Every time the displacement conversation gets uncomfortable, someone raises the retraining argument. Workers displaced by AI will learn new skills. They will move up the value chain. History shows that technology creates more jobs than it destroys. We have been here before and we got through it. I have been hearing this for twenty years and I want to be direct about what I think it is: a way of acknowledging a problem while declining to think about it seriously.
The historical argument is the weakest part. Yes, previous waves of automation displaced workers who found new roles. The industrial revolution did eventually produce more employment than it destroyed. But the mechanism that made that work was that humans remained necessary. There was always a ceiling the technology had not reached, a domain where human judgement, dexterity, or creativity was still required. Workers moved up because there was somewhere up to go. The retraining argument imports that assumption into the present without examining whether it still holds. The question it does not ask is: retrain into what, exactly, when the competition is a system capable of learning any domain faster than a human can?
I am a software developer. I have been watching AI eat into the work I do for several years now. The domains that were supposed to be safe, complex reasoning, creative problem-solving, professional judgement, are the ones being compressed fastest. The advice to "move up the value chain" assumes a stable position above the waterline. The waterline is rising. I am not describing this from a comfortable distance; I am describing the conditions of my own working life, and I think it is dishonest to pretend the professional class is somehow insulated from what is happening to warehouse workers and call centre staff. We are all on the same trajectory, at different points along it.
The retraining argument also has a quiet class assumption embedded in it. When a factory closes and workers are told to retrain, the implicit message is that their predicament is a personal problem with a personal solution: acquire new skills, adapt, compete. This framing places the burden of a structural shift onto the individuals it is displacing. It is convenient for everyone except the people being displaced. A fifty-year-old logistics worker whose job has been automated does not need career advice. She needs an economy that has thought seriously about what she is going to live on. The retraining argument is what you say when you want to appear constructive without doing the harder work of answering that question.
What frustrates me most about the argument is that it is unfalsifiable in practice. When you point out that the promised new jobs have not materialised for a given group of workers, the response is that the transition takes time, that we need to be patient, that the new roles are coming. This can be said indefinitely. It was said about manufacturing communities that have not recovered in thirty years. It will be said about the current wave of knowledge-work displacement for as long as it is convenient to say it.
I am not saying there are no people who successfully retrain. There are. I am saying that "some people successfully retrain" is not a policy response to mass structural displacement. The question is what happens to the large majority of people for whom retraining is not a realistic option, whether because of age, circumstance, the pace of change, or simply the fact that the roles they are supposed to retrain into are also being automated. That question deserves a direct answer, and the retraining argument is specifically designed to avoid giving one.
The US tech layoff numbers climbing through the first quarter of 2026, explicitly attributed to AI efficiency, are not an anomaly. They are the beginning of a pattern. The honest response to that pattern is not to tell the people affected that the answer is a coding bootcamp. It is to grapple seriously with what an economy looks like when human labour is systematically optional across an expanding range of domains. We do not yet have good answers to that. But we will not find them while we are busy pretending the retraining argument is one.