A week of AI coding cut a quantum-safe bitcoin transaction estimate from $320 to $66
StarkWare, which conducted the research, says contest participants using AI tools made the computation about five times faster, though the $66 figure has not yet been demonstrated in another transacti...
A week of AI-assisted coding has brought the estimated GPU cost of preparing a quantum-resistant bitcoin transaction down to $66 from about $320, according to StarkWare, the company behind the research.
The earlier figure came from a transaction mined on Bitcoin in August. Preparing it took roughly 3,100 hours of computing time across a fleet of about 100 graphics processors. That $320 was the cost of the computing work done before the transaction reached Bitcoin, separate from the fee paid to put it on the blockchain.
StarkWare then opened the code to a competition. Participants, many working with AI coding tools, tried to make it search through possible solutions faster. The leaderboard shows a leading submission checking about 881 million candidates a second, against roughly 146 million for the starting code on the same benchmark graphics processor.
The work addresses a possible future threat to Bitcoin. A sufficiently powerful quantum computer could use a wallet’s exposed public key to work out its private key and steal the coins. StarkWare’s method adds a form of protection based on hashes, which are expected to withstand that attack. It fits within Bitcoin’s existing rules, so trying it did not require the network to approve an upgrade.
As such, the method is an emergency option for moving eligible coins if the quantum threat arrives before Bitcoin adopts a broader fix. Using it widely would be expensive at $320 of computing per transaction — and a lower bill could make the option more practical.
The $66 figure remains an estimate drawn from a test of the computation. The improved code has not been shown preparing another transaction mined on Bitcoin.
Applying the speedups displayed on the contest site to its listed $320 cost breakdown yields about $83, according to CoinDesk calculations. The site says later record-setting runs go beyond those measurements, without showing which results bring its transaction estimate to $66.
Meanwhile, the method also has limits beyond cost. These transactions have to be sent directly to a miner because they do not travel through the network in the usual way, and they do not protect coins whose public keys are already exposed, which is the group a quantum attacker would reach first.
As stablecoins move into regulated finance, APAC is becoming a key proving ground. This report maps the region’s rules, use cases, and RLUSD’s role.
Asanat Analysis — Why it matters
StarkWare's research reveals a 79% reduction in quantum-safe Bitcoin transaction costs through AI-assisted optimization—dropping from $320 to $66 in a week-long coding sprint. This matters because post-quantum cryptography adoption remains one of Bitcoin's longest-standing infrastructure challenges. The cost barrier has historically deterred implementation; even at $66, post-quantum transactions are significantly more expensive than standard ones, but the trajectory suggests feasibility curves are compressing faster than previously modeled.
The AI acceleration signals two second-order effects: first, that quantum-readiness may become economically viable sooner than consensus estimates predicted, potentially shifting timeline expectations for Bitcoin's cryptographic migration. Second, it demonstrates how AI-driven optimization can rapidly solve cost problems in cryptographic primitives—a pattern likely to repeat across other chains facing similar scalability constraints. However, the $66 figure remains unvalidated on mainnet, and the contest format may not reflect real-world engineering constraints, warranting skepticism until independently replicated.