In this first 2021 episode of the Breaking Smart podcast, I want to talk about something that’s been on my mind a lot lately that I call anti-network effects.
Covid vaccine (source: Wikimedia Commons, CC-BY-SA 3.0)
1/ As I am recording this, governments around the world are working out the logistics problems of distributing billions of vaccine doses. It feels like a symbol of the times we are entering into, times that I think will be defined by anti-network effects.
2/ Vaccinations and mask-wearing are examples of anti-network effects. I’ll define these as effects that can slow down, regulate, arrest, or reverse the operation of network effects, and which might themselves be network effects.
3/ For almost 50 years, since the invention of the PC, the world has been riding one network effect after the other, all of which ride on top of the infrastructural network effect of the internet.
4/ To review, a network effect is when the power of a system grows faster than its size. The original form was called the fax-machine effect. One person with a fax machine is useless. Two people is one connection, three people is 3 potential connections, 4 is 6, and in general n nodes is n(n-1)/2 unique connections. So the power of the network grows as the square of its size.
5/ Network effects in computing infrastructure are deeply connected to Moore’s Law. When computers get cheaper, network nodes get cheaper too, and more things can get attached to computers, and then to each other via the network.
6/ This leads to a double effect. Every 18-36 months, the density of transistors doubles. This makes the cheapest computers based on the cheapest chips much cheaper, and what’s more, this allows whole new classes of even cheaper computer to be invented ever decade or so. So one compounding effect rides another. That’s how we we went from mainframes to Raspberry Pis, and from an internet of 2 computers to an internet of billions.
7/ There’s in fact a third effect. If computers get more powerful at a steady geometric rate, and industrial mass production lag is minimal, then the rate at which the network can grow is itself a function of network size. You don’t have to add 1 node at a time, you can bulk-add n nodes.
8/ Everybody with a home computer already had an internet connection, so with a router, everybody can add a wifi connected device at the same time. With software it’s even easier: everybody with a phone can download an app at nearly the same time, creating near-instantaneous soft networks.
9/ We’ve been riding this triple-punch meta-network effect for nearly 50 years. You’ve got Moore’s law, the basic network effect of devices, and then the network-size-proportionate growth effect. And for people about my age and younger, our entire lives have been spent on this ride. To the point where it is second nature.
10/ This is actually pretty unnatural. Think about the classic brainteaser to teach exponential or geometric thinking. If a lily plant doubles in size every day and on the 30th day covers the whole pond, when did it cover half the pond?
11/ The answer is of course, the day before, on the 29th day, and to people like us, this barely even counts as a brainteaser. In fact, for kids today, I suspect the expected wrong answer, which is the 15th day, will feel unintuitive. They deal with fewer important things that work that way.
12/ There’s a lot more to say about network effects and various formulations like Metcalfe’s Law and Reed’s Law, but we’ve been doing that for my whole life, so enough said. I’ll just add one more point: network effects are pretty dumb. I mean even viruses and lily plants on ponds can embody them.
13/ This is in general not true of anti-network effects. While some anti-network effects are themselves driven by network effects, most work on other principles. So let’s ta