Justin Drake Sets Ethereum (ETH) Anti-Quantum Deadline at December 2028

Devcon 8 Sets a December 2028 Clock

Devcon 8 ran from November 3 to 6 at the Jio World Convention Centre in Mumbai, the first time Ethereum's flagship developer conference has been held in India. On the opening morning, Ethereum (ETH) Foundation researcher Justin Drake set the heaviest target of the event: the network should complete its anti-quantum and post-AI transition by December 2028, a full year ahead of the Foundation's published schedule. The announcement re-bases Ethereum's quantum roadmap around three hard forks in the 24 months after Glamsterdam, roughly one every eight months. Recent cadence makes that a steep ask. The last three upgrade intervals ran 11, 14 and 7 months, Glamsterdam itself has now gone 11 months without shipping, and the Foundation's own arithmetic puts the requirement at one fork every 7.2 months to land fork L* by December 2029; at a 12-month pace, only J* arrives on time. The roadmap draft lines the forks up in sequence: Hegotá (H*) in 2027, then I*, J*, K* and L*. Hegotá is not the quantum fork, but it introduces framework transactions that let each account pick its own signature scheme, the exit route from ECDSA. I* adds leanVM, a minimal proving virtual machine, alongside quantum-resistant lean transactions. J* brings hash-based signatures for validators plus quantum-resistant data sampling, which the Foundation labels its minimum viable quantum-resistance milestone, while K* and L* add enforced proofs and the rest of the Lean re-architecture. The formal plan has not moved: the Foundation's September priority article still commits to three-layer quantum resistance by December 2029 and states that date will not change before an external expert review in January 2027. Researchers are working to the earlier number, with 2029 as the fallback. The conference itself ran leaner, with 370 talks and 399 speakers, about 43% fewer sessions than Devcon 7 in Bangkok, after the Foundation cut 54 roles, roughly a fifth of staff, earlier this year.

EIP-8037 Breaks the Delivery Rhythm

The gap between design and delivery showed in the same week. On October 22, Glamsterdam was pulled from a second testnet, and on the conference's final day the Foundation explained why: one change to gas fees broke assumptions that deployed code depends on. Glamsterdam's two headline changes were sound on paper. Embedded proposer-builder separation pulls builder bids into the protocol and moves execution off the slot's critical path, while block-level access lists declare each block's read and write set up front so clients can run in parallel. The problem came with EIP-8037, which prices new state at 1,530 gas per byte to slow state growth. Writing a new storage slot rises from 22,100 to 110,020 gas, and deploying a 24 kB smart contract costs roughly eight times more. Costs of that size cannot fit inside a transaction's gas limit, so the excess moves into a separate state-gas reserve pool that is forwarded in full to the called contract and refunded when the call reverts. Security firm ChainSecurity flagged that error handlers which previously aborted on out-of-gas would now run to completion, bypassing common anti-griefing checks. The Foundation listed eight broken invariants, rules the protocol must always hold, and coordination call summaries counted 13 affected DeFi projects. The Hoodi testnet was pushed back with no date for it or for mainnet, and All Core Devs coordinator Nixo wrote that the repricing “will likely push Glamsterdam mainnet into 2027”. The stakes reach past one proposal: Hegotá's framework transactions list EIP-8037 as a requirement, so the stalled change sits on the critical path of the quantum roadmap itself.

Two Clocks, One Bottleneck

The urgency has two documented sources. In March, researchers from Google Quantum AI, UC Berkeley, the Ethereum (ETH) Foundation and Stanford estimated that breaking the network's elliptic curve signatures requires fewer than 1,200 logical qubits and under 500,000 physical qubits, with runtime measured in minutes, fast enough to attack a transaction while it waits in the mempool. In July, Anthropic reported that one of its models cut the key-recovery difficulty of the lattice scheme HAWK-256 from about 2^64 operations to about 2^38. The tooling itself looks close to ready: leanVM already aggregates roughly 965 signatures per second on a laptop, folding 1,500 signatures into a 265 KiB proof that verifies in 7.4 milliseconds. What Glamsterdam exposed is that the bottleneck is delivery, not math, and the question Mumbai left open is whether a protocol that could not ship one fork this autumn can ship three by the end of 2028.