The four parts of a working loop
Before the diagnosis, the machine. A control loop has four components, and it is worth seeing them wired together before reading about how they break.
The model this borrows from
Charles Carver and Michael Scheier spent decades arguing that human self-regulation runs on the same logic as a thermostat — not as a metaphor, but structurally. Carver’s position is that the “logical elements that underlie homeostasis underlie any attempt to attain a desired goal.” The loop compares where you are against where you meant to be, and acts on the difference.
They compare these perceptions with salient reference values — whatever goals are temporarily being used to guide behavior. Carver & Scheier, on the comparator
It is a deliberately unromantic account of ambition, and that is what makes it useful. It moves the question from why am I not disciplined enough to which part of this loop is not running — and the second question has answers you can actually build for.
Fig. 2 — walk a stalled goal down this
Pick something you meant to have moved by now and answer four questions about it honestly. The first no is where your loop is breaking — and it is almost never the one people assume, which is the last.
- One real ambition
-
Check 01 · reference value
Can you name the very next step, with a date on it?
NO →Stalls — there is no set pointThe ambition never became something a step could attach to, so nothing can be measured against it and nothing begins.
Closed by — describing it in plain words and getting back a dated plan, every step owned and classified.
-
Check 02 · input
Could you say, right now, exactly where it stands — without guessing?
NO →Stalls — no reading, only a feelingProgress is being reconstructed from memory, which over-weights the bad fortnight and forgets the six quiet weeks entirely.
Closed by — the plan holding the readings as they happen, so the state is recorded rather than recalled.
-
Check 03 · comparator
In the last seven days, did anything actually compare the two?
NO →Stalls — the comparison never runsComparing requires stopping and looking, and stopping and looking is the first thing a busy week takes. This is the most common break.
Closed by — a review that runs nightly whether or not you thought about the goal that day.
-
Check 04 · output
Is the work that would close the gap actually getting done?
NO →Stalls — diagnosis without executionYou can see the gap perfectly well. Seeing it was never the bottleneck; every step of closing it is still yours to perform.
Closed by — the digital work running overnight, leaving you the real-world steps and the decisions.
- It moves
The four breaks in full
The diagnostic above is the short version. What follows is each component in turn — what control theory says it does, how it fails in an ordinary week, and what we built for it.
Reference value — the target isn’t a target
A comparator needs something it can compare against. “Grow the business”, “get the book written”, “go out on my own” are directions, not set points — there is no reading you could take tomorrow that would tell you whether you are on course.
So the loop never starts. This is the wall people describe as not knowing where to start: not absent motivation, an ambition that has never been converted into anything a next step could attach to.
What we do about it
You describe the ambition in plain words. It comes back as a dated plan — every step with an owner and a date, each one classified as digital (the machine’s), action (yours, in the real world) or decision (yours, and only yours).
That classification is the set point. It is also the division of labour, decided before any work starts.
Input function — you have a feeling, not a reading
The loop’s second part is perception — an honest account of the current state. For a personal ambition that account usually lives nowhere: not written down, not counted, reconstructed from memory whenever someone asks.
Memory is a poor instrument here, and it fails in a specific direction. A bad fortnight is vivid; six quiet weeks of nothing are not. People routinely feel behind while doing fine, and fine while quietly stalling.
What we do about it
The plan holds the readings, so you don’t have to. What moved, what didn’t, what has been waiting on someone else and for how long — kept as it happens rather than reconstructed later.
It knows only what you’ve shared with it, and you can export or delete any of it at any time.
Comparator — nothing compares while you’re not looking
This is the part that fails most often, and the reason is uncomfortable. Carver and Scheier found that the comparator engages when attention is turned inward — heightened self-focus produces “closer self-regulation to the person’s reference value.”
Which means the comparison only happens when you stop and look. For someone running their own operation, stopping and looking is the first thing the week takes. Weeks pass with no comparison run at all, and by the time one is, the gap is large enough to feel like a verdict rather than a correction.
What we do about it
The comparison stops depending on your attention. Every night the plan is reviewed against what actually moved: dates re-estimated, delays traced to a cause, the knock-on effects carried through the rest of the plan.
It runs at four in the morning whether or not you thought about your goals that day. You read the result at breakfast.
Output function — knowing the gap doesn’t close it
The last part is the action that reduces the discrepancy — and this is where an honest diagnosis stops being enough. You can have a real target, an accurate reading and a working comparison, and still be exactly where you were, because every step of the closing is yours to perform.
Most tools stop one part early. They will help you see the gap with great clarity. Seeing it was never the bottleneck.
What we do about it
The digital work runs overnight — drafts, research, plans, the preparation. You wake to finished pieces rather than a list of things to start.
What comes back to you is deliberately small: the real-world steps only a person can take, and the decisions that are genuinely yours. It cannot act on either without your say-so.
The fifth failure — being wrong gets mistaken for failing
There is one more way the loop dies, and it isn’t a missing component — it’s a misreading of a normal one. Every plan makes predictions, and predictions are wrong. In the control-theory account that is not a malfunction; it is the mechanism working.
Prediction errors aren’t problems. They’re a normal part of the operating instructions of your brain. Lisa Feldman Barrett
But a date that slips doesn’t feel like a signal. It feels like evidence about you. So the plan gets abandoned rather than re-estimated — and abandoning it is the only move that actually guarantees the outcome you were afraid of.
So re-estimation is routine here, not a verdict. A slipped date gets a new date and a one-line reason. And the loop also keeps score of itself: how far a goal’s target has drifted from where it started, and how accurate its own past estimates turned out to be. A plan that has moved five times should say so plainly — that is information about the plan, not about the person following it.
A loop with a missing component does not steer badly. It does not steer at all.
Where it stands — what we can honestly claim, and what we can’t
We are pre-launch. There is no user base to point at, no outcome statistics, and nothing on this page is a customer result — we’d rather say that plainly than dress up an anecdote.
What does exist is the machinery, running and measured on our own accounts. One recorded run of the plan-building step:
Example · measured on an internal account, not a customer outcome
What this does not fix
- It cannot want the goal for you. A loop steers toward a set point; it does not supply one.
- Control theory is a model, and a contested one — critics note it can oversimplify behaviour, and that it neglects social influences, emotions, and unconscious processes. A plan is not a life.
- Real-world action stays yours. If the step is a hard conversation, no amount of overnight preparation has it for you.
- Overnight execution is rate-limited, not unlimited: a couple of pieces of work a day on the free plan and more on the paid ones, so a large plan moves in steps rather than all at once. The nightly review itself runs on every plan.
The division of labour is the point
AI is the engine. You set the direction. The part worth building is the gearbox between them — the layer where an ambition becomes work that actually gets done, part by the machine and part by you.
Other AI tools complete a task you already knew to ask for. That still requires you to know the task, which is the barrier in the first place. This works out what the tasks are, then runs the ones it can.
Your iSpirit is the right hand you raise: it runs your goals and daily tasks, learns only what you choose to share, and grows with you — so every month it works better than the last. It costs you a few minutes of chat or voice a day.
Sources. The control-theory account here follows Carver & Scheier’s work on behavioural self-regulation, itself built on Wiener’s cybernetics, as summarised by Psychology Fanatic. Quotations are from that summary; the stated limitations of the model are its own.
psychologyfanatic.com/control-theory