The Ethereum Foundation’s Protocol cluster has officially published its first unified tier list evaluating 62 Ethereum Improvement Proposals (EIPs) proposed for inclusion in the upcoming Hegoté network upgrade. Rather than compiling isolated opinions from individual client teams, approximately 60 researchers, engineers, and domain experts across research and engineering collaborated to forge a single, cohesive perspective. While execution layer client Geth will still issue its own standalone assessment, the broader Protocol cluster integrated feedback from every participating team and contributor to establish a comprehensive data point for the ecosystem.
The priorities driving these evaluations are detailed in a companion publication titled EF Protocol: Current and Emerging Priorities, which outlines the cluster’s overarching roadmap and core commitments. The release of this unified list marks a significant shift in how upgrade scopes are debated, emphasizing rigorous peer review, shared delivery expectations, and strict necessity bars.
I. How We Tiered
The collaborative evaluation process involved contributions from nine core teams and several individual domain experts across the cluster, resulting in 16 submitted contribution templates. Out of these, 12 teams and experts provided explicit tier grades, frequently focusing their scoring on EIPs within their respective domains of deep expertise. This yielded a total of 397 individual tier grades across the 62 reviewed EIPs, averaging about 6.4 grades per proposal, with the most heavily debated changes drawing up to 9 separate assessments.
The methodology heavily incorporated hard-learned lessons from post-upgrade retrospectives following Glamsterdam. Key takeaways included the realization that EIP sizes should be measured by their integration depth, that technical complexity compounds rapidly, that the available testing surface remains the most scarce resource during development cycles, and that proposal champions frequently underestimate the true complexity of their ideas. To finalize the list, the cluster held half a dozen group calls and an intensive 90-minute working session to debate contested items before arriving at a final consensus view.
Under the scoring system, contributors graded each EIP independently using a standard scale: S-tier equaled 4 points, A-tier equaled 3 points, B-tier equaled 2 points, C-tier equaled 1 point, and D-tier or DFI (Deferred for Inclusion) equaled 0 points. Final averages accounted strictly for cast grades, meaning abstentions never penalized a proposal. Rather than relying solely on raw averages, the published tier list represents a steelmanned consensus: it originated from the quantitative grades but was refined through rigorous EIP-by-EIP debate during the working session, resulting in adjustments in both directions. Individual per-team grades were kept private to encourage candid internal feedback.
Every tier on the scoring ladder carries explicit delivery commitments for the cluster. Furthermore, the accompanying notes for each proposal follow a strict structural pattern: entries expected to ship highlight their inherent merits; items sitting on the bubble explicitly outline the specific technical requirements necessary to elevate them to a higher tier; and proposals placed below the line or formally declined detail their associated risks and concerns without embellishment.
II. The Tier List
While the complete visual representation of the rankings is hosted externally on Forkcast, the evaluated proposals span both the Consensus Layer and the Execution Layer, categorized clearly by their readiness and strategic fit for Hegoté.
Consensus Layer Highlights
Leading the consensus layer is EIP-7805, commonly known as FOCIL, which secured a unanimous S-tier grade with an average score of 4 across 7 votes. FOCIL serves as a primary headliner for the consensus layer, providing any network user with a guaranteed path to include eligible transactions without depending on centralized block builders. The cluster noted that FOCIL should ship in tandem with EIP-8369, which defines VOPS Profiles for FOCIL Eligibility, securing an A-tier average of 3.5 to establish exact transaction validation rules and extend inclusion guarantees to Frames.
Other notable consensus improvements sitting comfortably in the A-tier include EIP-8015 for removing obsolete deposit and eth1data fields with no downstream dependencies, and EIP-8365 for initiating the early retirement of BLS withdrawal credentials tied to vulnerable cryptography. Additional A-tier inclusions are EIP-8383 to reduce the consensus layer block retention window, EIP-8334 to bundle attestation propagation and cut gossip network load, and EIP-8025 to upstream optional execution proofs directly into canonical execution specifications so that zkVM development no longer relies on divergent software branches.
