That concern, he said, is why Ethereum’s lean roadmap has moved to hash-only signatures. His advice is to prefer hash-based tools and, for long-term safety, to multiply lattice key sizes by 10.
The Rebuttal
Hoskinson rejected the argument point by point:
- Tested for decades: Lattices have faced 40 years of attacks from the field’s best researchers, and today’s standards are already sized to withstand them.
- Hashes are not immune: MD5 and SHA-1 were both broken. Ethereum’s own Poseidon hash is so mathematically structured that the Ethereum Foundation funded a bounty to attack it.
- The 10x rule is “numerology”: A 20% better attack would need about 25% larger keys, not ten times larger.
“‘Structure’ names nothing,” he wrote.
The Real Stakes
Hoskinson’s biggest worry lies outside crypto. Browsers and messaging apps already use lattice-based ML-KEM to protect against hackers who steal encrypted data now to decrypt later. Casting doubt on it, he argued, could keep internet traffic on older encryption for longer, the very risk Buterin wants to avoid.
He also dismissed Buterin’s idea of sending encrypted notes off-chain through a third party. That only adds a middleman who can see users’ metadata, he said.
He questioned Buterin’s record too, saying past posts led engineers away from ideas such as Plasma and Casper, only for the industry to return to them years later.
Neither founder pointed to an actual AI-found attack. For now, the question is whether crypto should guard against a break that might come, or trust defences that have held for four decades.