Intro to AI on Arbitrum Stylus
summary
Shows a Stylus + Rust runtime on Arbitrum compiling to WASM and running a 5x5 on-chain Q-learning agent entirely inside blockchain transactions. Demo builds an ERC-721 where tokenURI returns a dynamic SVG maze, training runs called via cargo stylus deploy and Remix, and Q-table updates after 10/20/30 episodes (event emitted on training_complete). The agent learns a correct path (down/right... to goal) and the demo proves on-chain reinforcement learning works end-to-end.
AI-written summary · about
chapters (10)
- 0:00Workshop introduction
- 0:55Arbitrum Stylus introduction
- 3:04Stylus EVM and WASM runtimes
- 7:46Onchain AI agents and RL
- 17:15Demo: Rust code walkthrough
- 33:29Deploying AI agent contract
- 37:03Training AI with Q-learning
- 40:16Demo results and NFTs
- 41:39Resources and next steps
- 42:26Q&A
related talks
09mHow to Build On-chain AI Agents
30mInteractive Artistic Thinking in the Blockchain Space
02h 08mIntroduction to AI Agents
14mBuilding AI’s Infinite Game
15mBuilding AI Agents the Web3 Way
18mStablecoins, AI Agents, and the Future of On-chain Apps
32mBuilding on Ethereum
20mCross-Chain Communication: Arbitrum's Role in a Multi-Blockchain World