A post on Hacker News this week revisited a question that has been gaining urgency as AI capabilities improve: what happens when AI matures and people have little work to do? The thread, from a user who asked the same question three years ago, argued that even AI skeptics now have to acknowledge LLMs are a valuable intelligence resource, and that the conversation needs to move past whether AI will match human capability to what follows if it does.
The framing is worth taking seriously. The poster's argument rests on a mathematical observation: neural networks can approximate arbitrarily complex functions. If human thought is computable, and there is no theorem proving it is not, then some program can replicate it. Even if non-computable physics exists, the poster argues, nothing stops building a computer that exploits those same physical laws. The question is not whether AI will reach human-level capability in most domains, but what the social consequences look like when it does.
The disciplines that resist automation
The poster draws a distinction between fields where the process matters and fields where only the output matters. Mathematics and art fall into the first category. In math, the value of a proof is not just that it is correct but that it reflects human understanding. A proof no one comprehends is not the same as a proof that advances mathematical thought. In art, the value is the human expression itself. An AI that generates a painting is not the same as a person who made one.
Engineering, the poster argues, is different. In engineering, "if it works, it works." The output is what matters, not the process of producing it. A bridge that stands is a bridge that stands, regardless of whether a human or an AI designed it. This makes engineering more vulnerable to automation than fields where the act of creation is the point.
This distinction maps onto a broader pattern. Fields where human judgment, taste, or comprehension are the product tend to resist automation. Fields where the goal is a correct or functional output are more exposed. The poster suggests that even in fields like engineering, where people genuinely enjoy the work, the incentive structure shifts once AI can produce the same output faster and cheaper. The fun does not disappear, but the economic justification for paying a human to do it does.
The problem with the "AI tools free you for important work" framing
The tech industry has a standard answer to job displacement concerns: AI tools will handle the trivial work so humans can focus on the interesting problems. The poster pushes back on this directly. The framing assumes that people want to do things because they are easy or routine, and that removing the easy parts leaves the satisfying parts. But many people want to do things specifically because they are hard, or because no one has done them before.
External incentives matter too. The poster acknowledges that even if intrinsic motivation is real, it is not the only reason people work. Paychecks, career advancement, social status, and the structure of a workday all motivate people. Remove the work, and you remove those incentives alongside it. The "AI frees you for creative work" argument assumes that creativity is what people want to spend their time on, and that is an empirical claim that does not hold for everyone.
The poster also raises a comparison to physical activity. Running used to be transportation. Now it is exercise, a sport, a hobby. Sewing used to be a necessity. Now it is a craft. The suggestion is that intellectual work might follow the same trajectory: what was once a profession becomes a recreational activity, something people do for fun rather than for survival. But that transition assumes people can afford to treat work as optional, which depends on economic structures that do not yet exist.
What the HN discussion missed
The thread received relatively few comments, but the original post contains an honest admission that is rare in these discussions. The poster wrote that they cannot even convince themselves of their own optimistic framing. They want to believe that AI will free people for meaningful work, but on reflection, they are not sure. The external incentives that drive most human activity are real, and removing them without a replacement is a problem that no one has solved.
The implicit challenge is for the programming community specifically. Developers have been the primary beneficiaries of AI coding tools. The tools genuinely do reduce the time spent on boilerplate, debugging, and boilerplate debugging. But the poster's question applies to developers too. If AI can write code as well as a human, the economic value of being a programmer drops. The fun of solving hard problems remains, but the ability to charge for that skill diminishes.
This is not a hypothetical for people building AI tools. The same models that make coding easier are the models that will eventually make many coding tasks unnecessary. The question is not whether this will happen but what the transition looks like, and whether the industry is building the economic structures needed to support people through it. The Hacker News thread did not answer that question, but it asked it clearly enough that the community should be paying attention.