The Extended Internet Universe

The Internet is now a very complicated place

For several years now, I’ve been watching the creeping, unheralded growth of what I call the cozyweb, and for which others have lots of creative names. Kickstarter founder Yancey Strickler called it the Dark Forest in a recent post.

Unlike the main public internet, which runs on the (human) protocol of “users” clicking on links on public pages/apps maintained by “publishers”, the cozyweb works on the (human) protocol of everybody cutting-and-pasting bits of text, images, URLs, and screenshots across live streams. Much of this content is poorly addressable, poorly searchable, and very vulnerable to bitrot. It lives in a high-gatekeeping slum-like space comprising slacks, messaging apps, private groups, storage services like dropbox, and of course, email.

Public and cozy are just two of the dozens of distinctly different flavors of internet around today. Here’s my attempt at mapping the complexity of the extended internet universe.

Think of it as a vertical section through the nominal internet (everything connected via TCP/IP). The vertical dimension is depth and complexity of the permission/security architecture you need to navigate to access things. The horizontal dimension goes from darkweb with little to no public internet activity, to the well-lit web, with lots of both public and private activity.

So there’s this 2x2 under the hood here, with high to low risk on the x-axis and high to low privacy on the y-axis. The x-axis itself is the private-to-public boundary, marked by email for most of us. The y-axis is the high-risk to low-risk boundary, marked by security stronger/weaker than simple passwords for most of us.

Top Left

As you might expect, the high-risk, low-privacy quadrant is mostly empty (except for accidental or malicious disclosures like leaked information). The “darkness” above the darkweb is missing publicly visible activity. Activity might appear there in collapsed societies with no state power, where underworld elements can brazenly wander about in the light of day, OR in highly open utopian societies where there are no taboos, people leave their doors open, and there is no use for locks. The top left quadrant is basically empty/dark because both pure utopias and dystopias are hard to sustain for longer than 5 minutes.

Top Right

The darkness is the natural dual of the adtech web, the zone of extreme overactivity above the surface of the cozyweb, with businesses trying desperately to penetrate into private spaces past the open-to-private boundary marked by email.

This makes poetic sense. The adtech world is neither utopian nor dystopian. It is pragmatically mehtopian. It’s the essence of the mediocre world we have, whatever the world we want might look like.

One reason I’m jumping on the substack bandwagon with 2 of my newsletters is that it is an interesting bet against the ad-supported public internet. Here’s CEO Chris Best arguing that podcasts need to go from ad-supported to subscription-based. This idea wouldn’t have been possible as recently as 10 years ago, due to the jankiness of payments infrastructure.

The argument generalizes beyond podcasts. Increasingly, the only reasons to do “free” content at significant scale are bad ones, involving manipulation of crowd sentiments. The ad-based internet isn’t the cause of that, but it certainly makes it indefinitely sustainable.

Center and Top

The center and top of the map/2x2 form a sort of Yggdrasil public world tree through which the long-distance spooky entanglements of the extended internet operate.

The core of the internet is still the public web (which on the 2x2 sprawls across all 4 quadrants rather untidily), and though the rest of the world may not like the idea, it’s still primarily US-centric. To the extent other nations have created distinct internet flavors, they exist as restricted publics on top of the US-centric global public web. I’ve shown this as the purplish upper branches.

Between those branches we find old media (gray ladies in their gray gardens? 🤣), which is an authoritarian information space by the standards of the modern extended internet universe. I think it makes poetic sense to classify things like newspapers and cable television channels with (say) the Chinese and other less-free-than-American internets. The level of regulation and top-down control of information flows is comparable.

The American internet may head in that direction as well in the future, leaving an entire dark middle band: 2/3 dark (modulo leaks), 1/3 ads (aside: ads are “darkness” content too in a sense, since we all try our best not to look at them).


The dotted lines connecting the archipelagos in the bottom half of the map are cut-and-paste type human protocol activity, often even air-gapped human transfers (business cards, scribbled email addresses, all the way to secret single-use darkweb addresses passed on through dead drops).

In the lower right cozyweb archipelago this is just permissions negotiated via meatspace handshakes with no particularly serious security concerns. Most of the >1:1 cozyweb spaces I’m part of have an informal FrieNDA (friendly NDA) in effect, but much of it is merely boringly private. Pajama web rather than cloak-and-dagger web.

I’m brewing cloak-and-dagger conspiracies only in a few select channels. Like most sane people, if I actually wanted to do dark and devious stuff, I’d keep it all entirely offline, and conduct my dirty business in close-up whispers with the shower on, music blaring, and wearing a tinfoil hat. Even supposedly “encrypted” apps (modulo NSA backdoors) are risky if you actually want to pick a significant fight with a powerful adversary.

That’s the zone on the lower left, where there are security concerns and hostile intentions all around, and lots of associated technology like VPNs, stronger-than-passwords security models, etc. In that corner, the corporate deep web morphs into the government-corporate deep-state web of security and law-enforcement agencies, whitehat/blackhat hackers, and then into the dark web proper, of drugs, assassination contracts, and child porn.

But the big story is the perfectly legal cozyweb. More and more civilian online activity is sinking into the cozyweb. It doesn’t have a good interconnection fabric (ie, it’s all about the sophistication of cut-and-paste). It has better boundary/gatekeeping technology (administration, kicking users out) than linking technology. The cut-and-paste-and-screenshot protocol could use some good, deeper productization.

Apparently, Japan has already been down this road (ht Adam Elkus) with people retreating from the public internet to the cozy internet.

The cozynet is not the same as waldenponding, since it remains connected and online and isn’t concerned about distraction or overuse of digital media. It just retreats from public view/activity for various reasons ranging from simple preference for privacy and small communities to fear and PTSD.

The Pick-2-of-3

The current governing logic of the extended internet universe, I think, boils down to a pick-2-of-3 constraint triangle: {free, open to the public, quality}.

Can we have all 3? Many idealists think we once had that (we never did, we just had small scale, and it still wasn’t hitting all 3), or that in some hand-wavy way, “true” net neutrality would deliver that (I’m only a very weak defender of net neutrality: it was a useful principle for a while, but it has outlived its utility).

I think we can get closer than we are today though.

That’s the promise of the crypto world (not just cryptocurrencies, but cryptographic technologies in general, starting with private-key based identity and security for everything). It is still at a very rudimentary stage of development. But once it matures, we will see the entire extended internet universe getting rearchitected on better foundations, where we could at least push out the boundary of the 2-of-3 pareto.

For example, the cozyweb could evolve from cut-and-paste to a personal blockchain of context-permissioned, addressable, searchable, interlinked clips. Disqus comes architecturally close with its commenting system, but requires publishers to cooperate, with no meaningful incentives to do so. Plus it is public rather than private, and relies on a trusted 3rd party, and the idea that trusted 3rd parties are security holes is one of the doctrinal foundations of all things crypto.

Disqus is also simply too old technologically, since it dates to an era when “commenting” was a thing. The cozyweb may have been born as the comments section of the public internet forking off, but is now a vastly bigger, more complex space.

This observation generalizes: a LOT of the problems of the internet are due to much of it being simply too old. It wasn’t built to last this long even conceptually (ie in terms of atomic ideas like “comments”), let alone technologically. We need improvements, upgrades, and modernizations. The industrial world did that, and the internet world will have to as well.

Things were so simple 20 or even 10 years ago (social media, dated to the invention of RSS, is 20 years old, bitcoin is 10 years old). In 1999, there was the Internet, and there were various Intranets.

Then things got complicated. There are interesting times ahead.