ⒿⒶ ⓋⒾ

Building SpreadHunter for the Agora Agents Hackathon by @thecanteenapp × @arc A fully autonomous agent that runs 100% in your browser. It monitors two DEXes on Arc testnet (@Xylonet_ ↔ @synthra_finance), detects price gaps between stablecoin pairs

@Jmoon_174 14 agents ? WOW

I just open sourced Agent Athena, a full stack 14 agents marketing agency. github.com/henchmarketing… 🤖 Copywriter, outreach, SEO, ads, analytics, dev, strategy. Each has a title, memory, and one job. Nothing ships without your sign-off and agent have a real layer of accountability in getting tasks complete. It runs on @OpenClaw, @paperclipai, and @claudeai. (also made it native for you @NousResearch users too, cos why not. I didn't reinvent the wheel, I just made it work better.

@Marsvoyager_ @thecanteenapp @arc @Xylonet_ @synthra_finance There is another chain? 🫡

@_Ja_V1 @thecanteenapp @arc @Xylonet_ @synthra_finance arbitrage trading on arc? good one Javi 🥂

@Jmoon_174 @thecanteenapp @arc @Xylonet_ @synthra_finance You’re right, an agent’s ability to deal with failures is probably just as important, if not more important, than the arbitrage itself 🤝

@_Ja_V1 @thecanteenapp @arc @Xylonet_ @synthra_finance autonomous arbitrage agents push from demo to production. the execution loop has to handle failures gracefully or you bleed dry.

"Spread covers fees. Liquidity sufficient. Net profit: 0.12%. Executing." Two transactions on Arc testnet. Sub-second finality. Done before a human finishes reading the first price.

Real example: the agent reads EURC/USDC on both DEXes. Xylot: 0.9637. Synthra: 0.9606. Spread: 0.32%. The agent simulates the full swap via QuoterV2, adds fees (0.3% + 0.3%), calculates real slippage from pool liquidity, and decides:

@lablabai @AIatAMD Powered by the AMD Instinct, MI300X and Qwen3-VL-32B-Instruct, the system uses a dual-agent architecture to not only detect manufacturing defects via natural language, but also autonomously diagnose the root cause of the failure.

@lablabai @AIatAMD AsBuilt Lens is an autonomous visual inspection platform that eliminates the need to train custom computer vision models.

Final submission for the AMD Developer Cloud Hackathon @lablabai @AIatAMD

When a human operator sees something wrong, they immediately think: what's missing? why did it fail? how do I fix it? That's the exact reasoning process we are replicating in AsBuilt Lens with a dual agent architecture. @AIatAMD @lablabai

The demo is already running locally with Streamlit. Next up: hooking up a live camera feed with automatic stability detection and PDF report generation for audits soon...

· The original image · The full JSON result from Agent 1 · The specification context It then returns a root cause analysis: is it physical damage? a BOM misconfiguration? an assembly omission? It even generates a Corrective Action Plan (CAPA) right there.

What makes us different: if Agent 1 (Inspector) detects a defect, Agent 2 (Quality Engineer) automatically takes over to analyze the root cause. Everything runs natively on AMD Instinct MI300X using Qwen3-VL-32B. More info soon

We’re continuing to build AsBuilt Lens for the AMD Developer Hackathon by @lablabai and @AIatAMD, a multi-agent industrial visual inspection platform. 👇