I was reading about network science not for any particular reason, and I came across the concept of hubs. It sounds really simple on the surface, but the more I searched and tried to understand it deeply, the more I kept thinking about the internet, and Google, and then LLMs, and I got into a chain of thought which I felt like I should write down somewhere.

This is that somewhere. (welcome to narang.dev)

What Even Is a Hub?

In network science, a network is just a collection of nodes connected to each other by edges. Nodes can be anything like websites or cities or people. Edges are just the connections between them. That’s really it.

Most nodes in any network have a modest number of connections, but every once in a while there’s a node that has an oddly massive number of connections compared to everyone else. That’s a hub.

The easiest way I could picture this was airports. When we try to book a flight from Indore to Coimbatore we almost never fly direct, we go through Mumbai or Delhi first and then to wherever we’re headed. Mumbai and Delhi are hubs in that sense. A huge chunk of India’s air traffic flows through them not because someone sat down and planned it that way but just because of how the network grew over time.

And this isn’t just an India thing. Dubai’s airport is probably the most extreme version of this globally. A massive portion of international flights between Asia, Africa and Europe just pass through Dubai. Not because Dubai is the destination but because overtime it became the hub everyone routes through.

The Rich Get Richer

The natural question after understanding what a hub is, is why do they form at all? Why doesn’t every node just end up with roughly the same number of connections?

New nodes don’t connect randomly. They connect to nodes that are already well connected. Not because of any grand plan but because those nodes are more visible, more useful and more trusted. This is called preferential attachment.

Think about LinkedIn, when a developer or a founder is new to the startup world and wants to build connections they don’t reach out randomly, they gravitate towards the people who already have the most followers and the most engagement. Not necessarily because those are the only smart or helpful people out there but because they are the most visible nodes in that network, and because everyone connects to them they just keep getting more connected over time.

The node that already has connections keeps attracting more connections. Over time the gap just keeps growing and this is how hubs become hubs. Not through any special design but just through the natural way networks grow. And what’s wild is that this pattern shows up everywhere, in biology, in social networks, in the internet. It almost feels like a law of nature.

And honestly the internet is probably the best example of all of this. Which is where Google comes in.

Google and the Internet’s Hub

The web is a network. Websites are nodes and links between them are edges. And like any network that grows organically it was always going to produce hubs. But I believe that nobody fully anticipated in the early days of the internet, just how dominant a single hub could become.

Google figured out something that sounds obvious in hindsight. The more websites that linked to a page the more important that page probably was - which was the core idea behind PageRank and it was essentially Google’s way of reading the network structure rather than just indexing what was on websites.

Because it did this better than anyone else it became the default place everyone went to find anything online. More users meant more websites optimising for Google which meant more users. Preferential attachment in full effect.

But here is the part I find really interesting looking back at it. That entire system was legible, if some person ran a website, they could more or less understand why it ranked where it ranked. One could see who was linking to your website, you could study what content was performing well, you could build backlinks and optimise your pages. The rules of the game were visible even if they were complex. SEO as an entire industry exists because the network was readable enough that people could figure out how to navigate it.

It felt almost democratic in a strange way. Maybe not perfectly fair but at least transparent enough that a small website with genuinely good content and some effort could find its way up. The hub was powerful but you could see it and work with it in some shape or form.

This feels like it is changing quite a bit.

LLMs Change the Game

Here is where things get a little unsettling.

LLMs like ChatGPT and Claude are increasingly becoming the first place people go when they want to find something or understand something. Instead of typing a query into Google and getting a list of links to traverse you just ask a question and get an answer. It feels faster and more natural and for a lot of use cases it genuinely is.

When you search on Google you are still traversing the network. You get links, you click through, you visit websites. The hub is still routing you through the rest of the network, the nodes are still getting visited, the structure is still alive and functioning.

When you ask an LLM a question that flow stops. The answer just comes back to you. No links to click, no websites to visit, no network to traverse. The LLM has essentially absorbed a huge amount of what the web knew and now surfaces it without ever sending you back into the network.

What comes to mind is that with Google you could at least see the hub. You could understand roughly how it worked. You could trace why certain things ranked and others did not. The network was visible enough that people could interact with it, contest it, build around it.

With LLMs that legibility seems to have disappeared. Nobody really knows why an LLM surfaces what it surfaces, there is no link structure to trace back, no ranking to reverse engineer, no clear set of rules to work with.

The hub still exists, arguably it is more powerful and more concentrated than Google ever was, but now it is invisible. And that sounds like an extremely huge change.

Why This Actually Matters

The old hub i.e. the Google world being visible had a consequence, It meant the power was contestable. A small blog could rank above a big company if the content was genuinely better. An independent creator could build an audience without needing anyone’s permission. SEO for all its flaws was essentially people learning to read and work with something they could see.

People are already trying to do the same with LLMs. I was reading about new fields called AEO (Answer Engine Optimization) and GEO (Generative Engine Optimization), which are basically attempts to be visible inside the new hub the same way websites tried to be visible on Google. But nobody really knows if any of it works. There is no clear feedback loop. You cannot tell if you are being surfaced or ignored and you have no real way to find out.

And that is the part that feels significant to me. When the hub is invisible the power it holds becomes very hard to question or challenge. The concentration of what people know and how they come to know it is moving into systems that offer little to no transparency about how they work.

That is not necessarily alarming but it is a pretty big shift from where we were.

The Honest Uncertainty

I want to be clear that I don’t have any strong conclusions here. I’m a software engineer who was reading about network science and ended up with more questions than answers.

What I do think is that hubs are not going away, If anything LLMs might be creating the most powerful and concentrated hubs the internet has ever seen. But unlike Google which was visible enough that we could at least understand and interact with it these new hubs are….opaque.

We are now in the very early days of a shift and the new structure is not visible in the same way. Maybe that changes as these systems mature. Maybe tools and frameworks emerge that give us back some of that legibility. Or maybe this is just what hubs look like from now on.

I genuinely don’t know. But I think it’s worth paying attention to.