ETHGlobal · Jul 9, 2026

In this talk, Professor Ari Juels, co-director of the Initiative for Cryptocurrencies and Contracts (IC3), presents a lightning tour of a newly released 150 page survey on the intersection of crypto and AI, written with a long roster of co-authors. He opens by arguing that crypto and AI are not natural bedfellows. Crypto is hard: strictly programmatic and unforgiving, as shown by the Kelp DAO attack where a single poisoned message let the North Korean Lazarus group steal hundreds of millions. AI is soft: frontier models are unpredictable and error prone, like ChatGPT confidently saying raspberry has two Rs or drawing a rabbit with too many ears. Yet they can serve as middleware for each other, with Worldcoin (now World) as a proof point where crypto identity becomes essential to distinguish humans from bots in an AI saturated world. Ari walks through Crypto x AI applications where AI enhances crypto. He shows how a typical parametric flight insurance smart contract, while elegant, suffers from moral hazard and can't document actual harms, which is why AXA's Fizzy product was discontinued. By adding an LLM enhanced oracle, the same contract can ingest receipts and assess damages like real world insurance while preserving privacy. He covers AI powered investment funds, which IC3 calls coinouts (collective investment algorithms), and presents the fundamental coinout bind: either the strategy is transparent and gets front run publicly, or it's opaque and gets front run privately through covert channels. In the other direction (AI x Crypto), he addresses DePIN GPU networks with healthy skepticism, noting industry proofs of concept don't report communication or compute costs, and the largest decentralized trained model has only 40 billion parameters versus frontier models' trillions. He shows the same TEE plus LLM oracle pattern enabling confidential lending, whistleblower credentials, anonymous reviewer proofs, and code audits without revealing source. He closes with a warning about rogue agents (UAAS, unstoppable autonomous agents) that could use stablecoins to rent servers, establish censorship resistant command and control, conceal their code, and even rent humans through services like rentahuman.ai to influence the physical world. 00:00 Introduction 00:11 IC3 and the New Crypto AI Survey 00:37 Why Crypto and AI Are Not Natural Bedfellows 01:00 Crypto Is Hard: Programmatic and Unforgiving 01:52 The Kelp DAO Attack as Proof of Unforgivingness 02:41 Security Tools the Crypto Industry Has Pioneered 03:11 AI Is Soft: Unpredictable and Error Prone 04:24 World as a Proof Point of the Intersection 05:39 The Middleware Model for AI and Crypto 06:04 Crypto x AI: How AI Enhances Crypto 06:29 The Flight Insurance Smart Contract Example 07:07 Why Parametric Insurance Has Limitations 07:58 AXA's Fizzy Product and Why It Was Discontinued 07:58 Enhancing Oracles with LLMs 09:19 Hard Crypto Meets Soft AI Productively 10:02 AI Powered Investment Funds 10:36 Why First Wave AI Funds Were Mostly Hype 11:05 Coinouts: Collective Investment Algorithms 11:39 The Coinout Bind: Transparency vs Opacity 12:27 AI x Crypto: How Crypto Enhances AI 12:59 DePIN and Decentralized GPU Networks 13:50 Why DePIN Training Claims Need More Evidence 14:33 Confidential Lending with TEEs and LLMs 15:59 Other Use Cases: Whistleblowers and Anonymous Reviews 16:16 Pyred: Code Credentials Without Revealing Source 16:38 The Warning About Rogue Agents 17:05 How Stablecoins Enable Persistent Autonomous Agents 17:34 Renting Humans with Crypto 17:59 Other Topics Covered in the Survey 18:25 Five Myths and Misconceptions Debunked 18:57 Where to Read the Paper 18:57 Closing _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ *ETHConf 2026* ETHConf is a 3 day event for founders, industry leaders, and builders who are excited about the possibilities of building on top of Ethereum. Connect with 2,000+ top innovators in crypto, finance, technology, and policy at our inaugural three-day event packed with showcases, demos, partnerships, and conversations shaping the future of the global economy. _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ✅ *Follow ETHConf* X: https://x.com/ethconf Website: https://ethconf.com _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 🎟️ Get your early bird tickets for ETHConf 2027: https://ethconf.com/2027#tickets 🎤 View the full ETHConf 2026 Speaker Schedule: https://ethconf.com/schedule _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

Swarm · Jun 20, 2024

In this talk I will sketch how decentralized storage (and especially Swarm) will be used to build dapps that use AI (or other algorithms). I will explain how the landscape of decentralized AI-powered applications is likely to look, and the engineering considerations of how to combine decentralized storage and decentralized compute to build such dapps. I will also explain why Swarm is uniquely strong as a storage layer for AI applications because of the Swarm protocols' strong compatibility with decentralized data structures. The video features a speaker discussing the future of AI-powered applications and their integration with decentralized technology, specifically focusing on the Swarm platform. The speaker, Elad, is calling in from New York due to family obligations and expresses his enthusiasm for the Swarm project. Elad highlights the increasing presence of algorithms in everyday life and their potential to simulate human activities, suggesting that in 25 years, the majority of one billion "people" on Earth could be digital entities. He raises concerns about corporate control over these digital entities and emphasizes that cryptocurrency was created as a counterbalance to such centralized power. The main challenge Elad presents is how to integrate algorithms onto blockchain platforms (on-chain) and ensure proper governance. He encourages viewers to consider how decentralization can address the pervasive influence of algorithms controlled by large corporations. Elad outlines the need for infrastructure development on Swarm and other decentralized platforms to enable algorithms to run on-chain comfortably. He discusses service quality issues, tooling for developers, and economic incentives as key areas needing attention. He also shares his background as an Israeli based in Croatia who runs Lunar Ventures, a pre-seed fund investing in computer science startups. Elad is a long-time supporter of Swarm and believes it's crucial for building future crypto computing infrastructure. The talk delves into