I think this is cool. This has given a use case for llm at work, using them to prescreen ideas and generate FAQ material.

First time it worked great even with only a link. Second time I had it generate some comments based on a small prompt and kept telling me something went wrong, alternating between comments and that error state.

Good work!