We are clever primates on a wet rock, drifting through a dark sea of space, and one of our most persistent illusions is that the universe is impressed by our dashboards. We take something messy and alive—human attention, judgment, sleep, dignity—and we pour it into tidy containers called targets, and then we act surprised when the container leaks. Somewhere in a bright meeting room, under lighting designed to flatter neither skin nor conscience, we declare that “output” can be optimized the way one optimizes a delivery route, as if a mind were a warehouse and a day were a conveyor belt, and the great joke—cosmic, quiet, and very old—is that nature has never once signed your service-level agreement. There is, in every organisation, a small altar built from slides and acronyms, and on it we place offerings: a little more speed, a little less headcount, a little extra stretch, and we call it “discipline,” because “panic with a haircut” sounds unprofessional, and then we add a new tool, ideally one with a name that suggests inevitability, as if technology were a weather front rolling in from the west, and now, conveniently, any discomfort can be reframed as “change management,” which is the polite way of saying: you will adapt, and the system will learn nothing.
From far enough away, it is almost beautiful: billions of neurons firing, building meaning from noise, trying to keep reality aligned with the stories we tell each other. From up close, it looks like an inbox at 23:17, a set of numbers that must be right because the world outside the window depends on them being right, and a person whose job is to prove, again and again, that the system did not hallucinate. We like to imagine that work is a simple equation—fewer people plus better tooling equals the same result—and we say it with the serene confidence of someone who has never tried to reconcile two stubborn truths that were born in different systems and raised by different definitions. But the universe has a sense of humour, and it expresses it through variance. It expresses it through late changes, missing fields, shifting rules, new interpretations, upstream “minor” adjustments that arrive like meteorites, and the quiet miracle that any complex process runs at all, even on good days, even with normal staffing, even before we decide to run it at 97% capacity forever and call the remaining 3% “fat.”
Here is the uncomfortable part: “efficiency” is often just a nicer word for converting slack into hope. Slack is the breathing room where errors are caught early, where people can read a change request twice, where someone can ask, without fear, “Are we sure this is true?” It is the margin that allows a system to absorb the ordinary unpredictability of reality—because reality is not polite enough to arrive on schedule. When slack disappears, the system does not become streamlined; it becomes brittle, and brittle systems do not fail loudly at first. They fail the way stars die: quietly at the core, long before anyone on the surface notices. The output can look fine for months. The charts can be green. The deadlines can be met. And all the while, the cost is being shifted into places that are hard to measure: the quality of evidence, the clarity of lineage, the depth of understanding, the time to think, the willingness to escalate, the courage to say “unknown.” This is not corruption; it is thermodynamics. You do not get order without paying for it. You can only decide where the payment is made.
And because we are social creatures, we pay attention to rewards. We notice what gets praised. We notice what gets promoted. We notice which stories get told at the top. If the story is “we did more with less,” then the system will learn how to tell that story, and it will become very good at it. It will move work into shadows. It will rename unfinished as “in progress.” It will transform controls into rituals and rituals into checkboxes, and checkboxes into soothing noise. It will create a museum of “temporary” spreadsheets that never close, like abandoned satellites still orbiting the process, debris from prior expeditions that nobody dares to decommission because no one remembers what they’re holding together. It will become a place where heroic effort is treated as normal operating procedure, because heroism is cheaper than redesign, and because it photographs well. If you want a harsh definition of organisational maturity, it is this: do you build systems that protect people from routine heroics, or do you build myths that require them?
Now add a new tool, something powerful and fast, something that can write, classify, summarize, generate, explain. We love it immediately, because it feels like leverage, and leverage is intoxicating. But we should ask: leverage for what, exactly? For truth, or for speed? For understanding, or for appearances? If a tool produces an answer faster than we can validate it, have we accelerated work—or accelerated risk? If it drafts an explanation that sounds confident, is it helping us communicate reality, or helping us decorate uncertainty with adjectives? There is a particular kind of danger in fluent outputs: they can make the organisation feel smart while it becomes less precise. They can make ambiguity look resolved. They can make it easier to stop thinking, which is, in many places, the most aggressively optimised activity of all. The tool, meanwhile, does not care. It does not get tired. It does not fear audits. It does not wake up at 04:00 remembering a missing sign-off. It does not understand consequences. It emits. We decide.
So why do organisations overpush when they know it ends badly? Because the end is later, and later belongs to someone else. Because the rewards are now, and now is legible: a cost line drops, a metric rises, a narrative clicks into place. Because humans discount future pain, especially when it will be distributed across many hands and many calendars. Because complexity hides causality; when failure finally arrives, it rarely comes with a neat label reading “caused by last year’s headcount cut.” It arrives as an incident, a finding, a restatement, a late-night escalation, and the system will call it an exception, an outlier, a one-off—anything but the bill for the loan it took out against resilience. And because in competitive ecosystems, restraint looks like weakness. If everyone else claims to be running faster, you feel compelled to run faster too, even if you are sprinting on a bridge made of matchsticks, even if you can smell the smoke, even if the alarms are inconvenient.
Which raises questions that feel impolite precisely because they are practical. If a process only works when everyone is slightly overextended, is it a process or a trap? If the official capacity plan assumes perfect inputs, no surprises, and people who never get sick, what species is it designed for? If “doing more with less” is celebrated, what exactly are we doing less of: checking, understanding, documenting, escalating, sleeping, telling the truth? If we remove redundancy and then praise the absence of redundancy-related “waste,” have we optimized a system—or optimized the conditions for a single point of failure to become a single point of catastrophe? If the metrics are green but the explanations are thinner, the evidence messier, the runbooks outdated, the subject-matter knowledge concentrated in two exhausted people, what exactly is being measured? Output, or the ability to postpone reality?
And here is the satirical twist the cosmos would appreciate: even when the overpush “works,” it often works by spending something precious that does not appear on any balance sheet—trust. Trust that you can raise a risk without being punished. Trust that the organisation prefers a hard truth today over an elegant story this quarter. Trust that when you say “I don’t know,” you will be given time to find out rather than asked to improvise confidence. Trust that quality is not a slogan used when convenient and ignored when costly. Once that trust is spent, people stop donating their judgment. They do only what is asked, exactly as asked, and the system may become more compliant while becoming less intelligent, which is a very efficient way to build a machine that runs smoothly right up until it doesn’t.
We should, perhaps, take a lesson from the night sky. The stars do not hurry. The laws do not negotiate. If you overload a structure, it holds—until it doesn’t—and the moment it fails is not evidence that the structure was suddenly mismanaged; it is evidence that the margin was quietly removed long ago. The question is not whether humans can be pushed harder. They can, for a time. The question is: what are you borrowing against, and who will pay, and will you even recognise the payment when it comes due? If the point of optimisation is to serve human aims—accuracy, reliability, fairness, safety, meaning—then at what moment did the aims become subordinate to the optimisation itself? When did the map become more important than the territory? When did we decide that a system that produces numbers quickly is better than a system that produces numbers that can survive contact with reality?
If you’re tempted to answer, “We have no choice,” consider that “no choice” is often the most expensive story an organisation tells itself. It is a story that stops curiosity, and curiosity is the only force I know that reliably prevents brittle systems from becoming monuments. The universe is patient. Reality keeps excellent records. And it will, eventually, reconcile.
