I’ve been running across a lot of people lately who struggle to see where LLMs are useful, whether on social media or in everyday life. Less often, I also run into folks who use them everywhere, even in cases I think are suspect. So I thought I’d write up my current thinking on where these tools are helpful, at least in mid-2026.

The important theme here is that I think they’re best for tasks where the work is expensive but the validation is cheap and easy.

Continue reading →