A crowdsourced competition involving more than 100 human participants and AI coding agents reported an 86.1% reduction in a benchmark for a quantum circuit primitive relevant to Bitcoin’s cryptography. The score fell from approximately 10.75 billion to about 1.496 billion, although the researchers emphasized that the result does not represent an executable attack on Bitcoin.

The findings come from the ECDSA.Fail preprint, an arXiv paper posted on September 9, 2026, using a frozen benchmark cutoff of July 26, 2026, at 09:21:55 UTC. The work sits at the intersection of artificial intelligence and cryptocurrency: an open leaderboard allowed AI agents to iterate on quantum circuit designs relevant to breaking elliptic-curve cryptography.

Eigen Labs launched the competition in late May 2026 with contributors affiliated with Trail of Bits, StarkWare and the Ethereum Foundation. Over roughly eight weeks, it attracted more than 400 promoted submissions. The activity comes as institutions and companies, including NIST and Galaxy, assess post-quantum risks. Those discussions have occurred alongside broader market developments, including Bitcoin’s decline below $77,000 during the current correction.

Key points

• The ECDSA.Fail preprint reports an 86.1% reduction in a benchmark score for a reversible secp256k1 circuit primitive relevant to Shor’s algorithm.
• The paper documents the metric and the participation of AI agents, but its observational design does not isolate AI’s causal contribution.
• The result does not establish a practical attack on Bitcoin because it excludes error correction and complete fault-tolerant resource estimates.

What the 86% benchmark reduction measures

The competition optimized reversible secp256k1 mixed point addition, a primitive relevant to Shor’s elliptic-curve discrete-logarithm algorithm. According to the preprint, the score is defined as S = Q × T, where Q is the peak logical-qubit width and T is the average number of executed Toffoli gates.

Using that metric, the paper reports a reduction from approximately 10.75 billion to approximately 1.496 billion at the July 26 cutoff. The rounded 86% figure therefore describes a change in the benchmark score. It does not describe a reduction in attack time, attack cost or Bitcoin’s security margin.

Changes in the circuit benchmark

The baseline circuit used 2,715 logical qubits and 3,960,753 average executed Toffoli gates. The cutoff winner, identified by commit 8e9c9a2, used 1,151 logical qubits and 1,299,453 average executed Toffoli gates. Those two values are the multiplicands behind the reported score reduction.

Peak logical qubits: baseline → cutoff winner
2,715 → 1,151

The benchmark supplies one addend classically. A separate coherent variant compatible with windowed addition used 1,162 logical qubits and 1,684,161 average executed Toffoli gates. It recorded an empirical success probability of 0.99809 across 100,000 random inputs.

A later, post-cutoff submission reduced the reported approximate score to 1.259 billion, using 1,321 logical qubits and 952,707 average executed Toffoli gates. Those figures belong to a subsequent entry and should not be conflated with the cutoff winner’s 1,151-qubit design.

Where AI agents contributed

More than 100 participants working with AI agents produced the competition’s over 400 promoted submissions. That makes the effort an AI-agent-assisted optimization workflow rather than a purely human exercise.

However, the paper characterizes the progression as observational. Participants were not randomly assigned to human-only and agent-assisted groups, there was no common attempt budget, and the researchers did not have a complete record of prompts or wall-clock time. The study therefore does not isolate AI’s causal contribution to the 86.1% improvement.

Claims circulated in some newsletter coverage that AI independently drove the entire gain are unconfirmed and unsupported by the paper’s methodology.

The scoring process also includes a separate caveat. Some 4.8% of scored source commits fall into a verifier-search bucket, in which nonce searches can lower the sampled score without producing a semantic circuit optimization. That percentage represents the share of commits in the category; it is not a discount applied to the headline 86.1% figure.

What the result means for Bitcoin security

The authors explicitly state that the reported circuits do not constitute an executable attack or a complete fault-tolerant resource estimate. The work omits physical error correction, architecture-dependent compilation, full non-Clifford accounting and validated integration into windowed Shor arithmetic.

“This work improves a reversible point-addition primitive relevant to quantum attacks on elliptic-curve cryptography, but it does not constitute an executable attack or a complete fault-tolerant resource estimate.” — ECDSA.Fail paper authors, including Jieyi Long et al., as reported by Decrypt

A percentage reduction in one circuit primitive cannot by itself establish attack feasibility or provide a timeline for Bitcoin’s exposure to quantum attacks. The optimized metric covers only one component of the required arithmetic, while the logical-qubit counts exclude the substantial physical overhead required by fault-tolerant hardware.

That distinction separates an improved benchmark from a compromised key. No Bitcoin private key was cracked, no funds were affected, and the work does not establish an imminent Q-Day.

Evidence needed to assess the risk

Determining the practical significance of the result would require absolute resource estimates that include error-correction overhead, explicit hardware assumptions and validated integration into complete Shor arithmetic. The preprint does not claim to provide those estimates. Independent reproduction of the circuits and peer review also remain outstanding because the work has not been externally replicated.

Institutional preparation predates the paper. NIST’s IR 8547 is labeled an Initial Public Draft and was published on November 12, 2024, with a comment deadline of January 10, 2025. It outlines a planned migration away from quantum-vulnerable standards and does not establish a binding Bitcoin mandate.

Galaxy launched a Bitcoin Quantum Readiness Initiative in July 2026 with up to $5 million in developer grants, a research program and a Quantum Advisory Council. That activity represents prior preparedness work, not a demonstrated market reaction to ECDSA.Fail. Galaxy’s announcement described the initiative in the context of preparing for the Bitcoin quantum threat.

During the research window, Bitcoin traded at around $76,982 and recorded a 24-hour decline of roughly 0.96%. The broader Fear & Greed reading was 56, or “Greed.” Neither measure establishes a causal market response to the paper.

For the AI-crypto sector, the unresolved question is whether agent-assisted circuit searches can materially shorten the timeline for post-quantum migration. That issue intersects with debates over on-chain Bitcoin issuance and long-term protocol governance. Whether the reported benchmark improvement changes practical attack requirements remains unresolved pending absolute resource estimates and independent replication.

Source: https://aicryptocore.com/ai-agents-bitcoin-quantum-attack-benchmark

Disclaimer: This article is for informational purposes only and does not constitute financial or investment advice. Cryptocurrency and digital asset markets carry significant risk. Always do your own research before making decisions.