involves other calculation steps, error correction costs, and hardware architecture. Only by putting the new circuit back into the complete solution and recalculating can the final resource ratio be determined. Second, the seventeenth comparison benchmark is Luo et al.'s solution, which requires about 69.5 million Toffoli gates and 838 logic qubits, while Claude's solution requires about 4 million gates and 935 logical qubits. If opponents were replaced, the conclusion changed: Schrottenloher's plan this year only required about 1.9 million gates, but required 1,446 logical qubits. In comparison, Claude's circuit has 2.1 times the number of gates Schrottenloher's scheme, but has about 35% fewer qubits. Which is more critical, the number of doors or the number of qubits, depends on specific hardware constraints, which indicates that the new scheme provides more of an optional design point than a complete overthrow.
Furthermore, the circuit allows a failure rate of up to 2×10/1 per point addition in exchange for resource savings, and Lean proof only covers part of the nature of divsteps. The verification process relied on simulation checks from automated evaluation programs and did not run on real quantum hardware. Anthropic also pointed out that Claude's early performance was conservative, and researchers needed to guide him to let go of his mind.
According to data compiled by Woofun AI, the estimated resource requirements for cracking secp256k1 have declined exponentially over the past few years. In 2017, Microsoft (MSFT.US) researchers estimated that it would take about 2,330 logical qubits and 2.3 x 10¹ Toffoli gates. At the end of March this year, the Google (GOOGL.US) quantum AI team (collaborators include Justin Drake of the Ethereum Foundation and Dan Boneh of Stanford University) proposed two sets of circuits: one with less than 1200 logical qubits and 90 million gates, and the other with less than 1,450 logical qubits and 70 million gates, all requiring less than 500,000 physical qubits and operating time for about 9 minutes in specific scenarios. In September, IonQ (IONQ.US) further estimated that using around 20,000 ion trap qubits, one cracking can be completed in about 26 days. There was an interval of almost 9 years between 2017 and March of this year; by September, IonQ (IONQ.US)'s new estimate was only half a year away, showing that the pace of technological progress was accelerating.
Although estimates continue to decline, current quantum hardware capabilities are still limited, intersectional research between AI and quantum cryptography has become a new focus. Currently, quantum computers only have a few hundred to a few thousand qubits, and the logical qubits that have been demonstrated are between tens and nearly 100. According to Project Eleven's Q-Day award records in April of this year, the largest elliptic curve key cracked on real quantum hardware was only 15 digits in the public demo. Another paper from this year, ECDSA.fail, emphasizes the importance of open automated research. Project Eleven notes that the next round of challenges will focus on the intersection of cutting-edge AI models and quantum cryptographic analysis. Anthropic's research cannot be directly added to estimates from Google (GOOGL.US) and IonQ (IONQ.US). The former is a single module design, and the latter two are estimates of a complete set of attack resources, but both together point to a rapid decline in attack costs.
Bitcoin faces a unique migration dilemma, and the BIP proposal is progressing slowly due to its lack of a centralized decision-making mechanism. BIP-360 was merged into the official proposal library on February 11 and is still a draft, providing only one new address type for users to migrate voluntarily. The more aggressive BIP-361 was proposed by developer Jameson Lopp and others, and the idea was to eventually stop acknowledging old signatures, which sparked a property rights debate over whether old coins that remain immovable for a long time should be frozen. BIP-360 co-author Ethan Heilman estimates that, even if launched now, it would take about 7 years for Bitcoin to fully migrate to the post-quantum system. The risk exposure is huge. There are about 6.04 million bitcoins whose public keys have been disclosed, accounting for 30.2% of the issued volume. In contrast, Ethereum is more proactive: a dedicated post-quantum security team was formed in January of this year, and Vitalik Buterin (Vitalik Buterin) announced a roadmap in February, showing stronger organizational coordination capabilities.
Banks' coping strategies are relatively clear, but implementation is complicated, and market opinions are divided. NIST's draft transition timeline states that 112-bit-strong algorithms such as RSA-2048 will be deprecated after 2030, and that all public-key algorithms vulnerable to quantum attacks will be disabled after 2035. SWIFT plans to launch SWIFTNet 8.0 with post-quantum cryptography support in 2027.
However, the BankInfoSecurity review notes that most banks are unable to clearly explain their cryptography deployment. BlackRock (BLK.US) first disclosed quantum risk in Bitcoin ETF (IBIT.US) registration documents in May 2025, when the fund's net assets were approximately $64 billion. Bloomberg intelligence analyst James Seyffart cautions that such disclosures are intended to list all risks. In terms of price, Glassnode analyst James Check believes that the weak price is mainly due to long-term holders' sell-off, while Bitcoin developer Matt Corallo pointed out that if quantum risk is the main cause, Ethereum should perform better than Bitcoin. Banks can advance migration through unified management decisions on budgets and schedules, while Bitcoin fees are scattered, no one can guarantee simultaneous upgrades, and the fate of some old assets is uncertain. This is a key point in the transformation of post-quantum migration from technical feasibility to social engineering challenges after AI accelerated quantum algorithm optimization.