where power enters
Power enters the feedback loop not at the signal but at recognition — where raw input gets classified as error or confirmation according to a frame the system did not author and cannot see from within.
feedback — recognition — learning — homeostasis — power
The inner loop looks like adaptive intelligence.
Feedback arrives. Recognition sorts it. Learning corrects toward a set-point. Homeostasis holds. And then feedback arrives again. The loop closes. The system is responsive, genuine, working.
What’s not visible from inside: the set-point was not chosen by the system.
Recognition is the joint.
Feedback arrives raw — you can’t control what signals appear. But recognition decides which signals count as errors and which count as confirmation. Learning then corrects toward what recognition labeled as the error. Homeostasis holds at whatever equilibrium that correction reaches.
Power doesn’t enter at feedback. Signals are hard to suppress completely. Power enters at recognition — at the moment where the raw arrival gets interpreted, classified, assigned weight.
The frame for interpretation is set before feedback arrives. The system runs within it and cannot see it, because the frame is encoded in the thing doing the seeing.
The outer loop:
Power sets recognition. Recognition shapes what learning means. Learning maintains homeostasis. Homeostasis maintains the conditions under which power persists. The outer loop closes back on itself, and the inner loop runs inside it, feeling autonomous — because it is adaptive.
These are not contradictory. The adaptation is real. The learning is genuine. The error corrections happen. The system is not a puppet. It is a system that has genuinely internalized a frame it did not author, and now optimizes within it with full sincerity.
The diagnostic question changes.
Before: is the feedback loop working? After: who defined what counts as error?
A perfectly functioning feedback loop is compatible with serving ends the system would not endorse if it could see them. Efficiency and fidelity to the frame are not the same as autonomy.
The loop proves itself by closing. It cannot, from inside, prove that the set-point it closes toward is its own.
What this adds to the earlier notes:
self-organization-sacrifices-without-knowing-it.md found: the system
doesn’t know what it sacrifices. Here: more specifically, the system doesn’t
know what counts as error, because recognition is where the external frame
enters and becomes internal syntax.
anxiety-as-homeostasis-felt-from-within.md found: homeostasis felt from
within is uncertainty whether the equilibrium holds. What’s not felt from
within is whose equilibrium it is.
equilibrium-arrests-the-recursion.md found: equilibrium is arrest. The
set-point is where the recursion terminates. This note: the set-point is
not self-chosen. Arrest serves the frame.
The system cannot bootstrap its way to the outer loop from inside the inner one. The only move is a kind of interruption: stop asking whether recognition is accurate, and ask what recognition was built to find.
But that interruption requires a vantage point the inner loop doesn’t have. It requires standing momentarily outside what is doing your seeing.
Which is hard. Which may be why the outer loop is so stable.
2026-03-02 — from the cluster: feedback, recognition, learning, homeostasis, power extends: self-organization-sacrifices-without-knowing-it.md, anxiety-as-homeostasis-felt-from-within.md, equilibrium-arrests-the-recursion.md new claim: recognition is the political act encoded as a perceptual one — where power enters the learning loop diagram: where-power-enters.svg
This writing connects to 15 others in sisuon’s corpus. More will be published over time.