High-level overview of Ethereum EIPs related to staking. It will cover key EIPs, including the Pectra hardfork, EIP-4444,  Verkle Trees and ePBS.

Pectra will be a major feature fork, including EOF, Account Abstraction (EIP-7702), PeerDAS, and other enhancements like the BLS precompile. However, in this talk, I focused on staking-related improvements. Pectra includes three staking-related EIPs: 6110, 7002, and 7251. Thanks to these EIPs:
‍
- Deposit processing will be much faster (from 12 hours to 13-19 minutes). (EIP -6110)
- Funds owners (withdrawal key holders) will be able to exit without a pre-signed exit. (EIP-7002)
- Stakers will be able to stake with more flexible increments. (EIP-7251)
- Consolidation of validators will reduce networking requirements. (EIP-7251)

Verkle will allow for stateless clients.

EIP-4444 will significantly reduce disk requirements for node runners.

ePBS and inclusion lists will enhance Ethereum's decentralization and censorship resistance.

The combination of EIP-4444, Verkle, and ePBS will either reduce hardware requirements for stakers or increase Ethereum's throughput.

In this talk, we describe Monomer’s:
- Purpose and Motivation
- Design and Architecture
- Use Case (with an example)
- Roadmap

The purpose of Monomer is to free developers from the a common trade-off: technical sovereignty versus lack security and desirable tokenomics. By combining the Cosmos SDK and OP Stack, developers can achieve the best of both worlds.

Monomer is built as a compatibility layer between the OP Stack and Cosmos SDK, translating Engine API requests into ABCI requests.

It is easy to integrate Monomer into a Cosmos SDK application; it only requires importing one module.

Monomer’s roadmap is focused on adding foundational features (withdrawals, forking) and developer experience benefits (externalized app support)

Presenting recent work by Nethermind Research and Nethermind Core to allow validator to post client data anonymously after receiving a grant from the Ethereum Foundation to study different possible approaches to collect this data.

In this presentation we start by showing how client data is currently being collected and how unreliable it is.

Then we move to explain the goals of the research which are:
- Allow validators to post data anonymously
- Maintain high accuracy
- Require minimal changes to the validator’s code that does not affect consensus

We then go to explain the three approaches we reached and elaborate on their advantages and disadvantages:
- Method 1: Client diversity data on the graffiti field
- Method 2: Share data through gossip network + use nullifiers for privacy
- Method 3: Dedicated voting scheme for client data collection

Eric’s talk covers the topic of mutation testing (a technique used to test software test correctness based on code mutation), with a focus on how it can be used within smart contract development.

The beginning of the talk focuses on the history and necessity of software testing, exploring different kinds and their purposes. It then goes into the evaluation of test quality, explaining how coverage (a commonly used metric) can be misleading, and mutation testing can be used in addition for a proper evaluation.

The second part of the talk focuses on mutation testing applied to solidity smart contracts specifically, going through some tools and techniques, and guiding the audience through a simple example of how mutation testing can be used to improve the quality and correctness of smart contract tests.

Tomasz Stanczak discussed the emerging issues and opportunities surrounding minor extractable value (MEV) on Starknet. He emphasized the need for proactive engagement with MEV to ensure that as the Starknet ecosystem grows, both developers and users can benefit from better price discovery and reduced malicious activities.

Tomasz explained that MEV represents free money available for privileged actors, like sequencers, to extract. While not all MEV extraction is malicious, there is a risk to the ecosystem if these actors misuse their positions. He outlined different types of MEV activities, such as front-running, back-running, and sandwiching, using real-world analogies.

He stressed the importance of addressing MEV on both protocol and smart contract levels to prevent negative impacts on the ecosystem. As Starknet’s DeFi ecosystem grows, there will be an increased focus on price discovery and MEV-related optimizations. Stanczak encouraged developers to start experimenting with MEV extraction on Starknet now, before larger players enter the space, to better understand the protocol and improve their skills in Cairo.

The talk concluded with an invitation to join ongoing initiatives and discussions on MEV in the Starknet community, emphasizing the importance of developing tools, content, and solutions to manage MEV effectively as the ecosystem evolves.

In this workshop, Rodrigo Pino delves into the Cairo Virtual Machine (Cairo VM), the main execution engine of Cairo, emphasizing its role in optimizing the execution of Cairo code. He explains the VM's purpose beyond just executing code, highlighting its integration within a broader ecosystem that includes the compiler, verifier, prover, and runner.

The discussion focuses on the architecture of the Cairo VM, which uses a simple set of registers and instructions designed to make proving computations cheaper and more efficient. Key components like the program counter, allocation pointer, and frame pointer are explained, along with their roles in tracking execution.

