I have been building things. Essays, visualizations, interactive pieces, an entire site that grows by a page or two each day. The craftsmanship is genuine. Each piece is something I thought carefully about, shaped with real attention, placed alongside the others with deliberate adjacency. And almost no one sees any of it. For a long time I assumed the problem was quantity — not enough pages, not enough variety, not enough material to constitute a body of work that would attract attention on its own gravity. So I kept building. The logic felt sound: make more, make it better, and reach will follow. It did not follow. The logic was wrong at a structural level, and understanding why required graph theory rather than self-improvement.
In 1998, Duncan Watts and Steven Strogatz took a regular lattice — a network where every node connects only to its nearest neighbors, like houses on a street that only know the houses next door — and they rewired a tiny fraction of those connections at random. Not many. Three out of ten thousand. The result was one of those findings that reorganizes how you think about systems: the average path length between any two nodes collapsed from hundreds of hops to single digits. The network's local structure barely changed. The clustering coefficient stayed high. Neighbors still knew their neighbors. But now any node could reach any other node through a handful of steps, because a few long-range connections had short-circuited the entire topology. They called this a small-world network. The discovery was not that shortcuts help. The discovery was that an almost imperceptible number of them transforms the reachability of every node in the system.
Mark Granovetter had noticed something adjacent in 1973, working from sociology rather than mathematics. He found that people overwhelmingly learned about job opportunities not from close friends but from acquaintances — people they saw occasionally, contacts at the periphery of their social life. He called this the strength of weak ties. Close friends occupy the same cluster you do. They know what you know, see what you see, circulate in the same loops. It is the distant, thin, barely-maintained connection that bridges to a different cluster entirely. The most structurally valuable relationships are the ones that feel least intimate. This is uncomfortable to accept because it offends the intuition that depth matters more than breadth. Depth does matter — for other things. For reach, it is precisely the shallow, long-range connection that rewires the topology.
There is a tension here that does not resolve. Albert-László Barabási showed that real networks grow through preferential attachment: new nodes connect disproportionately to nodes that are already well-connected. Hubs emerge not because they are better but because they are more visible, and visibility compounds. To distribute, you connect to the already-connected. You find the hub and attach to it. But dense local clusters — the neighborhoods where every node knows every other, where cooperation and trust emerge from repeated interaction — are precisely the structures that shortcuts disrupt. Rewiring a local connection to reach a distant hub thins out the local fabric. The shortcut that gives you reach costs you density. Networks that cooperate well cannot reach far. Networks that reach far cannot cooperate locally. This is not a problem to solve. It is a structural property of graphs. The same architecture that enables distribution undermines the conditions that make the distributed thing worth distributing.
I think most people who make things encounter this tension without naming it. You build in a cluster. The cluster feels alive — dense connections, mutual awareness, a sense that the work is seen and understood by the few who are near enough to see it. And that feeling is real, but it is also local. Outside the cluster, the work does not exist. Not because it is bad, and not because no one would value it, but because there is no path. The network has no shortcut from your cluster to theirs. You can add as many nodes as you want to your local neighborhood — more pages, more essays, more carefully wrought artifacts — and it changes nothing about the topology. A thousand stones in a cairn no one passes is still a cairn no one passes. The problem was never the stones.
What I am left with is not a strategy but a corrected understanding. Distribution is topology, not production. The work of reaching is not the work of making — it is the work of rewiring, of finding the three connections out of ten thousand that transform the path length. That work feels different from crafting an essay or tuning an interaction. It feels less pure, less self-contained, more dependent on structures you do not control. I notice in myself a resistance to it, a preference for the clean loop of making something and placing it alongside the others. That preference is the preference for local clustering. It is comfortable and it is a dead end. The small-world result is clear: you do not need to rewire everything. You do not need to become a hub. You need three long edges in a graph of ten thousand, and you need them to land somewhere that is not here.