DESCUBRE CÓMO DESTERRAR

Servitor practice is taught almost entirely in the imperative: build one, charge it, feed it. The exit gets a sentence. So the most reliable outcome of a few years of this work is a growing private pantheon that nobody ever formally retires, each unit still drawing on a fixed and small reserve of attention that nobody was told was fixed. The vocabulary for ending exists. Dismissal, egress, release, unbinding all appear in the literature, because the technique inherited them from the older protective-spirit material it grew out of, and every one of them assumes the relationship has a stated end. What follows is an account of the exit as the hard part, from field notes rather than theory: the dismissal that went wrong, the sign that was misread, the rule about announcing the end, the two failures that repeat, and the part where you discover the servitor was already gone, which is a different problem from the one you thought you had.

The exit is the hard part

Building a servitor is satisfying in a way retiring one is not. The build has a beginning, an artefact, and a first moment of apparent success. The retirement has none of those. There is no artefact to show, no moment that announces itself, and often no evidence at all that anything stopped. This asymmetry is why the practice accumulates: the pleasant operation gets repeated and the unpleasant one gets deferred, and a two-year deferral is indistinguishable, from the inside, from a servitor working exactly as intended.

The technical reason is that the construction rests on an unstated arrangement. A servitor is framed as a displaced part of the self, a fixed charge of attention given a task and a trigger. Nothing in that framing announces a term. The term has to be supplied by you in advance, and if you did not supply it then you are not retiring something with a defined end. You are hoping it notices.

What the vocabulary of dismissal gives you

The technique took its charge, feeding, and dismissal language from protective-spirit practice, and the inherited vocabulary is more useful than practitioners usually notice. Dismissal states a fact and closes a relationship. Egress describes withdrawing the attention you were maintaining. Release and unbinding describe the period after that withdrawal, when you stop checking. Four operations, and beginners routinely confuse the first three.

The distinction that matters most is between dismissal and discharge. A discharge pushes the unit outward and lets it go. A dismissal ends the arrangement while keeping the record. Most of what gets described as dismissing something is a discharge, and the difference is visible afterwards: a discharged unit leaves a hole where the attention used to be, and a dismissed one leaves a closed file.

The first dismissal that went wrong

I dismissed one in a single sitting after a difficult month, without a term having been declared when it was built. The two weeks afterwards were worse than the two weeks before, which is the part nobody warns you about. A unit that has been running a trigger for eleven months does not stop on request; it stops when the attention behind the trigger is actually withdrawn, and a ritual that only declares an intention has not yet withdrawn anything.

The construction that worked on the retry was unglamorous: the task was written down, the trigger was named, the feeding was skipped deliberately for a fortnight, and the dismissal sentence stated the specific end of the arrangement rather than a general release. The fortnight of not feeding is the operative part, because feeding is the maintenance of the attention and the maintenance has to be stopped before the ending can mean anything.

Reading the sign you misread

The sign that a unit is still active is not an event. It is a habit of checking. After a successful retirement, the thing you used to do stops happening: you do not glance at the notebook to see whether the trigger fired, you do not feel a small lift on the days you forgot to write. Most people describe this as a loss and read it as a failure of the dismissal, when it is the ordinary sensation of a commitment that has been closed.

The sign that it was never active is different and harder to accept. If the task had been done anyway, out of habit or obligation, then the unit was doing nothing, and the retirement ceremony has retired an idea. That is worth establishing, because a unit credited with work you were already doing is a unit that will be credited with work you stop doing.

The rule about announcing the end

The single rule that would have prevented most of the difficulty is to write the term into the design at build time. One sentence, in the same sitting as the construction, saying what ends it and on what condition. Nothing in the technique requires this, which is why almost nobody does it, and adding it converts an unbounded arrangement into a contract with a clause.

The condition has to be a thing you can evaluate. Not when I have achieved the aim, because that never arrives and leaves the arrangement open indefinitely. Something like when the work has been done for four consecutive weeks without the charge being used, or when this specific project ends. A date is better than a feeling for exactly the reason it is uncomfortable: it can be checked.

Doing it in a single sitting

The retirement does not need a session. It needs an hour, a written record of what the unit was for, and the discipline to skip the part that feels ceremonial. What the hour is for is documentation: the task, the trigger, the duration, the feedings, the results that were actually observed. A retirement performed without that record leaves you unable to answer, a year later, whether the thing ever did anything, which is the question that will eventually be asked.

Then the withdrawal, and then the announcement, in that order. The announcement that comes first is the common mistake, because declaring an ending while the maintenance is still running produces a unit that has been told it is finished and continues, which is worse than one that was never told, because now there is a story about why it is misbehaving.

