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!
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-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!
reply