Your time is the only thing you can't buy back

Two things have been rattling around in my head this week. The first is a book I finished, The Almanack of Naval Ravikant. The second is a Harvard commencement speech by comedian Ronny Chieng that went viral and that I've watched a few times now. On the surface they're completely different. One is a quiet, philosophical guide to wealth and time. The other is a profanity-laden comedy set about destroying AI. But they're making the same point from opposite ends, and together they've shifted how I'm thinking about what it means to use AI well.

Naval has a concept that sounds almost laughably simple: decide what your time is worth per hour and use that number as a filter for every decision you make. Pick a number, commit to it, and then ask honestly whether the thing you’re about to spend two hours on is actually worth that. If someone else could do it for less than your rate, pay them. If it doesn’t clear the bar, don’t do it.

The insight isn’t really about money. It’s about how seriously you take your own attention. Most of us, if we’re honest, don’t take it seriously at all. We fill our time with things that feel productive but don’t compound. We say yes reflexively. We consume endlessly without producing anything. We’re not protecting our time. We’re leaking it.

“AI is just going to end up making mediocre people dumber.”

That’s Ronny Chieng, at Harvard, to a graduating class that cheered him for it. The full speech is funnier and sharper than any summary I can give, but the serious point underneath the comedy is real. He cited a 2025 MIT study on cognitive debt, the idea that overreliance on language models doesn’t just make you lazy, it actively erodes the mental muscles you stop using. You outsource the thinking, and over time, the capacity for that thinking quietly atrophies.

His line that stuck with me most was about “minimum escape velocity of intelligence.” His argument was that only people who have genuinely built something real in their heads, actual knowledge, actual judgment, actual taste, will benefit from AI as a multiplier. Everyone else is just accumulating debt they can’t see yet.

Here’s where Naval and Chieng converge. Naval’s hourly rate concept is ultimately about protecting your cognitive output. When you decide your time is worth $500 an hour, you’re not just making a financial calculation. You’re making a statement about what your thinking is worth and what you’re willing to let other people or other systems do in its place.

The danger with AI isn’t that it does things for you. It’s that it does things for you that you should be doing for yourself, and over time you lose the ability to tell the difference. You start outsourcing judgment. You stop forming opinions because the model forms them faster. You use it to summarise the article instead of reading the article, and then wonder why you feel vaguely hollow when someone asks what you think.

That’s cognitive debt. And like financial debt, you don’t feel it accumulating. You only feel it when you need to draw on something that isn’t there anymore.

"AI can be the fuel. But fuel is useless if you can't kindle the fire."

Chieng's line. And it's the right frame. The question isn't whether to use AI. Of course you should use it. The question is whether you're using it to amplify something real, or to paper over the absence of something real. Those two things look identical from the outside and feel completely different from the inside.

I’ve been running a daily digest through Claude, a system that pulls signals from GitHub, Indie Hackers, Hacker News, and a few other sources and lands in my Notion each morning as a curated briefing. It’s one of the most useful things I’ve built into my routine this year. But I’m deliberate about what I use it for and what I don’t.

I use it to surface what’s worth paying attention to. I don’t use it to tell me what to think about those things. I use it to compress the noise so I can spend more time with the signal. I don’t use it to skip the part where I actually sit with an idea and form a view.

That distinction matters. The routine saves me hours. But the thinking at the end of it is still mine, and I think that’s the only way it compounds.

Naval’s book and Chieng’s speech are both, in different registers, making a case for substance over shortcut. For building something real in your head before you reach for a tool to extend it. For treating your attention as the scarce resource it is, not because it makes you more productive, but because it’s the thing that makes you, you.

We are in a moment where the fire hose is on full blast. More signal, more noise, more tools promising to help you process it all faster. And some of them genuinely do. But the ones that help you think faster are only valuable if you were thinking in the first place.

Protect the thinking. That’s the one thing you can’t outsource and actually get back.