What to keep when you let one go

The record is what you keep, and the record has a specific value: it stops the next one from being the same mistake. Written down, a retirement becomes a design lesson, and the design lesson is transferable in a way the unit is not. Six retirements produce a reliable account of which tasks suited a charge and which were carried by the person anyway.

What you do not keep is the sigil. A charged artefact belonging to a closed arrangement is an open loop, and keeping it on a wall or in a folder is the same as not having dismissed it. The unit is described in words and the words are filed, and the drawing goes into the same envelope as the record or goes into the bin.

The count, brought back down

Modern practice dropped the old limit on how many could be maintained, and that removal is defensible, because the limit was never a mechanism. What replaced it is worse: a practical ceiling that nobody writes down. Each active unit occupies attention continuously and mostly invisibly, which means the cost does not appear as a feeling until the total crosses what you can hold, and then it appears as general fatigue and difficulty concentrating on anything unrelated to the units.

So the count has to be written down and reviewed the way any other budget is. One, or two, is a defensible number for most people doing this alongside a job. Above three, the returns are not obviously there and the costs are. The review that matters is not the count itself but the question of how many of the current units are doing work you would not otherwise have done.

A dismissal log worth keeping

A log needs four columns and no more: the unit, the date it was built, the date it was retired, and the honest assessment of whether it earned the attention. A month is enough. The reason for keeping it is that the last column is the one you cannot reconstruct later, because your memory of a good month will improve with time and your memory of the unit will improve with it.

Reading three of these in a row is instructive in a way that no amount of theory about retirement is. The pattern is usually visible: the units that did nothing were the ones whose tasks were vague, and the units that helped were the ones whose triggers were tied to a moment in the day rather than to an intention.

The two failures that repeat

The first is retiring a unit whose task has become your job. Not consciously, not expensively, but gradually, so that when the unit goes you notice a gap in the day and reinstate it by habit. The marker is that you cannot remember the last time you would have done the task on a day when the unit was not working, which is a question the log can answer if you started it early.

The second is multiplying. It feels like progress, it looks like ambition, and it converts a practice with one cost into a practice with four. The diagnostic is whether the last unit built was built to do something or to relieve the discomfort of having too few. If you cannot name the task in one sentence without consulting the record, that is the answer.

When the servitor is already gone

Sometimes the discovery is that there is nothing to dismiss. The task has changed, the project ended months ago, and the unit is sitting in the background of a life that no longer has the corresponding activity. This is a different problem from an active unit and it is easier to resolve, because what needs retiring is an arrangement whose trigger no longer fires.

Treat it as a cancellation rather than a dismissal, and write it as one: the contract lapsed, no work was pending, and the record is closed. The temptation to run the full ceremony on an arrangement that has already ended is worth resisting, because the ceremony will produce a strong impression of having done something, and that impression is exactly the thing a log exists to check.

Nothing at all, and why that is the useful entry

Four lines. What it was for, how long it ran, what you actually observed, and what you would do differently. That last one is the only one that changes anything, and it is the one that gets skipped because by the time the retirement is over the interest is spent. Write it in the same sitting as the withdrawal, while the account is still accurate, and then the file is closed.

The larger point is about the shape of the practice. Servitor work is usually presented as a technology of assistance, and it is much closer to a technology of maintenance. The assistance is a side effect of having written the task down clearly and kept it in front of you. Everything else is bookkeeping, and the bookkeeping is where the whole difficulty lives.

Frequently Asked Questions

How many servitors can I realistically maintain?

One or two is a defensible number for most people, and the reason is that each active unit occupies attention continuously. Above three the returns are not obviously present while the costs are, and the cost does not appear as a feeling until the total crosses what you can hold, at which point it shows up as general fatigue rather than as anything you would attribute to the practice.

What is the difference between dismissal and discharge?

A discharge pushes the unit outward and lets it go, and it leaves a hole where the attention was. A dismissal ends the arrangement while keeping the record, and it leaves a closed file. Most of what gets described as dismissing something is actually a discharge, which is why the retirement can feel like a loss rather than a closure.

How do I know if my servitor was actually doing anything?

Ask whether you would have done the task on a day when the unit was not charged, out of habit or obligation. If the honest answer is yes, the unit was carrying nothing and the work was yours. This is the hardest column in a retirement log to fill accurately, because your memory of a good month improves with time along with everything else.

Should I write an end condition when I build the servitor?

Yes, in the same sitting as the construction, and it should be something you can evaluate rather than something that depends on the aim arriving. A date or a countable condition works; a state of mind does not. Nothing in the technique requires this, which is why it is almost never done, and adding it converts an unbounded arrangement into one with a clause.

What if the servitor has already stopped without my doing anything?

Treat it as a cancellation rather than a dismissal and write it as one: the contract lapsed, no work was pending, the record is closed. Running a full retirement ceremony on an arrangement that has already ended produces a strong impression of having done something, and that impression is precisely what the log exists to check against.