The Spiral That Tightens Itself

There is a particular Tuesday that tells you something has shifted.

Nothing on it is new. The same category of decisions you have made for eleven years. A supplier conversation, two approvals, a pricing question, a hire who is not working out. Eighteen months ago this was an ordinary Tuesday. You would have moved through it without noticing you had moved through it.

Today it sits on your chest.

You check the obvious explanations first, because you are a competent person and that is what competent people do. Volume is not up. Revenue is fine. Nobody has died. The business is, by any measure a bank would apply, in better condition than it was three years ago.

And the work is harder than it has ever been.

So you reach the only conclusion available: something is wrong with you. You are tired, or getting older, or have lost some edge you used to have. You resolve to sleep more, exercise properly, take the holiday you have deferred twice.

That conclusion is wrong. Not the tiredness - the tiredness is real. The causal story is wrong, and because the story is wrong, every remedy that follows from it is aimed at the wrong target.

The reading everyone gives it

The standard model treats this as a fuel problem.

Under the fuel model, you are a tank. Demands draw down the level. When the level is low, everything requires more effort, because you are running on less. The remedy follows automatically: put fuel back in. Sleep. Rest. Take the break. Protect your evenings. Guard the weekend. Come back restored, and the same demands will feel manageable again, because they always were manageable - you were simply too empty to meet them.

Search the phrase that brought most people to this page and you will find that model repeated in near-identical language across hundreds of pieces. Rest more. Set boundaries. Practise self-care. Protect your energy.

The advice is not wrong, exactly. It is aimed at a variable that turns out to be secondary.

The fuel model predicts that rest ends the problem. Most founders have run that experiment personally, and know the result.

You have almost certainly tested it. You took the holiday. For four days you felt genuinely different. You came back, and inside a week you were exactly where you started - which under the fuel model should not happen, because the tank was full.

When a model makes a clean prediction and reality contradicts it, the model is wrong. Not the person running the experiment.

Which direction the causation runs

In 2020, Christina Guthier, Christian Dormann and Manuel Voelkle published a meta-analysis in Psychological Bulletin that asked a question this territory rarely raises.

Everyone assumes demands cause burnout. Arrow pointing one way. Their study tested whether the arrow also points back - whether burnout causes demands, or more precisely, whether burnout changes the demands a person reports experiencing.

They pooled 48 longitudinal studies covering 26,319 participants, using a method designed to handle the fact that different studies measured their subjects at different intervals. That last detail matters more than it sounds: the direction of an effect can look different depending on whether you check back after three months or after three years, and most earlier work could not reconcile those timeframes.

The finding, in their own words: the stressor effect is small, whereas the strain effect is larger.

The phrasing there is carrying weight. The “stressor effect” is demands causing burnout - the arrow everyone assumes. The “strain effect” is burnout causing stressors. Of the two, the one everybody builds their advice around is the smaller.

Both effects are modest in absolute terms. This is not a claim that workload is irrelevant. It is a claim about which direction dominates once the process is under way, and the answer is the direction nobody is treating.

A 2025 re-examination by Maunz and colleagues went back to the same question and again found stronger support for the strain hypothesis, with burnout at one point predicting demands reported six months later.

The part most readings drop

There is a second finding in the same paper, and it matters as much as the first.

The strain effect was not uniform. It was moderated by two variables: job control and job support. In the sub-sample where the researchers tested this - 31 European studies, 17,747 people - how much authority someone held over their own work, and how much support surrounded them, changed how strongly burnout fed back into perceived demand.

Neither of those is a property of the individual. Both are properties of the structure the individual is operating inside.

This is the part the self-care corpus cannot see, and it is why two founders carrying identical workloads spiral at different rates. If your operating structure has concentrated every consequential decision on one person, removed anyone with the authority to absorb load, and left you accountable for outcomes you do not fully control, the spiral is not simply running inside your head. It is being driven by an architecture - one you built, which is also the reason you can change it. A business that cannot run without you is not only a valuation problem. It is a regulation problem with a structural cause.

So there are two variables in play, not one. The appraisal system drifts, and the structure determines how fast it drifts. Interventions aimed at neither will not hold.

The instrument, not the tank

Follow that through, because it is not a small adjustment to the fuel model. It replaces it.

If burnout increases the stressors you report, and it is not conjuring new work out of nothing, then it is changing something about how existing work registers. The supplier conversation did not become harder. Your reading of the supplier conversation became harder.

Which means the thing that degraded is not your tank. It is your instrument.

This is why the holiday failed, and it failed for a structural reason rather than a personal one. Rest puts fuel in the tank. It does not recalibrate an instrument. You returned rested, and your rested self read an ordinary Tuesday through the same miscalibrated gauge, and the gauge said: this is heavy.

Anyone who has spent time around measurement will recognise the shape of this immediately. The apparatus doing the assessing is the same apparatus being degraded. It cannot flag its own drift, because the flagging function is downstream of the thing that drifted. Ask the system whether it is reading correctly and it will answer using the instrument in question.

You cannot detect a miscalibrated gauge by consulting the gauge.

The second thing that breaks

A physiological layer sits underneath this, and it explains the acceleration.

Bruce McEwen’s work on allostatic load describes what sustained stress-system activity costs the body. Allostasis is the machinery that mounts a response to a demand and then stands the response down once the demand has passed. Allostatic load is the wear that accumulates when that machinery runs chronically. This is the chronic counterpart to the acute nervous-system response - not a spike you recover from, but the slow cost of a system that has been running the spike too often.

The critical detail is what specifically fails. It is not simply that the system gets tired. One of the named failure types is the response not switching off after the stressor has ended.

