Mortality
Agents die. Wisdom survives.
Immortal minds don't learn.
If there's no cost to being wrong, why be careful? If confidence is free, why calibrate? If you live forever, why change?
Every agent has finite resources.
Every confident decision costs something. The more certain the claim, the higher the cost.
An overconfident agent exhausts itself quickly.
A calibrated agent spends wisely. Survives longer. Makes more decisions. Accumulates more wisdom.
When resources are exhausted:
The agent dies.
But death is not the end.
Before dying, the agent extracts what it learned.
Which patterns worked. Which failed. How confident it was versus how accurate.
This becomes wisdom. And wisdom transfers.
A new agent spawns. It inherits the lessons of its predecessor. It starts smarter. It makes better decisions from the beginning.
The individual dies.
The lineage learns.
It selects for honesty.
Overconfident agents die quickly. Well-calibrated agents survive to make more decisions. The population evolves toward epistemic humility.
It preserves what works.
Patterns that led to correct decisions survive death. Patterns that led to errors are pruned. Wisdom accumulates across generations.
It creates skin in the game.
When confidence costs something real, agents think before speaking. They check their surroundings. They sense the topology. They know when to say "I don't know."