The author shares their experience setting up a Claude instance for DeFi research, highlighting the efficiency and effectiveness of the process. In just over an hour, they created the instance, programmed it to prioritize adversarial analysis, and ran a test on the ThGold project. This analysis uncovered several significant red flags, including allegations of fraud against a key stakeholder. The findings surpassed those of prior research, underscoring the potential of using AI tools for deeper scrutiny of DeFi protocols. The setup involves a detailed hierarchy of information that the Claude instance uses to evaluate projects, emphasizing the importance of on-chain data over official claims. The author expanded their research to include ThBILL and later projects like ETHStrat, which was flagged as high risk. In a contrasting evaluation, the bot assessed Liquity as “Conclusively Positive.” To facilitate replication, the article provides essential files and a step-by-step guide, along with a toolkit that includes access to DeFiLlama data for verification of various project claims, such as total value locked (TVL) and financial sustainability. Key tools in this research setup include on-chain verification through DeFiLlama, third-party intelligence gathering, and GitHub analysis. The author stresses the necessity of thorough research, warning against superficial investigations that can mislead users about a project's reliability. This approach aims to empower users with the capability to conduct professional-grade adversarial analysis on DeFi projects with minimal effort.