Taste Is the Last Edge
For most of my life taste sounded like a soft word. The thing people reach for when they can’t measure something. Engineers distrust it because it never shows up in a benchmark. I’ve come to think taste is the hardest, most valuable, most underpriced skill there is, and the age we’re walking into is about to prove it.
Definition first, because the word is slippery. Taste is the ability to tell, fast and without a rubric, which of two things is better. Which design is cleaner. Which idea is worth chasing. Which sentence lands and which is filler. Which feature matters and which is bloat dressed up as progress. It’s judgment that runs faster than analysis. You see it before you can fully explain it, and good taste is also being right when you finally do explain it.
Why it’s about to be the only edge that matters. When execution was expensive, taste was a luxury. Knowing which thing to build didn’t matter much if building anything took a year, because the cost of building swamped everything else. You could be mediocre at choosing and still win on grind. But as the agents drive execution toward zero, the bottleneck moves entirely to choosing. When you can build anything, the only question left is what’s worth building, and that question has no benchmark, no rubric, no agent. Pure taste. The constraint isn’t capability anymore. It’s discernment.
The uncomfortable bit: taste can’t be downloaded. You can’t read a book and have it. It comes from one place. A huge volume of attempts, each judged against reality, the feedback slowly recalibrating your internal sense of better and worse. I built taste in distributed systems by being wrong about architectures and watching them fall over in prod. I built taste in short-form by posting hundreds of times, watching most of it die and a few things spread. No shortcut through either. The reps were the curriculum. You buy taste with time and embarrassment, and it accepts no other currency.
Which is exactly why it’s such a durable edge. Anything cheap to acquire is no edge. If you can buy it in a weekend, so can everyone. Taste is expensive in the one way that can’t be arbitraged away. Years and a thousand judged attempts. No API for it, and there won’t be. The agents will execute any taste you hand them flawlessly. They can’t supply it. They’ll happily build the wrong thing with perfect craft, forever, until a human with judgment tells them which thing was worth building.
So I stopped treating taste as a soft afterthought and started treating it as the actual asset, the thing worth the most deliberate practice. Every time I choose, I’m training it. Every time I default to a rubric instead of building the judgment to not need one, I’m letting it rot.
In a world where everyone can build anything, the person who knows what’s worth building wins. That person isn’t the best engineer. They’ve got the best taste. Go get reps.
One of a series of essays. I’m Prajjwal Chittori. prajjwalchittori.com.