Fireside Chat
summary
Claims a working stack: secure enclaves running models with confidential computing (Nvidia Hopper, Intel TDX), programmable encryption via chain signatures, and an agent-to-agent protocol AITP. Benchmarks show enclave overhead at 1–10% overhead for inference. A demo agent managing $10,000 on a multi‑chain intent trading flow returns ~$4,000 (≈40%) in two months. Roadmap includes personal encrypted memory boxes, intent-based payments and legal arbitration primitives to let autonomous agents hold assets and act across chains.
AI-written summary · about
chapters (19)
- 0:00Introduction
- 0:46Overlap of Crypto and AI
- 1:51AI-Driven User Experiences
- 3:09Reliability and Trust Issues in AI
- 4:08Sovereignty and Centralization Risks
- 5:18Blockchain Solutions for AI Challenges
- 6:28Confidential Computing Advances
- 7:36Integrating AI with Blockchain
- 8:31Autonomous Agents and Decentralized Cloud
- 10:07Legal Contracts, Payments, and Real-World Integration
- 11:21Blockchain Network Evolution
- 12:39Confidential Compute and Data Privacy
- 13:31Programmable Encryption and Chain Signatures
- 14:38Personal Memory Boxes and Secure AI Models
- 15:07Intents for Payments and Contracts
- 16:46Building AI Agents: Near vs. General SDKs
- 17:44Agent Communication Protocols (AITP)
- 18:45The Future of Interfaces: AI-Driven Experiences
- 19:44Closing Remarks
related talks
19mBuild Agents with 0G + Office hours
36mSecurity for AI and Multi-Party Collaboration with Confidential Computing (Oh, and...
15mBuilding AI Agents the Web3 Way
32mProgress In the Decentralized AI Stack
44mGPU-less, Trust-less, Limit-less: Reimagining the Confidential AI Cloud
41mBuilding a Powerful Edge Using Blockchain
17mInfrastructure of AI
21mFireside chat with