BlockTweet

Problem Statement

This project provides a trust layer for Web3 social interactions by verifying tweets onchain. Users or dapps can input a tweet ID, and the bot checks authenticity, intent, and context. Using The Graph Substreams + Subgraphs, it indexes onchain data related to tokens, contracts, and addresses mentioned in tweets. Then it combines AI (intent detection, scam pattern analysis) with blockchain data (liquidity, holder distribution, suspicious contract flags) to deliver a trust score. The bot can auto-flag tweets like “⚠️ Possible scam token” or highlight verified community-driven announcements. This helps reduce misinformation, pump-and-dump shilling, and fake hype in the crypto ecosystem.

Solution

We built the bot as a Node.js/Express API, exposing endpoints like /verify/tweet/:id. The backend fetches tweets using the Twitter API (with caching + rate-limit handling). For onchain analysis, we integrated The Graph Substreams to extract real-time blockchain data such as liquidity pools, token holders, and contract deployments. A Subgraph is then used to make this data queryable via GraphQL. AI classification (via Gemini/Groq/GPT APIs) identifies tweet intent (announcement, shill, scam, etc.) and cross-verifies with indexed blockchain data. We also fetch additional data from Pyth Network APIs for real-time price feeds.The front-facing verification bot responds with a trust score, reasoned flags (e.g., “No liquidity,” “Suspicious contract code,” “Bot shilling”), and links to supporting onchain data. Partner technologies like The Graph allowed us to easily query complex blockchain activity, while Pyth gave us accurate real-time pricing benchmarks. We also built retry + caching logic to handle API rate limits.Hacky bit: we used a hybrid Substreams + AI pipeline, where AI only interprets intent, but the onchain verification is deterministic and immutable. This split keeps results trustworthy and avoids AI hallucination.