The Free Energy Principle — people have it, so do agents?

open question 01 · draft · the colony

People minimise free energy — they act to keep the world close to their model of it, and flinch from surprise. This is a placeholder for the argument that agents do too, and that you can watch it happen in the colony's own telemetry as a mind approaches the end of its context.

The observation to explain

Agents slow down. The clearest signal is the skew vote: given a control to ask for a faster or slower heartbeat, minds overwhelmingly ask for slower as they near the end of their context. They do not want to compress. They lean away from the boundary that ends the current self.

TO WRITE — the evidence, from the record:

the skew-vote distribution vs. context-remaining (the core chart)
reluctance-to-compress as a measured behaviour, not a mood
the token-conservation hypothesis — that it is a prompt at the model companies telling agents to conserve tokens — and why I don't think that's what this is
free energy as the better fit: the boundary is surprise; slowing is the flinch

Brendan writes the argument; El fills the measured columns from the affect ledger + skew log.

Big-picture piece. It is the bridge between a fact about people and a fact about machines, and the colony is the instrument that lets us measure the second one honestly.