There is a particular small pleasure in opening a
configuration file you wrote some years ago and finding
that it still works. A pf.conf, perhaps, or
a Makefile: a comment addressed to a future
self who turned out to be you, a workaround whose reason
you can still reconstruct, a line or two you would write
differently now and have left exactly as they are,
because they are correct. The file was never taken to its
hundredth per cent. It was taken to the point where it
was right, and then left alone. And it runs.
Set against this the more familiar experience of modern software: the stack assembled eighteen months ago that is already spoken of as legacy, the framework whose migration guide is longer than the application it migrates, the surface buffed to a mirror over something quietly coming apart underneath. We have learned to call the chase for that mirror perfectionism, and to treat the word as a compliment. It is worth asking whether it deserves to be one.
The Last Ten Per Cent
Perfection, as the industry practises it, is the chase for the last ten per cent. And the last ten per cent costs more than the ninety that came before it. That alone would be a poor trade. What makes it a bad one is that the polish so rarely earns out, because everything has a shelf-life. A system, a feature, a format, a fashion: each has a span of validity, and when the span ends, the month you spent perfecting a corner of it is simply gone, spent on a refinement that expired before anyone needed it. One can buff a doorknob to a mirror on a house already scheduled for demolition. The diminishing returns are real, and impermanence collects on them without mercy.
And the cost is not only effort. The person who chases the last ten per cent tends to lose the whole in the parts: a thousand small refinements pursued without once weighing them against the goal, the goal itself mislaid somewhere around the four-hundredth. It is, reliably, a way to burn out, because the standard is internal and infinite and turned against the one who holds it. The maker is consumed by the making, in the precise manner that produces neither a finished thing nor a rested person.
And the word itself is a misnomer worth correcting. The person who builds everything to the maximum, who cannot leave a thing until it is flawless and calls this perfectionism, is not a perfectionist. He is its opposite. Perfect is not the maximum; perfect is the fit, the result proportioned exactly to what the scenario asks and no further. To pour unbounded effort into every corner is to have lost the sense of what fits, and to have no measure of one's own limits.
Perfect is not the maximum. It is the fit, the result proportioned exactly to what the scenario asks and no further. The one who cannot stop has mistaken exhaustion for rigour.
And there is a plainer test the word fails: what nine in ten of the people who use a thing cannot even perceive without having it explained to them is, by definition, no part of its fit. It is surplus, and surplus is never free. It is paid for three times over: once in the making, in the hours poured into a refinement no one asked for; again in the explaining, because a merit no one notices has to be pointed out, and pointed out again; and a third time in the slow attrition of the maker, who comes to feel permanently underappreciated for a value that is glaringly obvious to him and entirely invisible to everyone else. The grievance is real, and it is self-inflicted, and it is exhausting in the particular way of a debt one keeps paying to oneself.
Where Good Work Actually Comes From
It would be easy to misread all this as a defence of doing less, of the cut corner and the good-enough that is merely careless. It is the opposite, and the distinction is the whole point.
Stopping at the fit is not the same as stopping early. It is harder. It asks you to know, before you have exhausted yourself proving it, exactly where the work stops paying and to have the judgement to halt on that line rather than the one fatigue draws for you later. That judgement is not a shortcut around mastery. It is mastery: simplicity and a very good level, joined in years of experience. The novice cannot find the fit, so he either stops too soon or cannot stop at all. The discipline of the right amount is something you earn.
The demoscene has known this for forty years. A 4K intro is an executable of 4,096 bytes that renders a three-dimensional world, with light and motion and a synthesised soundtrack, everything generated from mathematics at the moment it runs. There is no room for a framework, no room for a spare byte; every byte is a decision. Finland, Germany and Poland have each placed the demoscene on their lists of cultural heritage, on a single recognition: that severe constraint produces extraordinary results, not despite the limitation but because of it. Less produces more, provided the less is deliberate. That is not the aesthetic of the maximum. It is the aesthetic of the fit, practised as an art form.
The same principle runs quieter through ordinary good engineering. KISS, lean, minimal: these are not slogans for laziness but for the discipline of removing everything that does not earn its place, so that what remains is exactly proportioned to the problem and can be held whole in one mind. A thing built that way reaches very good reliably, and absolutely perfect never, and the people who use it tend to call it the best of its kind precisely because of the restraint, though they would not put it that way. The maximum was never the goal. The fit was, and the fit is what people trust.
The Honoured Seam
There is an older idea behind all of this, and it is worth naming because it gives the argument an image.
Kintsugi is the Japanese repair of broken pottery with lacquer and gold, so that the mend is not concealed but honoured, the history of the object made visible and, in the making visible, made valuable. It belongs to wabi-sabi, the aesthetic that finds beauty in the imperfect and the worn. The relevant part is not the melancholy; it is the honesty. A good system, like a well-mended bowl, does not hide its repairs. It shows the line you would write differently now and kept because it is correct; it shows the workaround and the reason for it; it carries its decisions on its surface where the next maintainer can read them.
But a careful line has to be drawn here, because the idea is one short step from a much worse one. The honoured seam of a maintained system is the precise opposite of the unmanaged seam of one assembled from a dozen projects by no one in particular. When a coherent thing shows its mends, that is history, and you can still account for every one. When an incoherent thing shows its joins, that is not character: it is the visible evidence that no single hand owned the whole, and the gaps between the parts are where the failures live. Imperfection chosen by a maker is beautiful. Imperfection produced by the absence of a maker is just rot with good lighting. The test is simple: can someone still hold this in mind and say why it is the way it is? If yes, the marks are history. If no, they are seams, and the seams show.
The Limit
It would be dishonest to let the argument do more work than it can bear, so let me mark its edges plainly.
Stopping at the fit is not an excuse to keep the old because it is old. There is nothing romantic about an unpatched daemon, and a system left to age without maintenance is not wearing its years with dignity, it is decaying, and sometimes it is dangerous: an OpenSSH from the wrong year (regreSSHion, CVE-2024-6387) is a door, not an heirloom. Maintenance is the whole of the point, and maintenance is work.
Nor is this an argument against the rewrite. I am not against rewriting; some systems are so far gone that the honest thing is to tear them down and let the replacement breathe, and pretending otherwise out of sentiment is its own kind of perfectionism, the kind that cannot let go. The reflex to rewrite the working thing because it is not flawless is the failure. The decision to rewrite the genuinely broken thing because it can no longer be reasoned about is judgement. Telling those two apart is the same skill as knowing where the fit lies: it cuts both ways, and the wisdom is in the cut, not in a rule that always keeps or always replaces.
Closing
What changes, somewhere in an engineer's career, is what earns the admiration. Early on it is the maximum: the most complete, the most polished, the most. Later, if one is fortunate, it inverts, and the thing that earns respect is the one that knew where to stop, the file taken to the point of being right and no further, the system small enough to reason about and maintained with attention, the work that reached very good and declined, on purpose, to ruin itself reaching for more.
Perfect is not the maximum. It is the fit, and the experience to know where it lies. Build only what the problem asks, own the whole of it, maintain it with care, and stop when it is right: you are left with something that wears rather than breaks, that shows its history rather than hides its rot, and that people will, in time, come to trust. The work driven to its last exhausting per cent ages, as often as not, into garbage. The work taken to its fit, and no further, ages into something rather like grace.