Last week, over 100 core Ethereum contributors gathered in the Arctic Circle (Longyearbyen, Svalbard) for Soldøgn Interop. It was an intensive week of work on the Glamsterdam network upgrade.
Soldøgn continued last year Berlin Teropbut it is returned in the format used by: Amphora 🏺, Edelweiss 🏔️and star ✨: A single-track week of focused multi-client progress towards a specific upgrade, in this case the Glamsterdam enhancement.
By Friday, the group had achieved three key goals: Gas restricted floor 200M after GlamsterdamStable ePBS implementation running with external builder and final EIP-8037 pricing figures fixed. Significant progress has also been made on Hegotá features such as FOCIL and default account abstraction, as well as on many other topics.
Why Svalbard?
Svalbard is one of the few places on earth where anyone, regardless of nationality, can live and work without a visa. It is also home to the Global Seed Vault and Arctic World Archive, two cold storage facilities tunneled into the permafrost outside Longyearbyen. In between, it contains backups of crops, books, movies, manuscripts, and source code that humanity may need for thousands of years to come, including snapshots of the Ethereum source code. Lastly, the sun never sets on Svalbard from late April to August. It has 24/7 uptime, just like Ethereum, with core developers taking full advantage of it during the week!
Harden Glamsterdam, Scale Ethereum
The goal this week was to strengthen the Glamsterdam implementation and derive targets for the lower gas limit after the upgrade. Raising the gas limit safely is a multidimensional problem, and Glamsterdam addresses several issues, including how blocks are built and proposed, how much headroom is loaded on client implementations, and how the cost of state creation scales with throughput.
In practice, this meant ending the week with a stable, multi-client Glamsterdam devnet running the latest ePBS, pricing and blocking access list specifications, along with benchmarking data to lock in reliable gas limit proposals.
Most of the time was spent head-on writing code, often into the early hours of the morning, with breakout sessions to coordinate design decisions and discuss long-term roadmap items.
Three EF teams provided infrastructure during this week. EthPandaOps ethIQ and panda MCP server supports your team’s agent workflow Protocol support Establish Soldogn.xyz as your single source of truth for Interop goals, schedules, and notes. and EF Digital Studio The team captured this week on film. Stay tuned for our first ever interoperability documentary 🔜!

ePBS
In addition to pruning the proposer/constructor relationship, ePBS reorganizes slots by adding deadlines for block composition, payload disclosure, and proof. This explicitly shows how much time can be allocated to the run, which increases the headroom required to increase the gas limit.
The team started the week with a goal of 4 EL × 4 CL Glamsterdam devnet by Monday evening. The first attempt surfaced enough issues to push the goal until Tuesday, when the 4×3 configuration was running stable enough to begin stress testing.
The rest of the week was then an ePBS hardening cycle: stress testing, exposing edge cases, fixing, and iterating. The Tuesday morning Builder API breakout greatly simplified specifications related to validator registration, bid/header/commitment flows, trust model for builder payments, and circuit breaker behavior. Debugging during the week focused on cross-client edge cases, particularly around invalidating execution requests for beacon requests, where the new test suite revealed gaps in all client implementations. By Thursday morning, the CL team was reporting stable ePBS and the bidding path on the EL side was still being debugged. It was resolved from Thursday to Friday. Two questions about ACD remain controversial. Whether request signatures must be committed to the receiving builder and how to keep the 1 ETH stake builder design resilient against peer-to-peer Sybil-based active attacks.
By Friday, almost all clients were running together. glamsterdam-devnet-2 End-to-end tested using an external builder pipeline!

BAL optimization
If ePBS is the consensus layer side of the Glamsterdam scaling story, its execution layer counterpart has two main parts: gas price adjustments and block-level access lists. By providing clients with sufficient information in advance about a block’s read/write configuration, BAL supports parallel execution, batch I/O, and parallel state-root computation, all of which determine the size of a block a client can comfortably handle.
The Soldøgn BAL track ran on its own devnet, separate from the Glamsterdam ePBS chain, so optimization benchmarks were not intertwined with consensus layer stabilization efforts. Each optimization is followed by its own feature flag, allowing attention measurement tasks to be compared individually rather than as a single batch. The BAL benchmark dashboard and leaderboard surfaced the worst-case scenarios for each client across the test suite. This means that by focusing on raising the slowest routes first, the team was able to increase gas limits across the board, not just for the fastest implementation.

gas price adjustment
Glamsterdam includes various EL gas pricing adjustments to adjust costs to better match resource usage at higher throughputs. EIP-8037The increased state creation gas cost is key. This increases the price of creating new states so that higher gas limits do not result in infinite state growth.
Heading to Soldøgn, equipped with 8037 specs dynamic Pricing per state byte is tied to block gas limits, making testing combinatorially difficult (one fuzz matrix per gas limit band) and benchmarking nearly intractable. The team agreed earlier in the week to reduce dynamic pricing in favor of fixed pricing. cost_per_state_byteAny future price adjustments will be processed at the fork boundary, not within the fork.
The accounting model itself took a more iterative path. Monday’s breakout moved state gas accounting from mid-run to the end of the calling frame. A follow-up on Tuesday ended account creation fees, code deposit fees, and CREATE transaction cancellations. On Wednesday, we saw a store refund/refill exception that made us rethink our inventory. Thursday’s breakout brought accounting back to the opcode level, concluding that the real complexity lies in the reservoir model, not the accounting calculations. By Friday, the specifications had stabilized. bal-devnet-6BAL Track will provide final price adjustment figures.
This entire arc highlights one of the most important aspects of interoperability: the ability to solve complex specification, implementation, testing, debugging, and design problems in hours rather than weeks. At its best, Interop Week is a month of asynchronous progress condensed into each day!
By Friday, three threads had converged on this week’s headline figures. 200M Post Glamsterdam Gas Restricted Floor. This significant increase is possible because ePBS structures slots to provide more execution time, BAL optimizations provide clients with throughput headroom under that structure, and 8037 ensures that higher gas limits do not translate into runaway growth.

