We're spending a lot of time and compute credits on the current back-and-forth of human-driven bug identification and AI refinement. Instead of manually testing and describing bugs to Greta AI, could you implement AI agents in the backend that autonomously handles the entire feedback loop: coding, testing, and deployment? What do you think about working on this feature?