Recently, I read an article posted on Mike Masnick’s techdirt.com that helped crystallize a line of thinking that’d been floating around half-formed in my head for a while. That article, “AI might be our best shot at taking back the open Web”, is an extension of Masnick’s “resonant computing” manifesto. In it, he posits that over time, building your own Web site became less accessible to average people as complex frameworks like React, Node.js, and modern CSS became so popular as to be the default. “AI agents”, Masnick argues, are the solution to this problem. Thanks to Claude Code, Cursor, and their many imitators, building a fancy-looking, modern Web site is as simple as “being able to write well and accurately describe things.”
The same applies to building other pieces of software that meet your personal needs. Masnick’s examples are a piece of task management software, a video conferencing system, and a tool that runs his writing past an LLM for editorial review. He was able to go from nothing to having those things relatively quickly and without another person involved, thanks to “AI.” He seems to think that’s a good thing, and that a world where all your computing needs can be met without calling upon the aid of others, where off-the-shelf solutions are replaced by “AI”-tailored one-off programs made to fit your exact idiosyncrasies, is a good goal to strive toward.
I vehemently disagree.
If you’ve never watched the YouTube channel How to Make Everything, I recommend giving it a shot. The channel’s main project (at least in its current iteration) is exploring the progress of human ingenuity from prehistory forward, offering an exploratory-archaeology look at humanity’s techne – its corpus of knowledge and ways of knowing – over the millennia. The rule (which gets broken sometimes, but usually not in egregious ways) is simple. Andy, the channel’s main host, can’t use a tool if he and his cohosts haven’t built it themselves. He can’t use a resource – think copper, wood, steel, rope, et cetera – unless he’s demonstrated a way to obtain it using only the tools and techniques he already has.
There’s a lot of “build tools to build better tools to build better tools” going on, but along the way, there are plenty of explorations of things like agriculture, architecture, and cuisine. As the channel’s technological progress continues, as Andy and his team go from flint to copper to bronze to iron and more, another kind of progress becomes evident. A colorful cast of characters starts to appear on the channel as frequent collaborators. Viewers learn to anticipate the bronze casting guy, or Adrian the blacksmith. Without explicitly saying it, Andy’s channel “invents” the most important technology humanity has ever had: community.
It’s revealing that Masnick’s example of the way the Web used to be is his ability, in those halcyon days, to copy and paste the code from a Web magazine, tinker around with it, and use that code – someone else’s code – to build something himself. The “AI agents” he extols are essentially doing the same thing, except they do it with a lot of extra steps and a lot of compute hours on a GPU somewhere. And to be clear, he’s right about the fact that it’s much harder to do this yourself these days. Web sites really have increased in technical complexity, and a non-expert can’t reasonably expect to be able to plagiarize a modern Web site by themselves over the course of an afternoon.
“Plagiarism” is maybe a harsh word for the process Mr. Masnick describes undertaking with an issue of Fray magazine. After all, the Web before Web 2.0 had a remarkable spirit of collaboration and community. What a young Mike Masnick did wouldn’t have been considered plagiarism, at least not if he wasn’t attempting to commercialize the code. A lot of the way the Internet works rests on this set of ’90s social norms and unspoken rules among techies, even to this day. The fact that the “AI” tools modern Masnick uses got their training data through massive violations of those unspoken rules and norms is still something of a controversy. But there’s another side to that spirit of collaboration. As an unrelated third party, I have no idea whether Masnick ever emailed the people behind Fray, nor do I know for certain that they’d have been willing to explain their techniques if asked. But I suspect they would have.
Community, not the relative simplicity of Web rendering, was the magic of the old Web. For every hobby, interest, or obsession, there were forums where people built communities. Usually more than one, often with silly inter-forum rivalries and dramas. From television series to particular car models to entire fields of scientific study, there were forums for enthusiasts. Many of those forums are still there, their social communities still extant – if a bit diminished. The Internet is a network of networks, not just in the technological sense but in a social one as well. These small communities were the default way of experiencing Online. Even on modern social media platforms, we see such groups coalescing. They struggle to build truly coherent communities – social media isn’t exactly built to enable that – but people gravitate toward that social model regardless.
The so-called democratization enabled by these “AI agents”, much like the so-called democratization of image models like DALL-E and Midjourney, is in actuality the opposite. Democratization requires a demos, a community. These tools are built to make that community obsolete. Instead of reaching out to a friend who’s an artist or a web developer, instead of finding a forum or chat room full of people to collaborate with, you simply tell the machine what you want, and after some expenditure of tokens, you get back something close enough. I understand the bargain being drawn here; community comes at a cost.
“AI agents” are merely the latest in a string of technological innovations to draw a Faustian bargain, getting users to trade away community for convenience. Take music streaming services like Spotify, Apple Music, and Pandora, for instance. Before the streaming services, people discovered new music by listening to radio stations. Those radio stations’ playlists were curated by people – DJs with idiosyncrasies and tastes. It was rare to find a DJ whose taste in music perfectly lined up with one’s own, and that was a good thing! Listeners were exposed on a regular basis to music that was outside their comfort zone, some of which they would dislike and some of which would surprise them. Now, algorithms curate playlists and “internet radio” stations, based on statistical analysis of the songs users have already listened to. Your comfort zone, on air 24/7/365. Streaming video services are replacing network and cable television with much the same effects.
Community comes at a cost. Other people have by nature the very agency that Sam Altman’s and Dario Amodei’s vast warehouses of GPUs can only pretend at from beyond the horizon. They’ll rarely give you something for nothing; they’ll expect some of your time, maybe some of your money, perhaps the listening ear or sturdy shoulder of friendship. Other people will disagree with you, push back on your ideas, and annoy the living daylights out of you from time to time. Communities can be full of drama, infighting, and toxicity. And yet, it is only by paying the cost of community – by forming, navigating, and sustaining relationships with others – that we make ourselves human to begin with.
“AI” is not our best shot at taking back the open Web. It won’t bring back the magic of the pre-Facebook era, because the magic of that era was in the myriad small online communities that made up the Web back then. It’s not even the key to help average people build good-looking Web sites; WordPress is freely available, cheap to host, and has a vast community of people developing, maintaining, supporting, and using it to build their corners of the Web. And WordPress isn’t even the only tool of its kind. There are dozens of options just waiting for people to try them out.
Our best shot at taking back the open Web is the same thing that helped build it in the first place: each other. It’s tech geeks with an old PC gathering dust, deciding to run a forum for their friends. It’s music fans, writers, car enthusiasts, sharing their little corners of the Web with other people who share their interests. It’s flamewars, it’s fanfiction writers, it’s that lady with the website full of GIFs of sandwiches rotating. It’s choosing a metasearch engine, it’s buying your favourite band’s CDs or MP3 albums instead of streaming them on Spotify, it’s Mastodon, it’s Blacksky and Northsky and thousands of blogs. It’s choosing to be human.
Leave a Reply