October 7, 2026
An elegant AI solution can arrive too late: Addy Osmani on the cost of quality
Choosing among suitable options means weighing deadlines and risk, as well as execution quality.
Addy Osmani
@addyosmani
You need both taste and judgment. Taste helps you understand what is good, and judgment helps you understand the cost of quality. Ask a model to suggest a component or migration plan, and it will quickly produce plenty of options. Most of these drafts look acceptable. Once you have several options that meet the basic requirements, taste alone isn't enough to choose between them. You need to make a decision. I like to think of this as two filters. One is taste, the other is judgment. I think many people confuse them. The first filter is taste. It's an instinct for what makes something good. Taste kicks in before you can explain in words what is good or bad about something. You've probably experienced this: something is bad, but you don't know why until you work through it and can describe the problem. "This was terribly awkward to use" or "that error message was incredibly annoying." Taste develops through experience and practice. It comes from seeing a lot, making a lot and remembering what worked well. One of Kant's important ideas about aesthetic taste was that you can appreciate form even if you have no stake in the outcome. Engineers care much more about the outcome than the form, but the feeling itself is the same. Taste helps you filter out a lot. It lets you quickly discard 90% of drafts. But it isn't enough to choose between the few options that remain. The second filter is judgment. It's an understanding of the cost of quality. The most perfect solution may not be finished by the deadline. The most sensible architecture may require users to migrate while you're still carrying out the final update. Judgment helps you understand how much time and risk good decisions cost users and the team. It comes from domain knowledge and experience with past decisions. People with great taste but poor judgment make beautiful things too late. They make the right thing at the wrong time or the wrong thing at the right time. People with judgment but no taste meet deadlines, but don't create things people love. You need both taste and judgment. Use taste to narrow the options down to a few, then use judgment to choose the one you'll see through to completion.
· 29.9K views
A model can quickly suggest several components or migration plans that meet the basic requirements. Addy Osmani separates the choice into two stages: taste helps rule out poor options, while judgment about costs helps choose among those that remain.
A flawless solution may not be finished by the deadline. A well-designed architecture may require users to migrate in the middle of a final update.
Osmani links the ability to assess these trade-offs to domain knowledge and experience with past decisions.