Connor AI

Demo

About

The problem it solves ## Current DAO Execution Landscape Currently, DAOs face challenges with manual execution models that are prone to centralization and human error. Alternatively, self-executing DAOs depend on pre-defined contracts that can be rigid and vulnerable to security risks. ## The Connor AI Solution Connor AI redefines this landscape by introducing a modular, AI-driven system with several key enhancements: ### Automated Execution Connor AI streamlines the execution process of DAOs, entirely automating what was once manual, thereby diminishing human error and manipulation risks. ### Decentralization Without Compromise Connor AI's modular structure is the foundation of its commitment to True Decentralization. This system isn't just in harmony with DAO policies; it is a distributed network of autonomous agents, creating a robust framework that's resistant to a single point of failure. ### Adaptive Policy Making Designed for agility, Connor AI can swiftly manage a wide array of governance activities, from minute operational decisions to significant policy overhauls, without losing effectiveness. ### Enhanced Security Specialized AI Agents work in concert, with each one responsible for a particular aspect of the governance process—from drafting contracts to performing security audits—bolstering the overall integrity of the system. ## Future of Connor AI By taking a modular approach, we can: 1. Faciliate the development of each individual agent, by finetuning each agents differently or using different underlying LLM models for each agent 2. Integrate with ease, additional AI agents or tools like oracles, enhancing the system's capabilities and adaptability.

How it's made

Challenges we ran into ## AI It was quite challenging to make an AI system to execute the DAO's decision in a way that was both robust and aligning with the DAO's decision. We achieved this by making a system of AI agents, designed in a modular manner. Upon the DAO's approval of a new proposal, Agent Blue drafts the initial contract. This draft is then evaluated by Agent Purple to ensure it aligns with the DAO's proposal terms and by Agent Red to check the contract for any potential security issues. Agent Green refines the contract, incorporating feedback from Agents Purple and Red. This modular approach allowed the AI system to execute the DAO's decision in a robust and secure manner. ## Cross Chain Signature Our team tried to integrate NEAR's Chain Signatures to our product and have seen examples of NEAR's Chain Signatures, but we felt there was a lack of SDK that could comfortably be used by developers familiar with EVM. Therefore, we decided to develop a simple SDK compatible with Viem for the purpose of easy use of NEAR's chain signatures. Initially, the major hurdle was dealing with the payload in the "sign" function for "multichainContract.requestSignature". We had to navigate through trial and error to learn that instead of using a Uint8Array format, a number array was required and that reversing it was necessary to generate the correct signature. Additionally, we encountered slight differences between the "big_r" and "big_s" values compared to the "r" and "s" values. Crucially, the absence of a provided "v" value necessitated receiving two signatures with "yParity" values of 0 and 1, and then verifying which was the correct signature through "verifyMessage", "verifyTypedData", and "recoverTransactionAddress" before sending it. ## Frontend On the frontend side, deploying and testing across multiple blockchains introduced its own set of challenges, including issues with Sepolia's default RPC and difficulties arising from blockchain disruptions.