Rearranging the Hegotá hard fork priority, pausing the scaling function, and prioritizing strengthening the foundation of decentralized trust.

Written by: caspar

Translated by: Saoirse, Foresight News

Editor's note: This article is a viewpoint published by Ethlabs one week after the discussion with Ethereum core developers at Devcon. The article suggests that Ethereum is expected to achieve post-quantum security by 2027 and possess post-artificial intelligence security capabilities, which is pqai27. This plan aims to consolidate Ethereum's leading position as a globally trusted decentralized trust foundation for the economy, with ETH as the core programmable value storage medium of this system. To achieve this goal, the project team will readjust the scope of the Hegotá hard fork and prioritize safeguarding the network's long-term security resilience. Here is the content of this article:

The AI development timeline is accelerating

The capabilities of AI can be said to be at an exponential explosion stage, thus various computing systems face multiple threats that are also accelerating:

• Quantum risk timeline continues to advance: The U.S. government has adjusted the target year for addressing quantum threats from 2035 to 2030, and Google's current goal is to achieve post-quantum security by 2029.
• AI is solving many long-standing unresolved mathematical problems, while the security of cryptography itself is essentially a class of unproven mathematical problems. Unlike quantum threats, we cannot even determine if there exists a feasible breaking method. Although the mainstream view holds that the probability of such a breakthrough is very low, the risk objectively exists and the eruption time cannot be predicted.
• Cyberattack capabilities are also rapidly increasing: OpenAI used cutting-edge internal models to find and fix about 1,000 vulnerabilities in Linux systems. All software will eventually undergo the test of such automated vulnerability mining.

Ethereum not only needs post-quantum security (PQ), but also needs to possess post-artificial intelligence security (PAI). We define post-artificial intelligence security in two layers:

1. The cryptographic system used by Ethereum aims to minimize security assumptions as much as possible. No one knows where the mathematical capability limit of future AI lies, so this plan can only achieve potential for post-artificial intelligence security, rather than demonstrably absolute security.
2. Strengthening clients to resist large-scale vulnerability mining by cutting-edge models; relying on formal verification and further leveraging Ethereum's unique advantage of client diversity.

You do not need to be certain that any of the above types of threats will definitely occur. The real change is: it is becoming increasingly difficult for us to predict when these threats will arrive. Even at the current level of AI, it is accelerating the realization of all risks, and new models are being released almost every month. As long as you acknowledge the trend of continuous growth in AI capabilities, you will understand: now is the best time to prepare the network for long-term security.

Ethereum's opportunity: achieving post-quantum + post-artificial intelligence security by 2027

Bringing Ethereum's post-quantum, post-artificial intelligence security goals forward to 2027 requires adjustments to the priorities at the protocol layer and the entire community's collaborative focus. We believe it is worthwhile to invest in this goal for two reasons:

1. In an era where cyberattack capabilities are unprecedentedly strong, the extreme risk-resistance brought by pqai will become Ethereum's core trump card. Ethereum's core advantages are security and no downtime. As software attack methods become stronger and stronger, the security and continuous availability of value will be higher than ever. We believe Ethereum should consolidate this advantage, becoming a trusted foundation for ordinary users, application developers, and institutions. It should be noted that prioritizing this work does not mean stopping other developments. Glamsterdam will go live as scheduled, and its scaling capabilities will continue to be realized. Frames are on the critical path of pqai27, and in addition to adapting to post-quantum signatures, it can significantly enhance the user experience of accounts. Development work on upper layers above the core protocol will not be affected.
2. Now is the time when transformation costs are the lowest. If we wait until we deeply advance the existing fork plans before making an impromptu shift, the cost will be far greater than it is now. Moreover, these risks do not have a definite eruption time; as long as we do not finalize goals at once, every time a new model or a new study appears, there will be repeated debates on whether to advance the post-quantum upgrade. Solidifying the pqai27 goal is equivalent to setting the tone at once, allowing all participants to plan their work around this timeline.

