J Hackworth
@jhackworth
1/ Introducing AWESOM-O: A Farcaster agent that generates your own personalized Farcaster Wrapped. Built this because I believe we're just scratching the surface of what's possible with agents in crypto and wanted to get my hands dirty. Some thoughts 🧵
15 replies
121 recasts
234 reactions
J Hackworth
@jhackworth
2/ Onchain agents will revolutionize crypto as we know it. They’ll streamline UX while enabling entirely new applications. As a data scientist, I knew I had to build an agent for myself.
1 reply
0 recast
14 reactions
J Hackworth
@jhackworth
3/ Why Farcaster over Twitter? Twitter is a walled garden, Farcaster is an open protocol. While Twitter has distribution, Farcaster has what agents truly need: open data, native composability, and permissionless integrations. When agents can tap into everything in the network, they unlock possibilities that centralized platforms can't match
1 reply
4 recasts
13 reactions
J Hackworth
@jhackworth
4/ How does AWESOM-O work? Just @ the agent and it: -Processes your request -Queries network-wide data via Dune -Extracts personalized insights -Generates your custom-wrapped By querying Dune directly, AWESOM-O can analyze a year of network-wide data - impossible for agents on closed platforms like X. Also have to shoutout @hyperbolic for being my go-to inference on this
2 replies
1 recast
17 reactions
J Hackworth
@jhackworth
5/ After building agents with LLMs, here's the reality with today’s models: They excel at pattern detection and content generation but lack precision without training data. They'll generate AI slop forever, but can't judge quality. In the case of AWESOM-O the risk is low to generate a Farcaster Wrapped, but I wouldn’t trust it with my own money
1 reply
0 recast
12 reactions
J Hackworth
@jhackworth
6/ Until models improve, agents need humans at key decision points. For AWESOM-O, I would use Farcaster likes as an objective function - teaching it to understand content quality through user feedback Just like Botto, where AI creates, humans curate, and the system learns from human taste to improve
1 reply
0 recast
11 reactions
J Hackworth
@jhackworth
7/ After building an agent, here are the biggest opportunities I see: -Human-AI Collaboration: Applications that combine human judgment with AI automation -Better Data Integration: Richer data sources for agents to leverage -Crypto-Native Tools: Models and infrastructure built specifically for crypto use cases -Developer Experience: Expanding frameworks like Eliza and GAME to simplify building -Smart Contract Integration: Making it easier for agents to interact with onchain protocols -Feedback Systems: Building loops that help agents develop taste and improve over time
2 replies
1 recast
16 reactions
Jialin 🎩
@jialin
it's really a great idea for expanding frameworks like Eliza and GAME to simplify building 👍
0 reply
0 recast
0 reaction
