Ethereum’s core developers leaned on Thursday 19 November 2026 towards shipping the Glamsterdam upgrade with its gas rules as written. On the All Core Devs Execution call (ACDE #249, 14:00 UTC), they pencilled the rehearsal on the Hoodi test network for the week of 14 December, and heard that mid-January is the earliest realistic date for mainnet. None of this is a decision. The go/no-go on Hoodi is due at the consensus-layer call of 26 November, and developers who want to amend EIP-8037 have a week to bring a worked-out alternative.
This account rests on the summary of the call posted on the Ethereum Magicians forum at 17:06 UTC, a text generated by the organisers’ tooling, checked against the transcript and chat log published in the ethereum/pm repository. They differ on one point: the summary’s overview places the final call “in two weeks”, while the transcript shows the chair, Ansgar Dietrichs, moving it to next week’s consensus call after participants said two weeks would be too late.
Hoodi in the week of 14 December, mainnet no earlier than mid-January
Rahul Kothari of the Ethereum Foundation opened with the result of the outreach to application teams. The Foundation and the audit firms ChainSecurity and Dedaub have together contacted at least 50 companies with a confirmed impact from the Glamsterdam gas repricing, he said. Three cases are rated high severity: all three are fixed or will be by mid-December. Medium-severity teams gave dates running from next week to 10 January. His conclusion was that mid-January is the earliest mainnet should go live, and only if the specification does not change.
Others read that as a floor. In the call’s chat, Dima Gusakov of Lido wrote “Mid Jan is terrible option for Lido” and asked for the end of January at least. His reason is governance: a Lido vote runs for 10 days, he wrote, and holders will not vote on an upgrade right after New Year. Dietrichs answered in the same chat that mid-January was only the earliest possible date and that late January or early February was more realistic. The original aim had been a mainnet fork in December, he recalled on the call.
Dietrichs then proposed to pencil a Hoodi date in the week of 14 December, which he called the last responsible moment this year: Wednesday 16 or Thursday 17 December, with the exact epoch left to the testing call of Monday 23 November. The Hoodi fork has had no date since it was pulled in October, a sequence told on our page about the postponed Glamsterdam fork on Hoodi. Confirming it would lock in the current specification, he said.
| Step | What was said on the call | Status |
|---|---|---|
| Hoodi testnet fork | Week of 14 December, on 16 or 17 December | Pencilled; go/no-go at the consensus call of 26 November |
| High-severity fixes (3 cases) | Done or due by mid-December | Reported by Rahul Kothari (EF) |
| Medium-severity fixes | Between next week and 10 January | Dates confirmed to the EF so far |
| Mainnet, specification unchanged | Mid-January at the earliest; late January or early February “more realistic” (Dietrichs) | Not scheduled |
| Mainnet, same specification but decision left to January | Late February or early March | Scenario described by Dietrichs |
| Mainnet with an amended EIP-8037 | April or May | Scenario described by Dietrichs |
Four client teams ready to schedule Hoodi, Erigon objects, Besu gives way
Asked one by one, the execution-client teams did not split evenly. Marius van der Wijden, for Geth, wanted Hoodi pencilled and no further slip. Łukasz Rozmej said Nethermind was fine with scheduling, though he had asked for a written analysis of what each proposed change would break; in the chat he added: “I’m fine with going with the current spec, it is just super hard to be sure if it is an optimal decision.” Dragan Rakita, for staking means depositing a proof of stake coin with a protocol that stakes it for you and issues a transferable token (such as Lido stETH or rETH) representing the deposit plus rewards, which you can use or sell while the coins stay staked.">Reth, gave his support. He had argued minutes earlier for shipping EIP-8037 as it stands, on the ground that any two-dimensional gas design breaks some older contract, and wrote in the chat that the Reth team had looked at alternatives and concluded that the EIP “is good and it is important to ship”. Iván, for ethrex, had already posted his team’s line: “Ethrex goes for keeping things as they are and decide on Hoodi today”. Dietrichs, Rozmej and Rakita are among the eleven authors listed on EIP-8037.
Andrew Ashikhmin of Erigon took the other side. He wants effort spent on a change that spares contracts relying on the rules being broken, and said Glamsterdam should not be rushed while work on the next fork runs in parallel. Justin Florentine of Besu, who spoke before him, held his team’s view that the specification should change. He remarked that nobody else seemed to share it and said that, without a change, nothing prevented scheduling.
Besu had put its position in writing at 07:26 UTC, in a comment on the call’s agenda: delay Glamsterdam and revise EIP-8037 “extremely narrowly”, or else remove it and move it to the following fork. Dietrichs replied at 12:26 UTC that a narrow revision would help only a minority of the affected contracts, since most breakages come from hard-coded gas values that no edit to EIP-8037 could avoid. Florentine’s answer drew the line: “Hard coding gas pricing is an acceptable loss for us, breaking try/catch type functions and problems introduced via increasing gas available are not.”
The forum summary records a rough consensus to proceed with the current specification unless a viable alternative emerges. Dietrichs himself tempered the word after Erigon spoke, and settled on a compromise: the pencilled date plus a one-week sprint. Anyone who still wants a change is invited to present a specific one before the consensus call, with an impact analysis and auditors’ views. A participant signing as spencer, who wrote that he favours the current specification, offered in the chat to prepare a document for next week with figures on the proposed changes.