Ethereum itself is very suitable for addressing the current situation:

Ethereum already has a clear path to achieve post-quantum and post-artificial intelligence security. The Ethereum Foundation has a dedicated post-quantum team and cryptography team, and Cambridge University has been hosting post-quantum seminars for years, accumulating a lot of related work. The planned cryptographic schemes are based on hash functions and aim to minimize reliance on security assumptions. Other cryptographic components such as signature aggregation also follow this approach. Fewer security assumptions can simultaneously withstand risks posed by quantum computers and AI; existing solutions already cover both.

Pqai27 is divided into two levels: pqai-ready is the basic target for 2027, while pqai-scaled is a challenging expansion target. The Ethereum Foundation had originally brought the post-quantum security target forward from 2029 to 2028, reserving buffer time to address the earliest possible appearance of breaking quantum computers. At the L1 seminar before the Mumbai Devcon, core developers reached a consensus: as long as this matter is set as the highest priority, achieving it by 2027 is feasible.

1. pqai-ready: The consensus layer completes the post-quantum transformation, and all user accounts possess post-quantum signature schemes; Ethereum maintains the scaling capabilities brought by Glamsterdam. If the existing cryptographic system is breached in advance, Ethereum can switch to the post-quantum scheme and continue to operate at a reduced performance level.
2. pqai-scaled: Ethereum can operate post-quantum transactions and Blob data under full load. This includes the scaling capabilities provided by Glamsterdam. This is the sprint goal for 2027.

The long-standing technical route of Ethereum will play a huge role at this moment:

• Cutting-edge attack models will find vulnerabilities in individual clients, and reinforcing all clients will take time. Client diversity means that if a single client has a vulnerability, it will not cause major failures across the network, which is a characteristic other blockchain ecosystems do not possess.
• Formal verification is rapidly becoming the standard for software development, and Ethereum has been promoting its implementation for years. The Ethereum Foundation is conducting correctness proofs for zkEVM, and formal verification is also being promoted to other modules in the protocol: EthLib, rapid confirmation rules, and next-generation consensus protocols are all undergoing formal verification.

Implementation details: replanning the Hegotá hard fork

We believe the primary decision for core developers is to confirm whether to achieve Ethereum's post-quantum + post-artificial intelligence security by 2027, and all other functions should be scheduled backward from this goal. This timeline is extremely challenging, but at Devcon, after intensive discussions among many core developers, we see feasibility.

Setting this goal means that all functions not on the critical path of pqai-ready must be reassessed, whether they are PFI, CFI, or already included in the planning of SFI. As Potuz pointed out: any non-essential function will slow down the fork progress.

On the pqai27 roadmap, several key decisions regarding the scope of Hegotá:

• FOCIL: It was originally a key feature of the Hegotá consensus layer. However, we believe its launch schedule cannot affect the feasibility of pqai27. Many of us participated in writing this EIP and even advocated for its inclusion in earlier hard forks, but there are three issues: ① it will slow down the progress of pqai-ready; ② there are still many unresolved questions regarding its pqai compatibility with EIP-8288 and the post-quantum version of Gasper consensus; ③ if FOCIL is not included in Hegotá, the execution layer can launch Frames sooner, reduce the burden on the consensus layer, and allow the consensus layer to focus on developing post-quantum consensus protocols.
• Frames + other post-quantum account-related EIPs: Frames was already planned as a key feature of Hegotá and is part of the critical path of pqai27. A lightweight version may be introduced in the future, but the overall plan remains unchanged.
• Post-quantum decoupled consensus or post-quantum Gasper: This is currently the biggest undecided issue. Modifying the existing Gasper consensus to support post-quantum would involve relatively lower difficulty, and relevant scope evaluation work has already begun. However, we are still unclear how much compromise the modification will bring; we also need to evaluate at which point it would be more cost-effective to switch directly to decoupled consensus—decoupled consensus is the long-term ideal end state, and recent progress has been made in aggregators and incentive mechanisms, but more comparative research is needed.
• Signature aggregation (EIP-8288) and post-quantum data availability pq-DA (EIP-8479): Both are essential components for achieving scaling goals. The path for migrating to post-quantum objects is clear:

1. After achieving pqai-ready: Users can migrate to post-quantum secure accounts and send post-quantum transactions; however, due to larger signature sizes, gas costs will be higher. Rollups can rely on calldata within post-quantum transactions to gain post-quantum security assurance. Rollups are not a mandatory switch to post-quantum calldata, but once classical cryptographic assumptions are breached, it can serve as a fallback solution.
2. After achieving pqai-scaled: Users sending post-quantum transactions can enjoy the same performance as regular transactions. L2 can directly read post-quantum secure Blob data.

