This is a really useful thread with posts from people actually using LLMs.

I wonder if Claude-100k could be used to ingest this entire thread and then answer questions based on it, or summarize or identify the pros/cons of certain aspects of Claude, large context windows, vector embeddings, etc.