Google found a 10x cheaper way to crack crypto wallets and hid the blueprint.
Two months later, 100 strangers used AI to beat Google in under 3 days.
Then AI cracked a US government backup lock in 24 hours.
This is the AI race forcing @Starknet to make wallet locks swappable.
March 2026
- Google's quantum lab finds a way to attack Bitcoin and Ethereum wallet signatures roughly 10x cheaper
- judges the circuit details too dangerous to publish
- locks the design away and only releases mathematical proof that it exists
Late May 2026
- developers open a public challenge at ecdsa․fail daring anyone to beat Google's secret mark
- over 100 regular people enter, most working alongside AI
- they pass Google's quantum team in under 3 days and push 69.4% ahead of Google's score
July 28-29, 2026
- for 3 years, experts vet 40 future-ready locks in a US government backup competition
- only 1 finalist in its math family survives 2 elimination rounds
- Anthropic's Claude finds a hidden flaw human reviewers missed for 3 years
- its authors withdraw the lock the very next morning
Remember this part.
October - November 2026
- OpenAI releases 722 AI-written math papers disproving decades-old assumptions
- Vitalik Buterin warns the main replacement lock many crypto teams trust today uses that same math family Claude cracked, and could take serious hits from AI within 2 years
- this week, OpenAI's cyber model found and patched 948 Linux kernel flaws in 1 run for $1.2M
I think most crypto holders are watching the wrong clock.
Everyone asks when quantum computers arrive, while AI on regular computers is already breaking the replacement locks.
On older chains, your wallet address is welded to one fixed lock. If AI weakens that replacement, everyone has to move funds to a brand-new address again.
Starknet built its wallets so you can swap the lock and keep the same address and funds, with open tests already cutting new lock checks to ~$0.023.
Its engine also runs on the hash-based math Vitalik named as his alternative, securing over $1T in volume since 2018. Anthropic open-sourced that same math this week to verify 400B AI model outputs in ~17 seconds.
No blockchain is quantum-safe across every part today. Starknet isn't quantum-safe yet either, targeting 2027 with opt-in upgrades for existing wallets.
And because Starknet currently runs on top of Ethereum (which aims for December 2028), 93.52% of Starknet voting power backed researching an independent L1 chain to control its own upgrade clock.
Welding one new lock onto a wallet only works once.
Until AI breaks that lock too.
Sources: @StarkWareLtd Nov 3 report & Quantum Hub:
cc @EliBenSasson @adiiHQ