Since antiquity, people have worried about technological unemployment. This term, introduced by John Maynard Keynes in his 1930 essay Economic Possibilities for our Grandchildren, refers to loss of jobs caused by technological changes. He writes:

I draw the conclusion that, assuming no important wars and no important increase in population, the economic problem [the problem of scarcity] may be solved, or be at least within sight of solution, within a hundred years.

Now that we have extremely capable AI, the debate about technological unemployment is ever more relevant. Indeed, AIs are already able to perform self-contained tasks with economic value.

METR’s measurements of task-completion time horizon provides strong evidence of models’ economic value. Claude Opus 4.6 had a 50% success rate on tasks that would normally take software engineers 12h. A crude but useful simplification: having two copies of Opus 4.6 is like having a coding intern¹.

AIenation #

While I’m somewhat worried about technological unemployment, I’m more worried about alienation – AI alienating people from their work.

For many jobs will come down to prompting LLMs. Or, if you’re doing something really sophisticated, you’ll make different LLMs coordinate with one another. If you discover you’re more effective at work in the role of an LLM commander, hoarding your army of LLMs, what does this do with your sense of self? So much for those years of specialised training.

What the scholars say #

Marx is always a subject of scholarly debate, so scholars have (obviously) thought about this.

Some argue that labor-replacing AI ultimately could have a liberating effect. In this article, Alexander Sidorkin offers a positive view:

We propose that as AI increasingly takes over both manual and routine cognitive tasks, humans are liberated to focus on uniquely human qualities such as creativity, agency, and the capacity for joy. This transformation is likened to an evolutionary process, where humans shed layers of false humanity tied to productive labor, revealing a more authentic core.

Without having looked deeply into the literature, I’ll offer some quick takes. I think Sidorkin is partially right: I fondly remember my first real vibe coding experience – it was liberating having someone else write boilerplate plotting code. There’s a reason it’s called vibe coding, after all.

However, if we’re forced to ‘vibe work’, it seems like the liberation Sidorkin speaks of won’t happen. Quite the contrary: if SWE job comes down to vibe coding, as it now does, there’s a point where you might feel less like an LLM commander and more like a keyboard slave.

Conclusion #

Keynes was wrong about scarcity being solved by 2030. However, he also acknowledged the risks of labor-replacing technology. In the famous passage where he predicted 15-hour work weeks, he also recognised the human need to work:

For many ages to come the old Adam will be so strong in us that everybody will need to do some work if he is to be contented. […] Three-hour shifts or a fifteen-hour week may put off the problem for a great while.

There’s something profoundly satisfying about doing good, honest work – human work. Humans like like honing a craft. Historically, work has provided just that, at least for most people²; if AI makes work feel meaningless for a large part of the population, this might cause social disruption.

Needless to say, we need a renewed discussion about the concept of alienation. Even if AIs don’t steal people’s jobs, leading to technological unemployment, they might steal their sense of self.