Hey HN! I'm the builder here. If you're interested in the full story and more details, I also wrote about this on LinkedIn: https://www.linkedin.com/posts/char

Hey HN! I'm the builder here. If you're interested in the full story and more details, I also wrote about this on LinkedIn: https://www.linkedin.com/posts/charlie-tianle-cheng-6147a4325_opentowork-newgrad2026-activity-7424451805322940417-H2wV https://www.linkedin.com/posts/charlie-tianle-cheng-6147a432... The technical implementation treated the iteration process like training a model: test the AI's responses, measure the "loss" against what I wanted, backpropagate by adjusting prompts/RAG/CRUD, and repeat. Happy to answer any questions about the tech stack, the AI architecture, or the broader vision!