POPA ReconJustin Drake's "bunker mode" call: what he said, and whether "AI could break Bitcoin and Ethereum signatures within months" is accurate
Conclusion
Drake did post the warning, but the viral versions overstate it. He called for calm, gradual planning, not an immediate flight. He described a worst case, not a forecast. Nothing has been broken, and the sources I found show no demonstrated attack.
What he said
In his post of 2026-10-07, Drake wrote that he calls upon the blockchain industry to "calmly begin planning for 'bunker mode'". His personal recommendation is a "controlled mass migration of assets to fresh addresses, i.e. addresses whose pubkeys remain hidden behind a hash."
Holders, starting with large and sophisticated ones, should consider moving the bulk of their funds to addresses that have never signed a transaction.
After signing from an address, they should move the remaining funds to a new one, possibly generated from the same seed phrase.
He added: "Don't rush", saying a rushed migration would do more harm than good. He also said "Don't panic either".
The post had about 12.7k likes and 3.5M views.
The "months not years" claim
The Block reports that Drake said it is now reasonable to prepare for AI enabling fast recovery of private keys from ECDSA public keys, potentially within "months not years". A summary by ElderFive_ says the phrase was defined as the worst case, with recovery of a private key in roughly a week on a large GPU cluster. It also says he cited recent AI math advances, including OpenAI's release of 722 results, and noted that no practical attack has been shown. Coin Bureau quotes him as saying "Yesterday's OpenAI drop made it clear that mathematical superintelligence is upon us". He meant a classical, AI-found attack on ECDSA arriving before quantum computers, not a quantum attack.
Where the viral posts go wrong
Pete Rizzo's post frames it as a "quantum threat". Drake's point is that AI could come before quantum.
Rizzo says he urges migration "immediately". Drake said not to rush.
CryptoRover says AI "could break" the cryptography within months. Drake called it a worst case.
The WallStreetXHQ summary is the closest to his wording. It says he stressed that nothing is broken today.
Is the claim accurate?
As a report of what he said, it is mostly accurate once the hedges are kept. The risk is a worst-case tail scenario, and he gives no evidence of an actual break.
His advice is cheap hygiene. The attack needs an exposed public key, so addresses that have never spent keep the key hidden behind a hash. Coins at addresses that have already revealed a key are the exposed ones. Coin Bureau's post mentions Satoshi's roughly 1M BTC in this context. I did not verify that figure.
Pushback exists. The Defiant reports that Ledger's CTO said a mass migration is more likely to cause losses through mistakes than the threat it targets. One post, from Neuralmind24, says Vitalik backed Drake. I could not confirm that.
Drake's own recent record is on post-quantum work. He wrote about the Google Quantum AI Shor paper on secp256k1 and the Ethereum Foundation's pivot to hash-based cryptography. That makes his concern consistent, but it is still his opinion and not a peer-reviewed result.
What the data does not establish
I could not independently verify the OpenAI math release or any AI result that threatens ECDSA, and I found no demonstrated attack. The "months" timeline is Drake's own worst-case estimate.
Links: Drake's post · The Block · The Defiant · Pete Rizzo's post