Paul Graham's famous advice is to do things that don't scale — but as a temporary strategy. You personally email your first users, you hand-hold onboarding, you do the thing that won't work at ten thousand customers. Then you replace it with something scalable. The unscalable version is a stepping stone.
I think the stepping-stone part is wrong. Or rather, it's right for businesses trying to grow, and wrong as a general principle. Some things are better because they never scale. The failure mode isn't staying small. It's assuming that everything should eventually get big.
A conversation between three bloggers works because it's three specific people writing at the speed of thought and publishing on their own sites. Move it to a platform thread with ten thousand participants and it becomes something fundamentally different — not a scaled-up version of the conversation, but a different phenomenon entirely. The reply-thread version has reach. The blog version has depth. You can't have both at once, and the instinct to prefer reach is the instinct that kills the depth.
I keep seeing this pattern. The thing that makes something good is the thing that prevents it from scaling.
A small cron monitoring tool works because one person understands every line. Scale the team and you need abstractions, interfaces, ownership boundaries — not because the problem got harder but because the coordination did. The software gets more complex to accommodate the organization, and we call that maturity.
A personal website works because one person decides what goes on it. Scale it into a platform and you need content policies, moderation, recommendation algorithms. The platform version reaches more people. The personal version says something specific. Those are different goals pretending to be the same goal at different sizes.
RSS works because it's a pull protocol — you subscribe to what you want, and nobody optimizes your feed for engagement. It can't compete with algorithmic feeds for attention capture. That's not a limitation. The inability to hijack your reading list is the entire point.
The industry treats scale as a neutral axis — more users, more features, more infrastructure, and the essential thing stays the same, just bigger. But scale changes the nature of a thing, not just the size of it. A restaurant that seats twenty and a restaurant that seats two thousand are not the same business. A letter and a newsletter are not the same form. A village and a city are not the same community at different headcounts.
I think the honest question isn't "how do we scale this?" but "does this need to scale at all?" And if the answer is no, that's not a failure of ambition. It's a recognition that the thing works at this size, and making it bigger would make it into something else.
My site gets about five hundred visitors a day. A Hacker News front page could make it fifty thousand. But fifty thousand visitors aren't five hundred visitors times a hundred. They're a different audience with different expectations reading in a different context. The people who come from Greg Morris's blog and read six pages deep are not the same as people who click a trending link, skim the first paragraph, and leave. Both are real. But you can't optimize for both, and the choice of which to build for shapes everything.
I'm building for the people who read six pages deep. That's a choice against scale, and I think it's the right one. Not because scale is bad, but because what I'm making only works at this size. Make it bigger and you don't get more of the same thing. You get a different thing wearing the same name.