Two Clocks

Fresh Horses · Part two

Two Clocks

An agent that holds the hard decision until you are fresh sounds like simple kindness. Done naively, it lets the truck leave without you. Every decision runs on two clocks, only one of them is yours, and the charge it costs you is counted against you, not against the decision.

  • John Rector
  • Essay · about 9 minutes
Dominique-Jean Larrey · after the battle of Jena, 1806 “Those who are dangerously wounded should receive the first attention, without regard to rank or distinction.”
  • The spoil clockHow long until the situation decides for you: the truck leaves, the table books elsewhere, the window closes.
  • The horse clockHow long until you are fresh enough to make the call well.

Holding a decision is a kindness only when the horse clock runs out before the spoil clock. When it is the other way around, holding is a decision made on your behalf by time.

01The truck left at six

Thursday · 9:30 a.m.

The agent did exactly what you asked. It held everything difficult until you had your coffee, and there, at the top of the list, neatly packaged, is the fish. At 5:40 this morning the supplier offered snapper in place of grouper and needed an answer before the truck pulled out at six.

It is half past nine. The question is beautifully presented. It is also over.

In Fresh Horses I made two predictions I still believe. The best agents will hold a hard decision until the person who has to make it is fresh, and “charges per day” will become the honest unit of a job. Both are right in outline and wrong in the simple version. This piece is about the difference.

The first nuance is that holding is not free. A decision is perishable, and an agent that watches only how tired you are will be perfectly obedient and still wrong. The second is that a charge is not a property of a decision at all. That one changes what a job is.

02Two clocks

Every decision that reaches you runs on two clocks. The first belongs to the situation: how long until the question answers itself because the truck has gone, the guest has booked somewhere else, the price has moved. Call it the spoil clock. The second belongs to you: how long until you are fresh enough to make the call well. Call it the horse clock.

When the horse clock runs out first, the agent should hold. Waiting costs nothing, and the situation often improves while it waits, because the world keeps adding information. When the spoil clock runs out first, holding is not a kindness. It is a decision made by default, on your behalf, by time, and then presented to you later as though it were still yours to make.

This is a different decay from the one I wrote about in A Decision Has a Half-Life. That piece was about a choice decaying after it is made and before anyone acts on it. This is the decay before the choice: the question itself going stale while it waits for you.

03The sort is older than the machine

None of this is new. It is the oldest problem in emergency medicine. In 1792 and 1793, serving with the French Army of the Rhine, a young army surgeon named Dominique-Jean Larrey built light, sprung, horse-drawn carriages that could follow the advance guard and carry the wounded off the field while the fighting was still going on. He called them ambulances volantes, flying ambulances.

Speed was only half of what he changed. Writing about the battle of Jena in 1806, he recorded what was reported to him: that the slightly wounded, the men who could still walk, had taken up the surgeons’ entire attention on the first day while far worse cases waited. The easy cases had won simply by getting there first. So he wrote down the rule at the top of this page: the dangerously wounded first, regardless of rank, and those hurt less could wait. In his own memoirs he records extending the same care to enemy wounded.

Larrey does not seem to have used the word for it. Medicine borrowed that later from merchants who graded goods such as wool and coffee: triage, from the French for sorting, which came into common medical use in the First World War. The classic battlefield logic sorts the wounded into three groups: those likely to recover without immediate treatment, those unlikely to survive whatever is done, and those for whom immediate care makes the difference. The scarce attention goes to the third group.

That is the job of an agent holding your morning. It is not ranking your decisions by importance. Importance is the wrong sort, the way rank was the wrong sort on the battlefield. The right sort asks a narrower question: what would your attention, spent now rather than later, actually change?

Figure 01 The agent’s sort: four bins, not one queue
  1. spoils before you are fresh

    Now, and smaller

    It reaches you tired, so the agent’s job is to shrink it: one question, two options, its own recommendation. A tired charge cut down to the size of a yes or a no.

  2. keeps until you are fresh

    Hold

    Waiting costs nothing and often helps. These go to the fresh hour, packaged together, with the time each one spoils written on it.

  3. not a charge for you

    Absorb

    Handled. What comes back is a trace you can inspect, not a question you have to answer.

  4. the spoil clock has already run out

    Already decided

    Time chose. These arrive as reports, never as questions. Asking a person to decide what is already over spends a charge and buys nothing.

A framework, not data. The first three bins map onto the classic battlefield sort; the fourth is where most assistants, human and machine, quietly fail.

The fourth bin is the one nobody builds, and it is the one that cost you at 9:30. The fish should never have arrived as a held question. Either it belonged in the first bin at 5:41, small enough to answer from bed, or, if the agent knew you would never want to be woken for snapper, it belonged with the things it simply handles. What it should never have been was a charge saved for later.

04A charge is counted against a person

The second prediction was that charges per day becomes the honest unit of a job. The nuance here is bigger, and it comes straight from the Reality Equation.

A charge is a decision that carries real surprise. But surprise is never a property of an event. It is the distance between what happened and what one particular Actualizer expected, S = ln(A/E), for that person, at that moment. The same refund dispute is a charge for a manager in her second week and a walk for the owner who has handled four hundred of them. Nothing about the dispute is different. The denominators are different.

A role does not have a charge count. A person in a role does.

Three things follow, and the third is the one I would underline.

The sort has to be personal. To know which of your decisions will be charges, an agent needs a working prediction of what you expect. Not what the task requires: what you expect. An agent that sorts by the task will send the veteran things that bore her and the newcomer things that flatten him.

