Something I can't stop noticing is that nothing works as well as it used to, and I don't mean that in the back-in-my-day way. I mean it structurally, across systems that have no connection to each other.
Airlines cannot keep to their own schedules. Hospitals fall over under ordinary, predictable demand. Software that used to be solid is now a subscription shell that gets buggier with every forced update. Government offices take months to do things that should take an afternoon. New buildings cost more than the ones they replaced and are visibly worse. Customer service has become a chatbot maze built to wear you down until you stop asking. This is not me misremembering a golden age. The infrastructure is aging faster than anyone is replacing it, the people who actually knew how the systems worked are retiring without handing the knowledge to anyone, and somewhere along the way the running of almost every institution passed to people who treat making a good thing as a cost center and treat the quarterly metric as the actual product.
Start with money, because it sits underneath the rest. Once a company's real job quietly shifts from serving the customer to serving the shareholder, quality stops being a standard to hold and becomes an expense to trim. Boeing did not forget how to build an aircraft. It was managed into forgetting, decision by decision, by people who looked at engineering and saw a line item rather than the whole reason the company was allowed to exist. Then there is the credential problem, where we built a world that demands a four-year degree for work a focused six months would cover, and then act surprised when the graduate cannot do the thing. The university turned into a credential vending machine that calls itself a school, and the actual competence still gets transferred on the job, which is a problem when the entry-level job that used to do the transferring is the first thing to get cut.
The deeper one is the war on slack. Every system has been squeezed to wring out spare capacity, redundancy, and buffer, which makes it look wonderfully efficient on a normal day and turns it into glass the moment anything goes wrong. A hospital running at ninety-five percent looks lean right up until a bad flu season, never mind a pandemic. A supply chain with no inventory looks brilliant right up until one ship wedges itself across the Suez Canal and the whole planet feels it. We stripped every system down to its theoretical minimum and then performed shock when it could not absorb the smallest shock. And running through all of it is Goodhart's law, the old observation that the moment you turn a measure into a target it stops being a good measure. Schools chase test scores instead of learning. Hospitals chase throughput instead of health. Police chase arrest counts instead of safety. The platforms chase engagement instead of anything you would actually want from them. Measure the wrong thing and you get more of the wrong thing, delivered with real precision.
What turns all of this from a list of complaints into an actual crisis is that the failures feed each other. An institution that loses competence loses trust, and once it has lost trust the capable people stop wanting to work there, and once they leave the competence drops further and the trust drops with it, and the loop tightens. Government is the cleanest example. Public pay has trailed the private sector for decades, so the strong engineers and analysts and managers go to tech or finance, and the agencies are left with a thin layer of genuine public servants slowly burning out among everyone who had nowhere better to go. Then the government fails at something, which it does more and more, and the response is to cut its budget, which makes it fail harder, which gets taken as proof it deserved the cut. Nobody with the power to break that loop has any incentive to.
The obvious question is whether AI breaks the pattern or deepens it. You can tell an encouraging story where it patches over the missing competence: the model helps the overwhelmed clerk fill the form correctly, catches the defect the undertrained inspector missed, holds the institutional memory that would otherwise walk out the door at retirement. I don't believe that is the story we are going to get. I have watched how organizations actually decide things, and they will reach for AI as a reason to invest even less in people. Why train anyone properly when the model can cover the gap. Why keep the expert when a junior with a chatbot is cheaper. That works right up to the first situation the model was never trained on, at which point you discover there is no longer anyone in the building who understands the system well enough to step in. The bet here is on management incentives, and management incentives I trust completely to behave badly.
Competence is unglamorous, which is part of why it goes uncelebrated until it is gone. It is the bridge inspector who actually looks at every joint. The nurse who catches the drug interaction because she read the whole chart. The programmer who writes the test before shipping. The clerk who chases down the one application that slipped through the cracks. It is held up by culture, by training, and by institutional memory, and right now all three are being eaten at once, culture by financialization, training by cost-cutting, memory by churn. The societies that work out how to protect and reward the boring, real version of competence are the ones that will still function in twenty years. The ones that keep optimizing the metric while the underlying capability quietly rots will be the case studies. We are making that choice right now, mostly by refusing to admit there is a choice to make.