Embrace your Laziness in the Age of AI

I have a confession: I’m a bit lazy. Not dysfunctionally lazy. But definitely a bit.

Early in my career, I thought this was a problem. There’s so much more I could do if I weren’t lazy! But, over time, I realized it’s the opposite: My laziness is an asset. It’s a regularizer.

When you write code, you have an infinite blank canvas. You could build anything. This is pretty distinct from lots of other domains: Code is inherently less constrained than, say, the work of a mechanical engineer because you don’t have to fight the constraints of physics every step of the way (of course a computer is still physical, so there are some real constraints; but you get the point). This is fantastic. But it’s also daunting.

Enter AI agents. These things are great and I use them all the time. The amazing thing is you can build anything now. The trouble is also that you can build anything now. They never get tired, they never get frustrated, they are infinitely patient and motivated. In other words, they reduce the execution cost of building something towards zero.

So a natural question arises: What should you build? This was always a question, of course, but AI agents make it a lot more urgent.

When it comes to deciding on what to build and how, taste is an important concept. A lot has been written about this. For example, there’s Paul Graham’s famous “Taste for Makers” essay and John Schulman’s excellent guide to ML research, which identifies taste as a critical skill as well.

But what they don’t talk about is laziness. Taste is about quality and direction and simplicity and beauty. Laziness is about restraint and friction and effort. Laziness regularizes taste—taste without laziness gives you beautifully overengineered systems nobody needed. Also, tastelessness without friction is by far the worst: that’s slop.

Remember when you had to write all code and do all research yourself and you went “ugh”? Because you really didn’t feel like doing something? That’s healthy! That “ugh” carries information. It’s telling you: “This is going to be work.” And it’s asking: “Is this work really worth doing or should I just… not?” The answer might be yes or no, but it really matters that this question gets raised in the first place.

AI agents remove this natural friction. They are so fast. They never stop. But they have a price: You have to remember your very human laziness. You have to remember to ask yourself: “Should I really build this?”—or should you just be lazy today.