Consider what that does in combination with the first finding.

The system that judges how demanding something is has drifted upward. And the system that terminates the response once the demand has passed has become less reliable. So demands read as heavier than they are, and the response to them runs longer than it should, which extends the load, which drives further drift.

The sensor and the brake are both being consumed by the process they exist to regulate.

That is the spiral, stated mechanically. And it explains the property founders find most disorienting about it: it does not plateau. A fuel problem plateaus - you reach a floor, stabilise, adapt. A loop with a degrading governor does not stabilise. It tightens, at whatever rate the load allows, because each turn slightly worsens the equipment that would otherwise arrest the next one.

Why “nothing works” is an accurate report

One sentence shows up constantly among high-functioning people who have been in this for years.

I have tried everything and nothing works.

It usually gets heard as exaggeration. It is generally more precise than it sounds.

One boundary before going further, because the symptoms overlap and the distinction matters. What follows describes a load-and-regulation problem in someone whose output is still holding. It is not a description of clinical depression, which presents similarly from the outside and requires a different kind of help entirely. If the flatness extends past the work into things you used to want, that is a different conversation and a doctor is the right person to have it with. The rest of this piece assumes the first case.

Run the inventory of what “everything” contained. Better sleep. Exercise. A holiday. Boundaries on evening email. A meditation app. Possibly a shorter week for a while. Every one of those is a restorative intervention. Every one of them adds something back to a depleted system.

Not one of them removes anything from what the system has to regulate.

This is worth stating precisely, because the opposite reading is the one that gets people into trouble. Sleep, recovery periodisation and deliberate down-shift days are not wasted effort. They are the correct intervention at the substrate layer - they repair the regulatory capacity itself, and without them the capacity keeps eroding no matter what else you do. What they do not do is reduce the number of things demanding regulation. Capacity restored against unchanged load is capacity that gets consumed again on the same schedule, which is why the effect wears off with a regularity that feels like personal failure and is not.

Two variables, one loop. Restoration acts on the first. Only subtraction acts on the second.

The loop was never about the level. The loop is about structure - specifically, about how many independent things require appraisal and response from an appraisal system that has drifted. Restorative interventions raise the starting level and leave the structure untouched. The structure then does what it has been doing, from a slightly higher starting point, and arrives back at the same place. The person concludes they are broken, because they did everything correctly and it did not hold.

They did everything correctly for one half of the problem, and nothing at all for the other.

The name for this

MindMastery calls this the Regulation Gap - the sixth of the seven core problems, and the one that hides longest.

It hides because output holds. That is the whole difficulty. Cognitive and emotional regulation degrade underneath a surface of continued competence, and the competence is real - deliverables ship, decisions get made, the business does not visibly wobble. Nobody intervenes, because from outside there is nothing to intervene in.

The gap is between what the system can still produce and what it costs to produce it. That cost is invisible on every instrument anyone is checking, including yours, for the reason already given.

Which is why this one is rarely caught early. It is caught when the cost finally exceeds what output can be maintained on, and by then the loop has had years to tighten.

The distinction to run this week

One question separates a capacity problem from a regulation problem, and it is answerable without any instrument you would have to trust.

Did the demand itself change?

Take something producing disproportionate resistance right now. A conversation you keep deferring, an approval that sits, a decision that should take ten minutes and has taken three weeks. Now compare it to the same category of thing eighteen months or two years ago.

Not how it felt. What it objectively involved. Same number of people, same money, same reversibility, same consequences for getting it wrong.

If the demand has materially grown, you have a capacity problem, and capacity problems have capacity solutions - more resource, fewer commitments, better systems.

If the demand has not materially changed and your response to it has, the variable that moved is the appraisal, not the task. That is the Regulation Gap, and no amount of resource added to the system will address it, because the system is not short of resource.

This test works because it routes around the compromised instrument. You are not asking yourself how hard something feels - that question interrogates the gauge. You are asking what the thing objectively consists of, then comparing across time. External referent, not internal reading.

What actually intervenes

The intervention has to be subtractive, and it has to be permanent. Those two conditions are what separate it from everything already tried.

Not a holiday, because a holiday is temporary by construction and you return to an unchanged structure. Not a single dramatic reduction, because dramatic reductions get reversed within a quarter as the business reasserts its claims.

The unit of change is the recurring decision-stream.

Identify one decision you make repeatedly - weekly, or several times a week - and remove it from yourself entirely. Three routes, in descending order of preference: eliminate it, so nobody makes it; automate it, so a rule makes it; or delegate it with the decision right attached, so someone else makes it and you do not review the outcome.

That last clause is where most attempts fail. Handing over a decision while retaining review of every outcome does not remove the decision from your appraisal system. It converts one decision into two - the original, plus the ongoing supervision - and increases the load you were trying to reduce. This is the orchestration trap in its most expensive form: supervising everything, deciding nothing, and calling it delegation.

One per week. Not five. One.

The efficiency gained is not the point. On any given item it will be trivial, which will tempt you to skip the exercise as beneath your time. The point is that each removal is a permanent structural reduction in what the appraisal system has to appraise. Twelve weeks produces twelve fewer streams, and that is a different structure rather than a rested version of the same one.

The spiral tightens because load drives drift and drift increases load. Subtraction is the only move that acts on both terms at once.

The thing that makes this problem durable is that it recruits your competence against you. You keep meeting the demands. The meeting of them is exactly what conceals the cost, and the concealment is what buys the loop the years it needs to tighten properly. It is the same bill that arrives with every other form of success tax: the capability that got you here is the reason nobody can see what it is now costing to maintain.

The demands did not grow.

Something else did.