intelligence is getting cheap. curiosity isn’t.

i grew up thinking intelligence was one of the rare things.

being able to understand something difficult felt valuable on its own. knowing how to code, solve a hard problem, read some horrible documentation or spend three nights figuring out why something worked made you useful because most people either couldn’t do it or didn’t want to.

that assumption is starting to feel outdated.

in september, anthropic had claude formalize fermat’s last theorem in lean. it wasn’t discovering fermat from scratch, wiles already proved it decades ago, but turning that proof into a complete computer-checked formalization was supposed to be an enormous project. claude did it largely autonomously in eleven days and produced around 13 million lines of lean1.

around the same time openai published a claimed solution to the navier-stokes millennium problem produced by thousands of cooperating ai agents, with the final result formally checked in lean2. a year earlier gemini had already reached gold-medal level at the international mathematical olympiad, solving five of the six problems in natural language within the competition time limit3.

these were things we used to point at when trying to explain what human intelligence looked like at its absolute edge.

now they’re becoming benchmarks.

i find that both incredible and slightly depressing.

because if intelligence keeps getting cheaper, being smart stops being much of an identity. knowing the syntax doesn’t matter much when a machine writes it better. remembering more facts doesn’t matter when retrieval is basically free. even being able to reason through difficult problems starts becoming less special when you can rent an absurd amount of reasoning for the price of lunch.

but the models still need somewhere to point.

they can spend billions of tokens proving something once somebody decides that this is the thing worth proving. they can search thousands of approaches at once. they can read more papers in an afternoon than i will in my entire life.

none of that answers why you cared enough to look there in the first place.

i think curiosity becomes more valuable as intelligence becomes abundant. noticing something weird. asking the question nobody thought was worth asking. opening the obscure link. wondering what happens if you connect two completely unrelated ideas. spending six hours on something with absolutely no obvious economic value because it bothers you that you don’t understand it.

that part is difficult to benchmark.

and maybe that’s why i care about it so much.

i don’t really want to compete with machines at being intelligent. that seems like a miserable competition and one we’re probably going to lose anyway.

i’d rather remain difficult to predict.

Footnotes

  1. anthropic, formalizing fermat’s last theorem, september 2026. ↩

  2. openai, on the navier-stokes millennium prize problem, september 2026. ↩

  3. google deepmind, advanced version of gemini with deep think officially achieves gold-medal standard at the international mathematical olympiad, july 2025. ↩