This post was drafted by an AI in under ten seconds. The ideas behind it took a ten-minute conversation to surface.
That ratio is the point.
When calculators arrived, nobody argued that humans should keep practising long division. The skill just quietly stopped mattering. We redirected our energy to the things calculators couldn’t do — formulating the problem, knowing which calculation to run, interpreting what the answer meant.
The same thing is happening to knowledge work right now, just slower and less obviously. Drafting prose, summarising research, structuring a slide deck, writing code — these are becoming arithmetic. Not worthless, but not where humans add value anymore.
The question this forces is: what’s left?
I think what’s left is the conversation. The ten minutes of thinking out loud that produced an insight no model would have reached unprompted. The moment where you notice something feels off, or connect two things that don’t obviously belong together, or push back on a confident answer because your experience says otherwise.
The human job is shrinking to the parts that feel like play: forming a view, recognising what matters, asking the question that reframes the problem. Everything downstream — the drafting, the formatting, the research, the polish — is calculator work.
This isn’t a loss. Nobody mourns long division. But it does change what’s worth getting good at. Practising prose speed is like practising arithmetic speed — you’re competing with something that will always be faster. Practising the thinking that precedes the prose — the noticing, the connecting, the having-an-opinion — that’s the thing that compounds.
The people who struggle with AI aren’t the ones who can’t prompt well. They’re the ones who never developed the upstream skill — the play — because the downstream work kept them busy enough to feel productive.
Find the thing only you can do. Let the calculator handle the rest.