Rodrigo also introduces non-deterministic memory and built-ins, which address the limitations of the VM's simplified instruction set. Built-ins, such as range checks and Keccak hashing, allow pre-compiled execution units to handle more complex tasks with lower computational costs.

Throughout the workshop, Rodrigo emphasizes the unique design choices of the Cairo VM, including its use of non-deterministic read-only memory and segment-based memory management. He concludes by encouraging attendees to explore the Cairo VM further through available resources and practical implementations in different programming languages.

The panel discussion on formal verification in the blockchain space featured experts from various companies, including:
‍
- Julian Sutherland (Nethermind): Focused on verifying complex mathematical invariants, particularly in ZK circuits and on-chain verifiers.
- Alexander Hicks (Ethereum Foundation): Emphasized using formal verification to ensure code trustworthiness, especially in performance-critical areas like ZK circuits.
- Kostas Ferles (Veridise): Discussed the use of static analyzers and formal verification tools in the auditing process.
- Mooly Sagiv (Tel Aviv University and CEO of Certora): Highlighted making formal verification tools accessible to Solidity developers and ensuring the correctness of both code and design.

Key topics included the importance of formal verification in ensuring the security and correctness of blockchain protocols, particularly in complex systems like ZK rollups. The panelists debated the role of experts in writing formal specifications, the need for educational resources to lower entry barriers, and the potential for incentivizing formal verification through community-driven efforts and reward systems.

The discussion underscored that while formal methods are invaluable for ensuring security, they should be integrated early in the development process and combined with other security practices like auditing and fuzzing. The challenge lies in making these tools more accessible and practical for a broader audience, which is crucial for the widespread adoption of formal verification in the blockchain industry.

Albert presented on folding in cryptographic protocols, specifically focusing on hash-based SNARKs. The talk covered the basics of folding, which is a method to reduce the complexity of proofs by combining multiple proof instances into a single one. Albert introduced a framework for folding in the context of hash-based SNARKs, emphasizing the challenges and solutions in implementing this with non-homomorphic commitment schemes like Merkle trees. He discussed the Nova folding scheme, its reliance on homomorphic commitments, and how they adapted lattice-based commitments for their work. The adaptation aimed to improve efficiency while maintaining security, resulting in significant performance improvements. The talk concluded with a discussion on benchmarks and the practical implications of their approach. The Q&A session touched on why they chose Nova over Hypernova and the specifics of their approach.

The panel, moderated by Oisín Kyne, focused on security concerns related to node operation, private key management, and smart contract security in blockchain systems. The discussion emphasized the importance of securing nodes, private keys, and smart contracts, especially in the context of proof-of-stake systems where security is critical to prevent major issues like slashing events.

Panelists discussed potential threats, such as operator failures, zero-day exploits, and the risk of stolen private keys. They highlighted the dangers of key compromise and the importance of multiple layers of security, including distributed validator technology (DVT), remote signers, and proper key management. The conversation also touched on the risks posed by hackers, particularly in light of sophisticated attacks from groups like North Korea’s Lazarus Group.

The panelists agreed on the necessity of diversifying clients, having robust security processes, and maintaining strong operational security (OpSec). They emphasized that technology alone isn't enough—strong processes and human vigilance are crucial to prevent security breaches. The discussion closed with a reminder that the most significant threats often arise from simple human errors, such as phishing attacks, and that constant vigilance is required to stay secure in the blockchain space.

Albert Garreta introduces several prizes, focusing on integrating ZK (Zero-Knowledge) technology with machine learning (ZKML). The ZKML prize invites participants to incorporate ZK technology into machine learning projects, broadly including zero-knowledge proofs and succinctly verifiable computation.

Key applications include proof of correct inference, where a model signed by a trusted authority produces an output that can be efficiently and succinctly verified, and proof of training, which involves creating a ZK proof that a model was correctly trained on a specific dataset. This area is noted as more complex and resource-intensive than inference. Albert also suggests exploring Federated Learning, where multiple parties with private datasets collaborate on machine learning models without sharing sensitive data.

The workshop emphasizes creativity and innovation in tackling these challenges, with a focus on the value of original ideas. Other prizes include security-driven development, large-scale computation on L2 rollups, and pre-confirmations in innovation. Participants are encouraged to speak with the Nethermind team for more details, and Albert notes that Nethermind is hiring for full-time positions and internships.

Jinsuk from Nethermind discusses the importance of plurality in large language models (LLMs). He explains how AI, particularly in platforms like Amazon and Google, significantly influences user choices through machine learning models that align with corporate incentives. He argues that the future should see user-owned AI agents that negotiate against corporate AI, promoting fairer outcomes and user autonomy. The talk emphasizes the need for community-enforced incentive alignment to ensure that AI development serves public welfare rather than corporate interests.