A number of things are currently deteriorating in ways that do not seem connected. Many professionals are quietly losing income without being laid off. The AI revolution is supposedly transformative, yet productivity barely moves. House prices in big cities keep rising even as people feel poorer. Voters drift towards authoritarian and anti-system parties.
Looked at separately, each of these developments has a tidy explanation. Looked at together, they tell a single story.
I have watched one part of this story unfold at close range. My wife works as a lexicographer and translator, professions that were once specialised, scarce and well paid. When AI systems became able to do most of the work adequately, instantly and at negligible cost, there were no mass layoffs or militant strikes. Instead, work thinned. Contracts shortened. Rates fell. Gaps between assignments grew.
People did not protest. They adapted. They drew on savings, relied on a partner’s income or quietly lowered expectations. They never talked about it publicly, because saying “I’m not getting any work” is bad business. The damage was absorbed privately, person by person. Society barely noticed.
This is the first strand of the story, and it is the quiet one. Journalists, translators, editors, designers and similar groups have not been marched out of offices or added to unemployment statistics. Instead, permanent jobs have turned into freelance work. Freelance work has turned into intermittent work. Income erosion has been gradual, individualised and largely invisible. Unlike factory closures in the past, there has been no single moment of rupture. The shock has been absorbed in silence.
Recently, OpenAI introduced a benchmark called GDPval, designed to measure how well general-purpose AI systems perform economically valuable professional tasks. Until very recently, scores were low enough to be dismissed. Over the past months, they have risen sharply into a range that is no longer trivial.
This does not mean that AI will suddenly replace professionals. It means something more subtle and more important. Substitution has become credible. Tasks that could not plausibly be automated a year ago are now within striking distance. In markets, that is enough. Once buyers believe alternatives exist, prices fall and bargaining power evaporates long before full replacement occurs.
What we are seeing now is not the end state. It is an early phase, concentrated in professions where work can be modularised and priced task by task. But the mechanism is general. The difference now is scale. What once affected a small, easily ignored group of language professionals is becoming relevant to business and professional work more broadly.
This leads to the second strand, the productivity puzzle. For some time, people have asked why all this new technology has not shown up in the numbers. The usual explanations are that productivity gains are delayed, mismeasured or waiting for organisational change. That may be partly true, but it misses something crucial.
So far, AI has not mainly increased output per hour. It has reduced the price of certain kinds of output. The same texts are written. The same analyses are produced. But the value attached to them collapses. From the point of view of productivity statistics, little changes. From the point of view of incomes, everything does.
This helps explain the third strand, the political one. Large groups of voters are reacting strongly to changes that economists barely see. This is often explained away as cultural anxiety, nostalgia or irrational anger. There is a simpler explanation. People are responding to economic signals that do not yet register on official dashboards.
If your income is shrinking, your career no longer progresses and your buffers are being quietly eaten away, it matters little that GDP is stable or employment is high. You feel downgraded. When mainstream politics insists that everything is fine, while offering solutions that do not address what you have lost, people look elsewhere for someone who at least acknowledges the loss. That does not make the resulting politics good. It does make it intelligible.
The fourth strand is housing, and it is more dangerous than it looks. For many Gen X professionals, rising house prices have not merely been a background asset boom. They have been the mechanism that made income erosion survivable.
People who bought property in major cities in the 1990s or early 2000s accumulated large amounts of paper wealth as prices rose. When professional income began to thin, that wealth could be tapped. Remortgaging, extending loans and releasing equity allowed past gains to be converted into present cash flow.
This created the appearance of resilience. People stayed afloat. Mortgages were serviced. Lives continued more or less as before. From the outside, nothing seemed broken. But this resilience was conditional. Housing did not absorb the shock. It postponed it.
Each round of remortgaging reduced future flexibility and increased dependence on stable prices and low interest rates. Crucially, this often happened after liquid savings had already been run down. By the time equity was being used to support day-to-day life, other buffers were gone.
If house prices were to fall meaningfully in this context, the impact would be far more severe than many expect. These households would not be facing a housing shock on top of strong incomes and intact savings. They would be facing it with weakened cash flow, depleted reserves and limited time to recover.
This is what makes the risk systemic. A single generation has been using the same asset to compensate for stalled careers and falling professional value. If that asset fails, the adjustment does not spread gently across society. It concentrates sharply among people who already feel economically downgraded and politically unheard.
The fifth strand ties everything together: value. As informational and cognitive outputs slide towards zero marginal cost, value does not disappear. It moves.
It reanchors in things that cannot be cheaply replicated or easily shared: land, energy, infrastructure, location, platforms and ownership of the systems that deploy the machines. Apps are cheap. Content is cheap. Expertise is becoming cheap. Housing in the right place is not. Energy is not. Access is not.
This is why so many policy responses feel misaligned. Retraining assumes there is a clear new ladder. Modest basic incomes assume the problem is the absence of a floor. But for many people, the shock is not poverty. It is the loss of status, trajectory and bargaining power. The floor is still there. The ceiling has collapsed.
Once these strands are seen together, many puzzles stop being puzzles. The AI revolution is real, but it shows up first as value compression, not rising productivity or mass unemployment. Voters react because lived experience changes before statistics do. Housing becomes dangerous because it concentrates delayed risk. Value flows away from labour not because people stop working, but because scarcity has moved elsewhere.
We already know what happens when AI can do most of a profession’s work well enough, instantly and at negligible cost. It does not trigger mass unemployment. It quietly destroys value. What GDPval suggests is that this pattern is no longer confined to a few language professions, but is becoming a general condition for business and professional work.
A society can only solve problems it understands. At the moment, we are still misreading what is happening. This is not primarily a story about unemployment, nor about a lack of skills, nor about people falling below a subsistence floor. It is about the quiet loss of value, status and bargaining power among people who were previously scarce, experienced and economically secure.
The people affected are not easily retrained. They already have decades of specialised experience. Nor are they likely to be satisfied with a modest basic income that does not service the mortgage or preserve a sense of trajectory. What has been lost is not work as such, but the conditions under which work once translated into security and status.
Until this is recognised, responses will continue to miss the mark. We will keep offering ladders where there is no longer a wall to climb, and floors to people who have not fallen through them. The first step in addressing any of this is understanding what is actually happening. Right now, that understanding is still lagging far behind lived experience.