For a while I kept suspecting that I was walking in place.
Every day there was motion: writing code, running, memorizing words, reading a little. It looked like progress. But when I looked back every three months, almost nothing had changed. I thought the problem was “not enough effort.” Later I read Wiener and realized the problem was not the direction of force. The loop itself had not been closed.
In Cybernetics: Or Control and Communication in the Animal and the Machine (1948), Norbert Wiener gave cybernetics its first definition: the study of shared structures of control and communication in animals and machines. The central object was not the motor, nor the nerve, but feedback: a system returning part of its output to its input in order to correct its next action.
Action without feedback is, in essence, a guess.
The contrapositive cuts deeper: if I am not doing measurement, everything I am doing is open-loop guessing.
Open Loop and Closed Loop
Engineering draws the line cleanly.
Open-loop: issue an instruction and do not inspect the result. A microwave runs for 30 seconds and stops, regardless of whether the food is warm.
Closed-loop: issue an instruction, measure the result, feed the gap back into the next instruction. An air conditioner keeps sampling the temperature until the room reaches 22 degrees.
Move that into life. My old language-learning routine was textbook open loop: memorize 50 words a day for 30 days and expect that one day I would “know it.” I never defined a measurable signal for knowing. I set no reference value, sampled no current value, and therefore had no gap to feed back. Ashby says it coldly in An Introduction to Cybernetics (1956): no measurement, no control. You think you are regulating temperature. In reality, you are only pressing buttons.
Code is the same. I used to treat “the feature is done” as the finish line. Whether it became more stable, faster, whether the user’s path got shorter - none of those signals returned to me. An engineer who ships and never looks back is an open-loop motor. It looks productive. It is drifting.
Feedback Is Not a Postmortem
In Chinese, feedback is often collapsed into “reviewing after the fact.” The translation is not precise. It carries an aftertaste: the event is over, now we sit down and reflect.
But feedback in cybernetics is in-flight. In Carver and Scheier’s model of self-regulation, a feedback system must contain three things at once:
- a reference value - the desired state;
- a sampling function - the sensor that reads current state;
- a comparator - the mechanism that keeps feeding the gap back to the executor.
Remove any one of them and the system degrades into open loop.
Count the Loops
In Brain of the Firm, Stafford Beer proposed the Viable System Model, breaking a living organization into five subsystems: operations (System 1), coordination (System 2), control (System 3), intelligence about the future (System 4), and policy (System 5).
What shook me was not the number five. It was his diagnostic logic: an organization usually dies not because nobody is working, but because a feedback loop has broken at some layer. Operations are moving, but middle control is not monitoring. Management is managing, but intelligence is not sensing external change. Intelligence exists, but policy refuses to choose. When any loop in the chain breaks, the thing may look alive from the outside while disintegrating inside.
Apply it to myself and the conclusion is ugly: most of my “failures” were not execution failures. Some layer of the loop simply was not wired. Fitness had training (S1) without long-range load sensing (S4), so injury repeated. Study had input (S1) without output verification (S3), so a year later nothing remained. Writing had action (S1) without a policy for “what kind of writer am I trying to become?” (S5), so I tried a little of every style and became none of them.
Second Order: You Are Inside the Loop
Heinz von Foerster’s 1974 distinction lifts the whole matter one level higher: first-order cybernetics studies observed systems; second-order cybernetics studies observing systems. The observer is also inside the loop.
For life, that elevation means this: when I say “I am measuring my training state,” I must also know that the tools I use, the reference values I set, and my interpretation of the gap are all designed by me. If the reference value is wrong, precise sampling only drives me precisely to the wrong place.
A well-measured open-loop system runs faster toward the wrong destination.
So second-order feedback asks the harder question: did I set this reference value because it is truly where I want to go, or because someone told me that is where I should go?
A Simple Exercise
I later gave myself a short rule. Before I begin anything I want to take seriously, I must answer three questions:
- What is the reference value? Write it in one sentence. It must be measurable, contain a number or a binary judgment, and may not be something like “get better.”
- What will I measure with? How often, with which tool, and who reads the signal.
- When the gap appears, what do I change? The input, the method, or the reference value itself - decide beforehand, not afterward.
If I cannot answer one of the three, my default action now is: do not start.
It sounds cold. But the other side of Wiener’s sentence is warm: once the loop is closed, the system begins to find the target by itself. You do not need to drag yourself there by willpower. Keep the gap visible, and action begins to arise.
I used to think growth was a matter of gritting my teeth. Now I know growth is a matter of wiring.
The remaining work is not to try harder. It is to take the loops scattered everywhere, not yet connected, and wire them back one by one.