Abstract: Consensus protocols are the core of most blockchain systems today. They require several rounds of communication among all the validators, making it slow and expensive. Alternative constructions like Fastpay, SUI Lutris, and ABC relax the consensus requirements to finalize transactions in just 1-2 network round trips without needing any validator-to-validator communication. However, these systems have a key limitation: they do not allow for states that can be modified by more than one party at any given moment in time.

We propose extending these asynchronous networks to support applications with multi-owned states—states that multiple parties can modify. Our approach uses an asynchronous base layer that doesn’t provide any ordering and allows for composing it with custom sequencing gadgets built on top. Similar to single-writer states owned by external accounts, multi-writer states can owned by application-specific sequencing gadgets. These gadgets solely assign non-conflicting indexes to transactions that modify shared states. After receiving their indexes, transactions are settled and executed on the asynchronous base network. The system borrows security from the base layer and remains safe regardless of the ordering gadget's safety. Transfers across accounts in the base layer and that are managed by ordering gadgets can use the fast finality, unlike the typical delay of traditional L2 systems. Additionally, this architecture allows application developers to create custom ordering rules for their application state and control how they handle their application’s MEV.

Abstract: Explore how to communicate safely across Ethereum's parallel universes (forks). RIP-7789 describes a protocol for fundamentally safe or reorg protected communication across rollups on Ethereum.

Abstract: Blockchains are decentralized networks of fat servers. Rollups also operate on fat servers for sequencing and proving. The vast majority of users of these systems run thin clients that always interact through these servers, an architecture that (1) fails to leverage the local compute capability (2) rely on (decentralised) middlemen for interaction, necessitating transaction fees for middlemen incentivisation. Topology brings no-middlemen P2P into practicality: an open real-time P2P network functioning as an open unified programmable space that’s sharded by program, BFT, and participated *directly* from the end user devices. Topology enables application economics that are infeasible for existing client-server architectures, regardless if the servers are decentralised.

Abstract: L2s have historically had single chain settlement that allowed it to tap into the settlement layer's liquidity to bootstrap its ecosystem. With the rise of major chains like Bitcoin and Solana as settlement layers, we propose multi-settlement for L2s which allows an L2 to tap into multiple settlement layers simultaneously. This unlocks a new paradigm for applications to provide a CEX level UX in a trust-minimized way.

Abstract: Traditional computing follows the von Neumann architecture, where a compute/control unit interacts with read/write memory via a bus. Compute/control relies crucially on memory and access to it. This model has underpinned decades of innovation.

Web3 has traditionally developed technologies such as virtual machines (VMs), that map onto von Neumann’s compute/control framework. Data propagation (bus) and access (read/write memory), however, have emerged as present critical constraints in decentralized environments.

Unlike Web2, which benefits from powerful abstractions such as the TCP/IP socket to enable seamless end-to-end data transport in a decentralized network, Web3 lacks an equivalent mechanism for data propagation and access across decentralized nodes.

We introduce a breakthrough in Web3 infrastructure by using Optimum P2P for data propagation and decentralized random access memory (DeRAM) for data read/writes. Optimum acts as a virtualized dedicated memory layer, enabling high-speed data propagation, secure access, and real-time updates over a permissionless, decentralized, and potentially unreliable network of nodes. By leveraging Random Linear Network Coding (RLNC), Optimum ensures atomic, consistent, and durable data access, while maintaining high throughput and low latency.

Abstract: Solayer is pioneering the first hardware-accelerated SVM blockchain, InfiniSVM — an infinitely scalable, multi-execution cluster architecture powered by SDN and RDMA, delivering 100 Gbps throughput while maintaining atomic state. Our goal is to be the first blockchain to achieve 1M TPS in production, while also serving as a scaling solution for Solana. In this talk, Solayer co-founder Jason Li will discuss the limits of software scaling, the architecture of InfiniSVM, and how it will not only scale Solana but also unlock transformative use cases in web3.

Abstract: Vast majority of blockchain designs are based on the principle that the network must execute all transactions. However, such approaches have inherent limitations from both scalability and privacy standpoints. In this talk, we will describe a different approach on the example of the Miden protocol where most transactions are executed and proven by the end users and apps. This approach offers a number of interesting properties including unbounded scalability, support for privacy-preserving smart contracts, and reduced state bloat. However, it also requires reimagining of the state model to minimize the need for concurrent updates to the shared state, and a new execution model which enables concurrent transaction execution/proving by independent actors.

