Taste is knowing what to leave out

Part of our series on taste and shipping fast.

At Apple's developer conference in 1997, freshly back at the company, Steve Jobs was asked why Apple had killed a technology developers liked. Part of his answer has been quoted ever since: focus means "saying no to the hundred other good ideas".

Note the word good. Saying no to bad ideas is easy. Taste shows when you say no to good ones, because they are not the one thing your product is for.

Why AI makes this harder

When a feature cost a month, the cost itself said no for you. Most ideas died in planning because nobody had time to build them. Now a feature costs an afternoon and a prompt. Nothing stops you any more, so everything gets built: a second dashboard, a settings page for a setting nobody asked for, an AI summary on top of a list of five items.

Each one is small. Together they bury the thing your product does well under things it does adequately. Users don't experience your roadmap. They experience the product, and every extra option is something they have to look past.

Good taste is real

Paul Graham argued in Taste for Makers (2002) that good taste is not just opinion: if there were no such thing as good design, there would be no way to get better at it. He later came back to the question in Is There Such a Thing as Good Taste?. Among the marks of good work he lists is simplicity, solving the right problem and nothing more.

A practice: the not-doing list

Next to your roadmap, keep a list of what the product will not do, and why. Write the good ideas there too. When someone (including you, at 11 pm with an agent open) wants to add one, the list is where the argument starts.

• Every item names the person it would serve and why that person isn't who you are building for yet.

• Read it before every planning session.

• Promote an item only when real users ask for it, in their words, more than once.

The cheapest code to maintain is the code you never asked a model to write.