Breaking new study: memory in LLM agents still can’t really be trusted, even after over trillion dollars has gone into the development of the field.

→ View original post on X — @garymarcus