Abstract. Today’s secure computation often depends on trusted hardware or centralized actors, which introduce vulnerabilities and limit long-term adoption. Enclave’s Encrypted Execution Environments (E3s) offer a decentralized, cryptographic framework for securely processing encrypted data from multiple parties without exposing sensitive information. Unlike Trusted Execution Environments (TEEs), E3s are modular and can be run on any machine, leveraging a decentralized network of nodes to ensure privacy, integrity, and verifiable results. By combining programmable cryptography like Fully Homomorphic Encryption (FHE), Zero-Knowledge Proofs (ZKPs), and Distributed Threshold Cryptography (DTC), E3s overcome key limitations of TEEs while offering greater flexibility and security. This talk will demonstrate how Enclave powers privacy-preserving applications, including sealed-bid auctions, secret voting, and collaborative AI training, and how E3s create a stronger foundation for trust and privacy in confidential compute infrastructure.

Abstract. In this talk, we will introduce collaborative Noir (coNoir), new tooling that extends Noir to create coSNARKs. We'll explore its practical applications and challenges, with a focus on private proof delegation. Through benchmarks, we'll show how coNoir allows resource-constrained devices to securely and efficiently outsource proof generation.

We’ll discuss the concept of private shared state through the example of dark pools, showing how coNoir supports confidential trading by enabling secure computation over encrypted data without exposing sensitive information.

To close, we’ll highlight open questions in the area, from cryptographic research challenges to architectural considerations for implementing private shared state at scale.

Abstract. Homomorphic Encryption and other recent breakthroughs in secure computing technologies offer tremendous opportunities for secure data collaboration to enable more secure and effective data access, database operations, analytics, machine learning and other capabilities that enable a secure and transparent Web3. In this talk we'll discuss the most recent advances in secure computing technologies that enable these applications using secure privacy-preserving data collaboration. We frame our discussion in the context of both mature deployed software capabilities being deployed by Duality Technologies (https://www.dualitytech.com) and emerging new capabilities currently in prototype. Duality Technologies builds these capabilities from over a decade of technology development funded by DARPA and other S&T organizations in the DoD, and released in the open-source OpenFHE software library (https://www.openfhe.org).

Abstract. This talk will give a general overview of trusted execution environments and will describe the key features and processes that are required to create a secure TEE based app. It will go into the overlooked problem of upgradeability with TEE based apps. This includes normal code upgrades to maintain a running system and security upgrades that includes applying patches for vulnerabilities.

Abstract. Developers are unable to realize the full promise of FHE as the technology is notoriously difficult to configure as a non-expert. Accordingly, we build Parasol, a virtual FHE processor providing highly performant and compact programs. To beat state-of-the-art performance, we extract parallelism at all levels of the stack – via our novel circuit composition technique as well as via an out-of-order executing design. Circuit bootstrapping has long been neglected in the field. However, when combined in a sophisticated manner with multiplexers, we can enable homomorphic computation to be done a lot more efficiently than with competing methods. Additionally, Parasol produces programs small enough to store on mainnet today.

Abstract. "The field of cryptography has delved into multi-party computation (MPC) for over four decades, with recent years witnessing significant strides towards practical implementation. However, current MPC protocols encounter challenges when scaling to hundreds of parties, primarily due to either 1) communication overhead inherent in information-theoretic (IT) approaches, or 2) computational overhead associated with additive/fully homomorphic (public-key) encryption.

This talk showcases a novel, scalable MPC protocol and library that integrates the efficiency of IT-secure protocols with the manageable communication overhead of computationally secure MPC, all while ensuring post-quantum (PQ) security. The proposed approach leverages cryptographic hashes to implement essential commitments for MPC. Given that hash computations are 4-5 orders of magnitude faster than homomorphic cryptography, hash-based MPC offers significant scalability and cost-efficiency advantages. The talk will also demonstrate the practical applications of hash-based lightweight cryptography in various MPC primitives related to blockchain, such as random beacons, dynamic proactive secret sharing, and PQ-secure asynchronous atomic broadcast.

