NotebookLM is one of the most interesting AI tools out there, with little competition to speak of. While it can be used by ...
LCLMs compress LLM context before decode — 8.8x faster at 16x compression, beating every KV cache method tested. Open-sourced by NYU and Columbia.