No Circuit Fits All
Over the last few days, two unrelated things kept bothering me.
The first was a feature in OpenClaw.
OpenClaw has a native REM / Dreaming system. It consolidates memories by looking at what gets retrieved often, then promoting those chunks into more durable memory. For many agents, this is exactly what you want. If an agent is doing information work, frequent retrieval is a useful signal. The more often something is needed, the more likely it should become part of the agent’s long-term context.
The second was a post by Andriy Burkov about LLM-based agents and rationality.1
His argument, roughly, was that an agent must be rational, and rationality means preferring the action with higher expected utility. If LLMs cannot do that reliably, then LLM-based agents fail as general-purpose problem solvers.
Both ideas are intelligent.
Both are reasonable inside their own frame.
And both made me realize that a...
Copyright of this story solely belongs to hackernoon.com. To see the full text click HERE