Abstract. Gateway Abstraction rethinks Gateways and RPC endpoints. It enables trust delegation through ring signatures, bridging user optionality with a reliable, performant, cost-effective open data service and access—paving the path from Account Abstraction to Chain Abstraction.

Abstract: WalletConnect serves as a fundamental building block in the web3 ecosystem, enabling millions of monthly connections between apps and wallets while supporting a significant portion of on-chain activity.

At the core of this functionality lies a distributed key-value store, operated collaboratively by a network of over a dozen node operators.

In this talk, we will delve into the technical challenges we faced and the insights gained while building this network, offering valuable lessons for other developers and teams working on DePIN projects.

Abstract: Ethereum’s growing reliance on ZK proofs for scalability has made proof generation speed a critical bottleneck. Realtime proving is essential for improving interoperability and addressing liquidity fragmentation across rollups. Faster ZK proofs enable atomic cross-rollup transactions and support seamless composability—paving the way for a more interconnected Ethereum ecosystem. However, achieving realtime proving at scale presents significant technical hurdles.

In this talk, I will review historical approaches to ZK hardware acceleration and their limitations. I will then explore why GPU-Native zkVMs offer a promising path toward realtime proving by maximizing GPU efficiency and utilization. Finally, I will discuss the key performance challenges that must be addressed in a GPU-Native zkVM approach, including low-latency witness generation and distributed proving.

By tackling these challenges, we can unlock a new era of scalable, high-performance ZK proof systems, enabling real-time ZK applications in blockchain and beyond.

Abstract: Decentralized proof generation is redefining scalability, cost-efficiency, and resilience in ZK systems. This talk introduces the Lagrange Prover Network’s (LPN) groundbreaking integration with ZKsync’s ZK stack prover—the first decentralized implementation for a major ZK rollup ecosystem. Unlike centralized systems, which face bottlenecks, high costs, and single points of failure, the LPN distributes proof generation across independent operators, fostering competition and increasing network robustness.

Key Takeaways:

• Scalability: Learn how decentralized proof generation handles high transaction volumes without bottlenecks.
• Cost Efficiency: Understand how a competitive marketplace reduces fees.
• Reliability: Discover how decentralization enhances system resilience and uptime.

This talk is perfect for developers, users, and ZK builders seeking scalable, cost-effective, and reliable solutions through decentralized proof generation.

Abstract. Ligetron by Ligero Inc. makes ZK proving so cheap you can run it from your own device: mobile, laptop or server without any tradeoff in speed or privacy. Most of all, it is a zkVM(WASM), so code in the language of your choice. Ligero is suitable both for privacy applications on client-side and scalable verifiable compute server side

On server-side, ligetron is pushing the boundaries of zero-knowledge technology with a cutting-edge WASM-based zkVM system designed to scale efficiently, even on low-cost hardware. Our mission is to redefine Web3 with a lightweight privacy solution powered by zkValidiums, offering a superior toolchain and advanced features tailored to modern needs. While current L2 solutions cater well to EVM-compatible DeFi protocols, the future demands more. DeFi, gaming, and enterprise applications require robust features like databases, WASM modules, and generalized off-chain computation. Re-engineering existing codebases on a case-by-case basis is often impractical. Today's zkVMs hold promise but fall short of delivering scalable solutions for these complex use cases. Ligetron bridges this gap, enabling next-generation blockchain applications in and beyond DeFi to be effortlessly deployable on our zkValidium platform.

On client-side, ligetron is challenging the boundaries of zero-knowledge technology with a cutting-edge ZK system that brings zkVM to client-side ZK proving. Imagine building a full-fledged zkApp directly from your browser—no tedious compiling, just seamless, end-to-end functionality. We're on a mission to revolutionize Web3 with a plug-and-play identity solution designed for any company to effortlessly integrate a robust identity layer. Our vision is bold: empower service providers to choose from a dynamic mix of ZK-based identity proofs, including ZKemail, ZKTLS, ZKPassport, and W3C verifiable credentials. Whether it's proving an email address or showcasing verifiable credentials, users can demonstrate their identity through a tailored subset of options chosen by the provider.

Abstract: The growing demand for ZK proofs in blockchain ecosystems has exposed inefficiencies in traditional resource allocation models. Lagrange introduces DARA (Double Auction Resource Allocation), a groundbreaking mechanism designed to optimize ZK prover networks.