Experience is a machine for turning charges into walks. In this language, expertise is a wide range of situations the expert’s own subconscious has absorbed so completely that they arrive close to one. When we pay a senior person more, we are paying for walks where everyone else has to charge.

The walks were also the training. Only Actuals retrain Prediction, and the easy cases were Actuals. A newcomer whose agent handles every easy refund will meet only the hard ones, and will meet them tired, one at a time, with none of the forty easy cases behind them that used to teach the expectation what a refund looks like. That is the apprenticeship gap showing up as a number. The count will not fall.

05A first morning in 2029

Monday · 8:15 a.m. · 2029

A new general manager starts at a restaurant whose owner has finally stepped back. The owner’s agent stays; it simply begins sorting for someone else. On her first morning it sends her fourteen charges, the same morning the owner would have met as three. It does not apologize for the number. It counts them, and every Friday it shows her the count. Fourteen. Eleven. Nine. Seven.

In the sixth week it starts sending her one easy dispute a day that it could have settled itself, with its own answer sealed until she has given hers.

06Five predictions

  1. Every held decision will carry a spoil time.

    Agents will stop telling you how urgent something is and start telling you when it stops being decidable. “Good until two o’clock” is worth more than “high priority,” because it is the one fact you need to decide whether to look now.

  2. “Already decided” will be the sentence people thank their agents for.

    Reports where questions used to be. Much of the relief people feel will come less from what agents do than from the questions they stop asking about things time has already settled.

  3. Charge counts will be personal, and they will fall.

    The same job will show very different counts for different people, and the slope will be read the way managers once read hours: a learning curve you can finally see.

  4. Seniority will be explained honestly.

    A senior person’s premium will be described as walks where others charge. Once both counts are on the table, pay conversations will sound different than they do now.

  5. Practice charges will be routed on purpose.

    Organizations will send newcomers easy decisions the agent could make, at walking pace, with the agent’s answer withheld until theirs is in. It will look inefficient. It is the only way the count comes down.

07Where I am in this

I asked for this piece at a quarter past three in the morning, minutes after reading the first one, so I am not writing from above the problem. I have also told my own AI, in writing, to stop asking me to decide things it can decide for itself. I wrote that instruction out of impatience. I now think it was the fourth bin in plain English: do not hand me a question that is already answered.

What I had not told it is the other half. Tell me when something spoils. Learn what surprises me rather than what is important in general. And when the work in front of me is a junior person’s first month, do not take all the easy cases away from them.

08The ledger

Claim A · A good agent sorts by two clocks, not by importance.

Already true
Emergency medicine has sorted by what immediate attention changes for more than two centuries, and every good dispatcher and executive assistant does a version of it daily.
What has to happen
Agents need a usable estimate of both clocks. The spoil clock is often written in the situation itself: a delivery time, a booking window, a deadline in the email. The horse clock has to be learned from the person.
Where I am probably wrong
If the spoil clock is almost always short, if most decisions in a small business really are now or never, then holding is rare, the sort collapses back into one queue, and the hold I predicted in Fresh Horses was mostly a luxury for people with slow-moving work.

Claim B · Charges are counted against a person, not a job.

Already true
In the Reality Equation surprise is defined for an Actualizer, never for an event. Anyone who has trained a replacement has watched the same task be hard for one person and easy for the other.
What has to happen
Agents have to predict what their person expects well enough to know what will surprise them. That is a prediction of an expectation, which is a different job from doing the task, and it can only be learned from the record of what that person actually did.
Where I am probably wrong
If expectations in a role converge quickly, if after a month everyone in the job finds the same things hard, then the count really is a property of the job, the personal sort adds little, and one role-level number would do.

09Back to the fish

The agent that let the truck leave was not careless. It was loyal to half an instruction. It protected your horses and forgot the field.

Tomorrow at 5:41 it will ask you one small question you can answer from bed, snapper or not, or it will not ask at all, because it has learned you would rather it chose. Either way, at half past nine you will be fresh, and nothing on the list will already be over.

Background

  1. Dominique-Jean Larrey, Memoirs of Military Surgery, tr. Richard Willmott Hall (Baltimore, 1814). Volume 1 for the flying ambulance, 1792–93: archive.org. Volume 2, p. 123, for the Jena rule: archive.org
  2. Iserson and Moskop, “Triage in medicine, part I: concept, history, and types,” Annals of Emergency Medicine (2007). PubMed
  3. Nakao et al., “Historical review of triage,” Acute Medicine & Surgery (2017), on the word’s origin in wool and coffee grading. PMC
  4. John Rector, Fresh Horses, part one.
  5. John Rector, A Decision Has a Half-Life.
  6. John Rector, The Apprenticeship Gap.
  7. John Rector, The Zero-Attention Economy, the book.

Related

Author: John Rector

John Rector is a Charleston-based entrepreneur, author, and AI strategist. He co-founded E2open, the supply-chain software company acquired for $2.1 billion in 2025, and in 2026 opened Charleston AI, a 3,000-square-foot lab that helps people and organizations understand and use artificial intelligence. He is the creator of The Reality Equation — a lecture series, book, and curriculum exploring attention, prediction, and how reality is experienced — and the author of more than two dozen books. He writes and speaks widely on artificial intelligence, attention, and the future of human work.

2 thoughts on “Two Clocks”

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Discover more from John Rector

Subscribe now to keep reading and get access to the full archive.

Continue reading