0xdesigner pfp
0xdesigner
@0xdesigner
LLMs are terrible with blockchain apps. I could use some help. i'm trying to build a database that has every design everydays token and its metadata: collection name tokenID name description image url date minted ai-description** **i also need a database entry for gpt-4o to describe the image so that the tokens can be searched using more than just the title/desc i created with the token. the ai-description is generated using this system prompt: `You are a UX expert analyzing a design concept image from a web3/crypto project. Focus on identifying: 1. The core UX problem being solved 2. The proposed solution's key features 3. UI elements highlighted Keep your response to one sentence and be concise.` between the 2 collections are ~753 minted tokens between: Season 1 eth:0x5908eb01497b5d8e53c339ea0186050d487c8d0c Season 2 zora:0x5abf0c04ab7196e2bdd19313b479baebd9f7791b i can offer 1M $comicsans to for generating the fetching script and populating a db. @bountybot otherwise, any guidance is appreciated.
9 replies
21 recasts
54 reactions

Philip Sheldrake pfp
Philip Sheldrake
@sheldrake
LLMs are attuned to natural language of course. (I know you know that!) Your challenge here then exemplifies the wider challenge of interfacing LLMs and web3. The stochastic and syntactic meets the deterministic and transactional. The bridge is you and me. We bring the meaning. And that’s a good part of why we’re building /addresso
0 reply
0 recast
0 reaction