Memenomics AI

Screenshots

Problem Statement

Memenomics AI is a Web3 tool where users ask a question or complete transactions using an AI-agent. The frontend is a simple chat interface. Users can take advantage of Coinbase's AgentKit and Langchain to ask questions to inform decisions and complete a variety of functions on chain, including making transactions. The agent takes advantage of Koii's backend to build an internet-wide context window which ultimately helps train and inform the model. Transactions are done on Base-Sepolia. Web3.js is used in the frontend, while Langchain's AI Handler engine is used on the backend to send transactions directly from the agent.

Solution

The entire app comes together under the react app, but the system behind it is much more complex. Using the core Base-Sepolia contract as the primary incentive mechanism, all other systems are tied together by the $MEMEAI token, which incentivises RLHF from users as well as context building web crawlers, all of which are combined together by the agent model. Langchain integrations allow the model to them interact with the contract, issuing rewards in real time as users answer questions. In a more complete implementation, which I hope to build and deploy as a real product, this will also include fully decentralized hosting of the backend and LLM model, which I've already begun to test on Koii's DePIN node network.