Meanwhile, several consensus proposals sit in the B-tier, requiring further specification before final inclusion. EIP-8146, focused on Block Access List Sidecars, has a strong case for the A-tier once its observation deadline is formally written into the specification. EIP-8237 explores independent consensus and execution layer syncing to roughly halve sync bandwidth post-ePBS, though a simpler non-EIP alternative is currently under evaluation. EIP-8198, proposing Quick Slots, garnered strong theoretical support from research and development teams but faced a D-tier rating from engineering groups concerned with cascading retuning complexities. Consequently, it requires comprehensive specifications, a working prototype, an ecosystem-wide impact assessment, and formal sign-off before A-tier consideration. EIP-8321, introducing a Hash-Chain RANDAO, was deemed sound in principle but missequenced, as it should wait for a complete post-quantum consensus design.
A large cohort of consensus proposals was designated as DFI (Deferred for Inclusion) or placed into lower tiers due to competing priorities or timing mismatches. Notably, EIP-7716 regarding anti-correlation attestation penalties was deferred, with the cluster noting that while the concept holds merit, its broader impacts on staker centralization forces and multi-client edge cases require significantly more research and community discussion before consuming scarce upgrade scope. Similarly, proposals like RowDAS (EIP-8371) sit in C-tier as balancing questions between operator overhead and available headroom, while architectural changes like Block-in-Blobs (EIP-8142) were unanimously rejected to avoid deepening long-term dependence on KZG structures.
Execution Layer Highlights
On the execution layer, EIP-8141 (Frame Transactions) emerged as a definitive S-tier headliner with a 3.89 average across 9 votes. Serving as a cornerstone for native account abstraction on security grounds, it provides a unified path to post-quantum signature schemes, enables cryptographic aggregation to properly price verification costs, and lays the groundwork for retiring legacy k1 keys. It is intended to ship alongside core components such as EIP-8250 (Keyed Nonces for Frame Transactions) and EIP-8272 (Recent Roots for Frame Transactions), both of which secured strong A-tier grades.
The execution layer A-tier also accommodates long-awaited cleanup and optimization proposals. EIP-3298 successfully secures a 3.17 average for removing gas refunds and eliminating complex metering edge cases. EIP-5920 introduces the PAY opcode, offering a low-overhead method for transferring native value without invoking recipient smart contract code. Additionally, security-focused measures like EIP-8279 and EIP-8131 establish strict worst-case cost floors for block access lists and transaction content, bounding adversarial construction scenarios.
Extension packages tied to account abstraction and contract delegation, such as Transaction Assertions via State Diff Opcode (EIP-7906), SETCODEFROM Code Reuse Instruction (EIP-8298), and Account Code Restricted ecRecover (EIP-8151), all earned solid A-tier backing as part of the broader strategy to modernize account structures and deprecate legacy authentication mechanisms. Notably, EIP-4758 for the final deactivation of SELFDESTRUCT secured an A-tier spot despite its 1.5 average, backed strongly by client engineering teams who maintain the EVM and would otherwise bear the ongoing maintenance burden of the deprecated opcode indefinitely.
Conversely, numerous execution layer changes were deferred or placed in lower tiers. Proposals such as reserving the EXTENSION opcode (EIP-8163) or introducing new control-flow instructions like EIP-7979 were deferred because they added testing surfaces and complexity without immediate consumption needs in this specific upgrade cycle. Precompile-related changes, such as EVMification initiatives (EIP-8200) and ML-DSA verification precompiles (EIP-8355), were categorized under C-tier or deferred as premature standardizations awaiting dedicated cryptographic audits and finalized specifications.
III. In Closing
Of the 62 proposals thoroughly reviewed by the Protocol cluster, exactly 2 have been classified as must-ship headliners, 15 are expected to move forward into the upgrade scope, and 28 have been formally declined with transparently stated technical justifications. Among the remaining proposals, 8 sit in the B-tier with clear requirements for potential inclusion, 7 reside in the C-tier below the implementation line, and 2 timing-sensitive proposals remain TBD while awaiting empirical mainnet data.
The cluster emphasizes that defining the success of the Hegoté upgrade relies just as heavily on the proposals it declines as on the ones it chooses to ship. To foster open dialogue, the Protocol cluster encourages community members, core developers, and EIP champions to challenge the published grades and participate in an upcoming public Reddit AMA scheduled for September 16, where the tier list will be the primary topic of discussion.

Comments (0)