The architect's hardest job is saying no to good ideas.

Most people remember The Mythical Man-Month for Brooks's Law, adding people to a late project makes it later. But Brooks said the most important idea in the book is a quieter one: conceptual integrity. A system should feel like it came from one mind, even when a hundred people built it. He argued that's the single most important thing in design, more important than any individual feature.

And the way you protect it is uncomfortable. Brooks says you need a chief architect who acts as the user's agent and decides what goes in and harder, what stays out. Not because the rejected ideas are bad. Often they're genuinely good. They get left out because they don't fit the whole. A system, he argues, is sometimes better with fewer features, because coherence is worth more than capability. A product nobody can understand is a product whose features go unused.

That reframed the architect's role for me. The instinct, especially from strong engineers, is that architecture is about adding the smartest solution to each problem. But the real work is the opposite. It's holding a single, coherent vision and being willing to turn away clever, well-meant ideas that would each improve a part while quietly fracturing the whole. Anyone can say yes to a good idea. The architect's harder job is saying no to one, in service of something only they can see: the integrity of the entire system.

I've felt this most in the SDKs and frameworks I've designed largely on my own at Hubtel. When a piece comes from essentially one hand, even with good help around the edges, it has a coherence that's hard to fake, a consistency in how things are named, how they behave, what they refuse to do. The ones shaped by a committee, everyone adding their preferred pattern, tend to feel like a patchwork, however individually clever each contribution was. Conceptual integrity isn't a luxury you get after the features. It's the thing the features hang on.