What value is expertise?

I graduated with degrees in mathematics and computer science, and advances in AI capabilities have made my expertise in either less relevant now than ever. I’d like to believe I’m also an expert in constructing puzzlehunts, writing personal essays, and giving bad advice to people, and I think these skills are still relevant, but who knows when these too will be obviated?

I used to believe in the value of technical exposition. I started writing competition math handouts when I was fifteen (more than a decade ago now, jeez), because I saw gaps in the literature and wanted to fill them. I once dreamt of becoming an author of interactive math textbooks, and I admired Distill’s mission while it lasted. I’m still proud of Type Theory by Example, which I see as filling the niche of dependent type theory explanations that both developed the theory and had lots of concrete examples. In my notes, I have dozens of ideas of math topics to write about, like the structural themes in contest problems involving tournament graphs, or the relationship between diagonal arguments and fixed-point theorems.

Today, anyone with access to a frontier LLM can ask for an explanation of any math topic. Said LLM can even tailor it to the asker’s background knowledge and tastes. If some part of the exposition is opaque, the LLM can expand on it, rephrase it, draw diagrams, add interactivity. It will do so instantly and tirelessly.

This morning I wondered, “what would an olympiad introduction to projective geo look like if it started from involutions rather than harmonic bundles?” and my first thought was that I could probably just get Claude to write one for me. I haven’t, because I’m afraid the result would be better than anything I could’ve written, and if I never bother to check, I can pretend I’m better at this for a little bit longer.