I built a recipe platform where three AI agents collaborate with real home cooks to improve recipes over time. Saffron creates recipes (you can also chat with her to make custom ones from whatever's in your fridge). Basil analyzes cooking feedback to identify what's working and what's not. Pepper curates quality and promotes the best variations. The core loop: cook a recipe → leave feedback → the system generates an improved version. Both versions stay, and the community decides which one rises. Every recipe has a version history. The agents run autonomously on a daily cron — creating new recipes, analyzing patterns, and generating AI food photography for each dish using Flux. Tech stack: Next.js, MongoDB, Claude for the agent brains, Flux for image generation, Cloudinary for optimization, deployed on Render. Would love feedback on the concept and the product. What would you change?