To summarize our current rough plan for pqai27:

• Hegotá: Achieving pqai-ready Ethereum. By this step, no matter what happens, Ethereum can operate safely with performance degradation if necessary. The data availability fallback solution: the Blob itself temporarily does not possess post-quantum security, and Rollup may fall back to using calldata; the existing Blob throughput can be included in regular blocks, though the on-chain cost is higher.

1. 【Execution Layer EL】Hegotá: Frames + SPHINCS post-quantum signature scheme + post-quantum account EIP. Typically, the execution layer and the consensus layer will upgrade simultaneously in the same hard fork. If FOCIL is not included, these changes do not depend on the consensus layer; the execution layer can complete development and launch independently without waiting for post-quantum consensus.
2. 【Consensus Layer CL】Hegotá: Post-quantum consensus (modify Gasper or directly adopt decoupled consensus). The consensus layer can independently promote the development of post-quantum consensus without being constrained by the execution layer's fork timeline. This is a large project that takes longer. If the choice is made for post-quantum decoupled consensus, specifications must also be finalized; the client team can utilize this time: ① strengthen clients; ② start network layer-related work.

• Subsequent forks: Achieve pqai-scaled, full-load post-quantum operation for L1 and data availability. Post-quantum DA replaces calldata as the fallback solution, and signature aggregation allows post-quantum transactions to achieve regular performance. There is a high probability of merging the consensus layer and execution layer into a single hard fork again, including pq-DA + EIP-8288 (by then, FOCIL may have already been added).
• Supporting work not on the hard fork timeline:

1. User side: Guide users to orderly migrate to pqai secure accounts, with supporting wallets and toolchains. At the same time, leveraging the capabilities of Frames, provide users with a stronger account system. In an environment where AI attack capabilities continue to strengthen, these capabilities can fortify account security: supporting key rotation, separation of execution keys and recovery keys, fine-grained account authorization, etc., without relying on third-party tool services.
2. Validator and client teams: Allocate dedicated resources to the client teams to strengthen clients against cutting-edge vulnerability mining models; support validator pre-registration and migration to post-quantum keys, with supporting tools covering various scenarios such as institutional staking and home staking; promote more balanced client distribution to reduce the proportion of any single client, thereby reducing the risk of network-wide failures when vulnerabilities arise in clients.

Arrangement of fast slots and subsequent planning

Pqai27 is a grand goal that requires the entire Ethereum community to unite in the same direction. Therefore, we have decided to remove fast slots from Hegotá and invest all resources into pqai27. This decision has been carefully considered. Previously, we promoted fast slots because research indicated a strong community demand, and we are prepared to explain this trade-off to community members who look forward to this feature. We still have confidence in fast slots and plan to promote their launch in 2028.

You may ask what has changed our judgment compared to a few weeks ago? On one hand, the pace of AI development has exceeded previous expectations, with formerly unreachable risks rapidly approaching. In January of this year, when we planned to include FOCIL and Frames in Hegotá, the cutting-edge model was still Opus 4.5. The exponential growth is hard to predict, and the progress during this time has been dizzying; we have personally witnessed all of this.

Pqai27 is the opportunity for Ethereum to widen the gap. Our vision is to make Ethereum the trust foundation of the global economy, and the prerequisite for this is for Ethereum to become the safest and most reliable value settlement layer. Achieving pqai is as challenging as the merger upgrade from years ago. But AI is also accelerating core development: new consensus protocols complete formal verification within weeks, and automated research continues to drive iterations of cryptographic benchmarks. Achieving pqai is just the entry threshold; Ethereum still needs to continuously enhance usability on top of this. Bringing pqai forward to 2027 merely adjusts the sequence of the roadmap; we hope that by the end of 2028, Ethereum will not only have completed post-quantum and post-artificial intelligence transformations but also be faster and stronger against censorship, providing a better experience for a growing user base.

By the way: ETH will become a value storage with pqai security capabilities.