I just did something my AI agents couldn't

A developer recently shared a frustrating experience attempting to refactor an ESP-IDF project using various AI coding assistants, including Gemini and GitHub Copilot. The goal was to reorganize the project structure by moving code into a 'components' folder. Despite repeated attempts and following the AI's instructions, the developer encountered persistent build errors. After wasting days relying on automated suggestions, the developer decided to manually intervene. By investigating the CMakeLists.txt files and header include paths, they identified that the AI models were consistently providing incorrect configuration parameters for the 'REQUIRES' settings. Once the developer manually corrected these specific lines based on external documentation, the project compiled successfully. This experience highlights the limitations of current AI coding tools in handling complex, project-specific build configurations, underscoring the necessity for developers to maintain a deep understanding of their codebase rather than relying solely on automated assistance.