Re:The Components of Agent Memory
by Otterly99 on Monday November 09, 2026 @09:12AM
(#50425898)
I agree that it is a bit disapointing that one of the biggest AI lab is being so conservative with their research.
One of the biggest flaw of most LLM-based piece of software is that rules and behaviors are "soft-coded" by just putting it into context. I would love to see some actual effort to create "memory modules" in the form of weight tuning / adpaters.
One of the biggest flaw of most LLM-based piece of software is that rules and behaviors are "soft-coded" by just putting it into context. I would love to see some actual effort to create "memory modules" in the form of weight tuning / adpaters.