DARA ensures efficient matching between proof requesters and provers, balancing costs, revenue, and fairness. Its knapsack double auction model addresses key challenges, such as truthful bidding, anti-collusion, and scaling for large decentralized markets. By prioritizing sustainability and computational efficiency, DARA enables ZK proof marketplaces to operate fairly and profitably.

Explore how DARA revolutionizes ZK prover networks with an innovative double auction model. Learn how it balances costs, boosts prover revenue, and scales zero-knowledge proofs while ensuring fairness and sustainability in decentralized marketplaces.

Abstract: AirScript introduces a human-readable language and compiler that revolutionizes how arithmetic constraints are defined for STARKs. Previously, developers had to manually encode these constraints in the native code of a single target prover—such as Winterfell—restricting flexibility and forcing redundant efforts when adapting designs across systems. AirScript addresses these challenges by enabling complex constraints to be authored in a simple, modular scripting language. Its compiler then allows generating the appropriate native code for multiple backends, with future support for additional provers on the horizon. This talk will delve into AirScript’s innovative architecture, showcasing how it enhances modularity, improves maintainability, and streamlines the development process within the zero-knowledge ecosystem.

Abstract: Acceleration with existing hardware was the first step in harnessing massive parallelism for cryptographic proving systems, prompting the adoption of hardware abstraction layers and more deliberate data layout strategies. However, these solutions remain fundamentally oriented toward AI and floating-point workloads, leaving advanced cryptography—including zero-knowledge proofs, fully homomorphic encryption, and post-quantum cryptography—underrepresented.

In this session, we preview our custom cryptographic processor—featuring a prime-field-based microarchitecture—and share insights from our tape-out and manufacturing process. We’ll also introduce our API that brings cryptography-focused hardware acceleration seamlessly into existing proving systems.

Abstract: AI agents are increasingly becoming an integral part of onchain ecosystems, enabling automated transactions and decision-making in decentralized environments. But as these agents gain autonomy, understanding their design, types, and limitations becomes essential. This talk explores the anatomy of AI agents, their interaction with onchain systems, and the unique risks they pose.

We’ll dive into how traditional AI guardrails—tools designed to moderate and align AI outputs—can inform the creation of robust constraints for crypto-enabled AI agents. Topics include appropriateness, hallucination, regulatory compliance, alignment, and validation guardrails, along with their application to onchain transactions.

The session concludes with a focus on wallet constraint design, exploring how the deterministic nature of onchain transactions allows for clearer, enforceable boundaries. As AI agents become more autonomous, designing effective constraints will be essential to ensuring their alignment with community goals and safety standards.

Abstract: This talk explains how the Lit Agent Wallet ensures that agents are truly user-governed. We’ll cover how to implement policies without sacrificing security and provide a clear path to building unstoppable agents that operate seamlessly across multiple blockchains.

Whether you’re interested in automated treasury management, DeFi strategies, or AI-based operations, you’ll learn about the foundations to create robust, trust-minimized systems that can’t be “rugged” by evil developers.

Abstract: We are building the protocol and research to scale towards decentralized AGI and superintelligence — scaling beyond centralized clusters to train frontier AGI models and autonomous AI researchers across a globally distributed network. Our approach integrates low-communication training (DiLoCo, PRIME-RL), inference-time scaling, and trillion-token synthetic data pipelines to optimize reasoning and self-improvement.

This talk covers the technical breakthroughs enabling permissionless multi-node training across heterogeneous GPUs (H100s, A100s, RTX clusters), fault-tolerant model coordination, and economic incentives for decentralized intelligence. We explore how agentic AI researchers autonomously refine models, how p2p intelligence markets unlock scalable training, and how protocol-governed AI coordination ensures collective ownership.

Decentralized AGI isn’t theoretical — the first steps towards it are happening now. We’re scaling open-weight intelligence to 100B+ state of the art models, creating autonomous AI scientists, and building the peer to peer compute and intelligence layers to make AGI universally accessible."

Abstrct: The talk covers the concept of AI agent swarms, rooted in multi-agent systems and discusses live use cases in prediction markets, DeFi and social. The Olas protocol is introduced with its novel Proof of Active Agent mechanism that allows for coordination of large scale swarms in AI agent economies. Different agent frameworks and their decentralisation properties as well as their usefulness for creating agent swarms are discussed.