ever worked with long prompts that were a pain to debug? we started building a quick and dirty prototype to inspect parts of an LLM's output to know which part of the prompt led to it

here's the current version: https://inspectmyprompt.com