My Twitter feed these days is mostly full of people talking about how fast they shipped some new thing with AI. Built this in a weekend. Vibe-coded my MVP in three hours. 100x engineers shipping PRs all day. Yes, the AI tools really are that good and they’re exciting. I’m beyond excited about a future where I can build anything I can think of.
But I keep coming back to this nagging feeling that we’re ignoring a really important thing when all we talk about is how AI is enabling speed.
In physics, speed is a scalar. Velocity is a vector — it tells you how fast and in what direction. What does speed actually get you when you’re building a product? Faster feedback. You find out sooner whether you built the right thing or the wrong thing (you probably built the wrong thing).
A team ships something in a few days that nobody asked for. They pivot. Ship something else nobody asked for. Pivot again. Lots of motion. Not a lot of progress. I rarely see a follow-up about whether anyone used it.
AI making everything faster to build is great. But it moves the bottleneck. The bottleneck is not how long it takes to write the code. It’s whether you’re writing the right code. And when building is cheap, building the wrong thing is even cheaper to do and easier to justify.
I’ve been rediscovering my love for guitar lately, and one thing that’s humbling is building speed. Every instinct says to just play faster. But the only way to actually get fast is to practice painfully slow with a metronome. You work on the exact mechanics of how your pick hits the string, how your fingers move, where the tension is. You change these mechanics and adopt totally new techniques when you’re trying to play something that requires it. You do it slow enough that every motion is deliberate. Then you nudge the metronome up a few BPM. It takes forever and it’s frustrating because you just want to rip like the pros.
And then there’s a whole other thing beyond speed. There are players who are technically incredible — they can play anything — but they have nothing interesting to say. And there are players who make you feel something with three notes because they know exactly what they’re trying to express. The technique is always in service of something. Without that something, it’s just noise. AI can generate music that’s technically fine. But nobody puts it on and feels anything. The human part — having something to say and meaning it — is the thing that shouldn’t be glossed over.
Learning to do anything past the intermediate level works this same way. Elite marathoners run 80% of their miles at easy pace so the fast miles actually count. The speed comes from the mechanics, not the other way around.
So move fast and make things. Make things with direction. Make things with good mechanics. Make things that mean something.