Other Glamsterdam threads
Besides ePBS, BAL and price adjustments, most of the remaining Glamsterdam coverage was hashed out throughout the breakout session.
The CL team has made the final decision on the small Glamsterdam EIP. EIP-8061 (increased exits/integration churn) included. glamsterdam-devnet-1; EIP-8080 (Exited via integration queue) Rejected for inclusion. EIP-8045 (removed validator obligations) has been reduced in scope to proposer obligations only within the preview window. and EIP-7688 (SSZ stable containers) remain within the Glamsterdam range but are stored off-site. glamsterdam-devnet-1 The team works through limited gossip message sizes for proofs against progressive lists.
Wednesday morning EL/CL synchronization architecture breakout postponed EIP-8237 At Glamsterdam, we are in favor of preserving the option for a long-term “top-up sync” architecture in future forks. On the spot, the room agreed to draft a regularizing EIP. Fork selection updated / new payload / getPayload Specifies sequencing, snap-synchronization startup handshakes, and enforces valid/invalid consistency between engine API surfaces.
Parasha has been a constant theme this week. Thursday’s session covered the Fork Selection Compliance Testing Framework. diamond Repository of reproducible CL edge case scenarios architectural statementPandaOps’ external builder test tool demonstrated mid-session a long stream of attack scenarios proposed on-site by attendees.

Beyond Glamsterdam
Several breakouts headed towards Hegotá and its subsequent fork.
Sessions intentionally agnostic to suggestions Basic account abstraction We began our work by addressing the requirements and constraints that any future design would need to meet. Feature set goals such as alternative signature schemes, aggregation, batching, recovery, gas sponsorship, flexible nonce, and key storage wallets existed along with strict constraints on common memory pool compatibility, statelessness, and L2 DoS resistance.
thursday gun room Breakout focused on implementation updates: The initial prototype was already working, and multi-client interoperability and a dedicated FOCIL devnet were the immediate next steps. Two notable design decisions were also made. Disabled FOCIL during second-generation non-finality (mirroring the proposer boost circuit breaker behavior) and adopted an index-based bookmarking approach for compatibility with frame transactions/EIP-7702.
Furthermore, long run ETH P2P track sketched a QUIC-based replacement for libp2p with native privacy and slot-aware integration, along with an erasure-coded broadcast prototype that simulates propagation up to 6x faster than GossipSub on a 2.4MB payload. The CL track has also expressed strong sentiments about not fully supporting integration, such as declaring a final fork supporting integration and then later forcing them to re-deposit after termination. This is a cleaner long-term answer to validator-set state growth.

ACD process
On Wednesday afternoon, two ACDE co-leaders, Nixo and Ansgar, led a session to gather input from key contributors to the ACD process. The session reviewed: Headliner compositionWe discussed the pros and cons of what we had. straw mapformulation EIP SFI standard. The room largely wanted to keep the headliner, but loosened its EIP vs. theme rigidity to embrace “theme + candidate EIP” as a viable pattern. Fork-specific year allocations in the straw maps beyond 2026 have been shown to be overly standardized and are likely to be relaxed. A new four-point SFI definition is presented, with ACDE/ACDC maintaining ACDT signal preparation and final call. The new prioritization process, created after the CFI decision and reflected in the meta-EIP, will replace SFI’s existing role in driving devnet inclusion, starting with Hegotá.
In terms of currency adjustments Alex Stokes He announced that he will be taking a three-month sabbatical starting next week. fly Temporarily addresses ACDC coordination Barnabas Write ACDT. All said: Nyxo and Ansga chair ACDE, fly ACDC is provisional; Mario, Barnabas, Dance Latops Cycle through ACDT adjustments.
everything else
In addition to all of the above, the team has also used their hands-on time to work on everything from better testing harnesses (condensing the Hive feedback loop from hours to minutes) to engine API plumbing improvements (gossip deduplication, batch calls, and lightweight client-centric head searches), hard tradeoffs for client diversity, and many other topics. A full list of session notes can be found at sologn.xyz.

next steps
From here, the team goes home and spends the week building prototypes and getting them ready for production. The next few weeks are expected to focus on hardening the client implementation to the new specification, finalizing test coverage, and converting Soldøgn’s PR draft into merged code.
As always, final decisions on values such as the 200 million gas cap target and final price adjustment numbers will be made and shared publicly in AllCoreDevs calls. Expect this to be a major topic next week!
Thank you so much to everyone who came to 78°N and made this week a success! Special thanks to EthPandaOps for forming the group each day and to everyone who worked under the midnight sun to help us achieve our daily goals. Etrex The crew joins for their first interoperation. It’s been an incredibly productive week. Luckily, we’ll have the entire short film ☀️ to remember it from.
