They say AI is coming for the work itself. In a recent interview, OpenAI co-founder Greg Brockman was asked how much of OpenAI's own code is now written by AI. His answer: it's hard to know what percent is not.
What's load-bearing:
the assumption that producing the output is the valuable part — that if a machine writes the code, the report, the analysis, then the human contribution is gone. Authorship as the unit of value.
But notice what his answer actually reveals: at the frontier, nobody is counting anymore. The question quietly shifted from "who wrote it" to "who decided it was right." Production became abundant. Judgment didn't.
So what:
when output becomes nearly free, what buyers pay for moves up a level — to the people who can tell good from plausible. Every knowledge profession is about to discover that its real product was never the deliverable. It was the codebook: knowing what good looks like before the market does.
The professionals at risk aren't the ones whose tasks AI can do. They're the ones who never built a way to show their judgment.
(They Say — a new series: one claim in circulation, its load-bearing assumption, and what follows.)
Source: