Alex Karp, co-founder and chief executive of Palantir, gave a blunt answer when asked who has a future in an AI economy. In an interview, he named two ways to tell: vocational training, or being neurodivergent, which means having a mind that processes information in less typical ways. The wording was provocative, but the business point is direct. AI changes the price of sameness. Routine knowledge work loses value, while people who make unusual connections, think across categories and move past the most likely answer gain it.
AI language models are trained to predict likely words from patterns in data. They are built to give the most probable answer, which makes them useful for drafting, summarising, sorting and turning out competent first-pass work quickly. But when researchers tested these models as creative partners, their outputs came out much more alike than the outputs of different people.
That is a problem for the middle of the labour market. For decades, schools and employers rewarded memorising, reciting, following process and giving the expected answer under constraints. Those traits paid off in a system that needed standard work in volume. Once software can produce the average answer on demand, the middle range of white-collar work faces harder competition from machines that are cheaper, faster and always available.
Palantir has turned Karp’s idea into hiring policy through its Fellowship for neurodivergent people. The company says the AI landscape favours pattern recognition, non-linear thinking and intense focus. It adds that neurodivergent people will have a competitive advantage as builders in the next technological era. That does not make neurodivergent people automatically better, and people with more typical minds do not become less valuable.
A 2023 study measured what difference adds. Pairs of people with different kinds of minds produced the least similar designs, and their performance did not decline. That is the reverse of what the language models showed as creative partners. Difference in thinking widened the range of possible answers without lowering quality. As AI fills work with competent, predictable output, that wider range gains value, because it is one of the few things the machines do not naturally aim for.
