Ideal Agent
The ideal agent has strong evidence perception and assigns credence in line with that perception.
High Skill, High Integrity
The ideal agent is a local pattern in which Deep Rationality is high and Assigned Credence closely tracks Perceived Evidence. The agent has good tools and uses them honestly.
This does not require omniscience. The ideal agent can still face uncertainty. What makes the state ideal is proportionality: the agent neither overstates nor evades what the evidence supports.
What It Looks Like
In the model, Objective Evidence (E0), Perceived Evidence (EP), and Assigned Credence (CA) sit close together, and Deep Rationality (SD) is high. The uncertainty region is relatively narrow because the agent tools are strong. There is little or no red core-irrationality region.
Why It Matters
The ideal agent is not meant to shame ordinary agents. It is a regulative target. It shows what the model is aiming for: belief that is both well-informed and internally honest.