Release notes, tutorials, and official announcements from the AStack team
How we implemented the Recursive Language Model (RLM) from arXiv:2512.24601 in AStack's core package, using context offloading and dynamic discovery to process 100MB+ contexts without OOM errors.