Every "best uses of AI" list in 2026 looks roughly the same: productivity, coding, drug discovery. Those lists aren't wrong.
But they're mostly describing what AI does for people who already have a lot.
I've spent the last few years thinking about a different set of uses. The health worker in a low-resource clinic who can now consult a diagnostic tool she couldn't have accessed before. The student without a strong teacher who finally has access to one. The farmer making decisions based on better data than she's ever had.
These aren't niche applications. They're the whole point.
Progress here is possible. But it requires being deliberate about who we're actually building for.