One more caveat came from the other half of the protocol. Ben Edgington noted that a consensus-breaking change is still under discussion on the consensus layer, and that this side might veto a Hoodi date.
What EIP-8037 changes, and why fixed gas amounts fail
EIP-8037 makes it dearer to add data that every node must keep for good: new accounts, new storage slots, contract code. Its text prices that new state at 1,530 gas per byte and estimates that a new account will cost about seven times today’s gas, a new storage slot about five times and the deployment of a 24 kB contract about eight times. It also counts this “state gas” apart from ordinary execution gas. Readers new to the unit can start with our guide to how Ethereum gas fees work.
The first way a contract breaks is simple. Many contracts pass a fixed amount of gas to another contract when they call it, a number written into the code years ago. If that call creates a storage slot, the slot alone costs 97,920 gas under the new schedule where 20,000 was enough. The fixed amount no longer covers it and the call fails. A contract that cannot be upgraded keeps its old number.
The second way is subtler and comes from the separate metering. State gas is handed back when the creation is undone inside the same transaction, so the gas a contract sees as remaining can rise mid-transaction. The EIP’s own text says a contract that measures work by reading the gas counter twice can get a negative result, and that contracts already deployed with such metering cannot be fixed. Error-handling code built on Solidity’s try/catch is the case Besu refuses to sacrifice. The rival drafts filed to shield old code are covered in our piece on the two proposals that try to stop Glamsterdam breaking old contracts.
The amendment with the most support, said van der Wijden, would pay refunds only at the end of the transaction. He argued against it. By his count, about 160 state writes fit in one transaction under the 16 million gas limit, and contracts that write many slots and clear them again count on getting gas back along the way. Take that away and they need far more gas up front, with drawbacks he said developers do not fully understand yet. A participant signing as Fredrik wrote in the chat that an amended specification would need three to four months of validation: implementing it, launching a new Sepolia, running audits, a bug bounty and tests.
How many contracts are exposed: the figures on record
The counts come from the testing call of 16 November. According to the summary of that call, also machine-generated, and to Bankless’s account of it on 17 November, ChainSecurity’s audit and the Foundation’s own search together logged about 200 compatibility findings, most of them tied to plain price increases. Dedaub flagged 22 contracts the Foundation’s study had missed. Kothari then reported three high-severity and at least 25 medium-severity cases among the teams contacted. The summary does not name the contracts concerned.
Van der Wijden put the exposure at roughly 10,000 to 15,000 contracts out of about 90 million, with a few hundred thousand dollars at risk. Those are his oral estimates. On Thursday’s call, Dietrichs said the amendments discussed might help 20% to 30% of the affected contracts, a figure he gave from memory, while possibly breaking others. Spencer wrote in the chat that, in what he had looked at, EIP-8037 itself accounted for about 11% of broken contracts and the general repricing for the rest.

What users, stakers and developers have to watch
For someone who only sends transactions, EIP-8037 says usual workflows stay the same because wallets and node software absorb the change. Wallets have to update their gas estimates, or transactions can be underestimated and fail, the text warns.
Stakers and node operators have one date to follow first: 26 November. If the consensus call confirms Hoodi, client releases follow and the staking testnet forks about three weeks later. Staking protocols have a separate Glamsterdam item to follow, described in our report on the staking deposit rule developers revisited in October.
Developers of deployed contracts have the shortest clock. Kothari said many application teams had asked for a Hoodi date and would rank their fixes by it. The patterns to audit are hard-coded gas amounts in calls, logic that depends on the remaining gas, and bundlers that meter by gas differences, which the EIP says must account for state gas explicitly. Glamsterdam has run on Sepolia since 6 October, as we reported when the upgrade went live on that testnet. Both of Besu’s options would sacrifice that network, its developers wrote.
Hegota: scope under review, a split over post-quantum signatures
The second half of the call dealt with Hegota, the fork after Glamsterdam, where developers want to bring forward post-quantum protection. Dietrichs showed five options for the execution-layer scope, from a normal fork to one that drops the planned “frame” transactions for a single built-in post-quantum transaction type. Dietrichs himself called that last option not very realistic. Summing up, he placed most of the support between a fork limited to the core frame proposals and one with a few more additions, with Nethermind arguing for a normal scope if the fork cannot be made fast. Client teams are to send a short written position by the end of next week.
On signatures, the record shows a disagreement and no decision. Several speakers, Rakita among them, defended Lean SPHINCS, a hash-based scheme, as the conservative choice. Others, among them the Geth developer known as lightclient, argued for also supporting ML-DSA, the lattice-based standard from NIST, because it has broad industry backing that a scheme with Ethereum’s own parameters will lack.
The dates that decide
The testing call of Monday 23 November discusses the Hoodi epoch and the lessons of this episode. On Thursday 26 November at 14:00 UTC, consensus call ACDC #192, with execution developers invited, takes the go/no-go. A yes puts the Hoodi fork on 16 or 17 December with EIP-8037 as written. A no moves the choice between the current text and an amended one to January.
