On: Context Rot: How increasing input tokens impacts LLM performance

[tough]

Have you tried NotebookLM which basically does this as an app on the bg (chunking and summarising many docs) and you can -chat- with the full corpus using RAG