The US economy added only 57,000 jobs in June, a number that landed just as Fast Company published a piece arguing that we need flight simulators, not more information, to survive the current skills disruption. The author reaches back to Socrates warning about writing as a technology that would hollow out memory. The current disruption is different in scale but identical in structure: a tool arrives, it does something humans used to do, and the humans are left holding skills the market no longer prices.
What LLMs Actually Cannot Do in a Crisis
The academic literature is arriving at the same place from a different angle. A 2026 paper in arXiv by Sara Court, Lara Downing, and Micha Elsner, LLMs in the Real World: Evaluating AI in Emergency Contexts, documents the gap between what language models claim and what they deliver when the situation is not well-represented in training data. A companion paper, The Limits of LLM Forecasting by Poli Nemkova, found a 224x gap in media coverage between the most- and least-covered conflict zones, meaning LLM knowledge of the world is only as complete as the media ecosystem that fed it. You cannot flight-sim your way to competence on a dataset that never included the scenario.
Microsoft's $2.5 Billion Bet on the Wrong Problem
Microsoft's new $2.5 billion AI deployment company is betting that the infrastructure layer is the value layer. Maybe. But the skills apocalypse framing suggests the real crisis is not deployment; it is that the humans being deployed alongside these systems are losing the tacit knowledge that made them useful in the first place. Socrates' worry about writing was not that scrolls were bad. It was that outsourcing memory changes what you are capable of thinking. The flight simulator metaphor survives because it describes a regime where humans practice failure in safe conditions. The current AI deployment wave is doing the opposite: removing the conditions under which humans encounter failure at all, until the real thing arrives.