ChrisF | Starholder
We will begin this essay not 20 minutes into the future, but 47 years in the past when a boomer god performing the most leviathan of classic rock songs, Stairway To Heaven, asked a packed Madison Square Garden "Does anybody remember laughter?"
Whether you've ever gotten the Led out or not, this question sounds very familiar. You've probably heard it recently, just asked a different way by very online Gen X-ers and Millennials.
Do you remember when the internet was fun?
There was a time when the internet was in fact fun. I know it's hard to believe. but it is true. I'm as old as Robert Plant's laughter quip and I can remember when the internet was a place of discovery and serendipity.
Most people use Tumblr as their touchstone for a kinder, weirder, funner internet, but honestly that's a Zoomer structuralist response. It's an answer given by people who only know of doomscrolling, of content being sandwiched into blocks and served up in factory-templated feeds. Yes Tumblr was weird and it was beloved, but it carried all the hallmarks of what has ruined the internet today. Aggregation, structural conformity, reactions and comments...in short, herding.
It's nearly impossible in this day and age to think of a website whose job it is to send you somewhere else (even search engines keep you contained as much as possible), but before Tumblr there was another beloved website / browser extension which did just that, StumbleUpon.
StumbleUpon operated on a premise of discovery. The internet was wide open, unmapped and full of surprise, only we often had no idea where to go or what to do. You'd use StumbleUpon by telling the site what subjects you are interested in, and then it'd launch you out into random (but topically related) websites. Here's a quick, super old YouTube about it:
A big part of the appeal of StumbleUpon was that we as users weren't entirely sure what the internet was capable of. Yeah, we had Yahoo and Google, but they didn't own mindshare like FAANG does today. We were more open to the possibility that the internet was infinite and it was worth seeking out all the nooks and crannies it contained, just to see what was there.
Honestly until generative AI and LLMs came along, I had forgotten what it was like to use networked tech just to see what was there, to see what we as a hivemind are capable of creating before being yoked down and herded by entrenched interests.
I know I titled this essay 20 Minutes into the Future, but we are going to do one last bit of time travel, and go way way back to ancient Greece to talk about the tension between the genos and the demos, along with the best solution they had for it in 750 BC. It's a good setup for when we do flash forward.
The genos is what you inherit when you are born into the world. It's literally a family bloodline and a term for the noble houses of classical Greece, but metaphorically it was more than that. The genos defined culture, tradition, and laws. It was the weight of history that guided how people moved through the world.
The demos were defined by their otherness to the genos. They were the common people, the ones who had to live with the weight of being outside the genos. When the population of Greece picked up after the Bronze Age Collapse, good land and resources became scarce. Tension between traditional rights and immediate needs flared up.
Did the genos give up their ancestral holdings to make room for newcomers? No. The answer was to go elsewhere and open up undiscovered country.
Flash forward to the present. Apple is currently defending, stalling and even disregarding attempts to open up its app store and payment monopoly. The demos is demanding better rights and treatment, and the genos is doing all it can to keep that from happening. Same as it ever was...or better said, same same but different.
When Cornith got too full, its demos and disinherited genos hauled their butts off to Sicily and started Syracuse. That same process is happening today. Crypto can be viewed as one early and pivotal exodus from mainstream net culture, and the next one is emerging with generative AI and LLMs.
Which brings us to Websim where we can finally step 20 minutes into the future.
Websim is named for what it does. It simulates the internet, treating the address bar in its virtual browser as a chat interface where users can prompt the platform to render virtual websites dynamically.
The project isn't even a month old, and was not an intentional creation. Its founders, Sean Lee and Rob Haisfield, met up at a hackathon where each was interested in another AI project engaged in simulation. Serendipity took hold and in a weekend, they built a tool that simulated internet pages.
Using Websim is quite easy. Lets walk through it. For starters, you need to login with a Google account and then you'll see a virtual browser within your browser:
You then click into the address bar of your virtual browser and use it to navigate a simulated internet which is created by an LLM (currently Anthropic's Claude) under the hood.
The address bar takes either a URL that you make up or you can just use it like a chat window and type text into there which the AI uses to create virtual web pages with. I find both work equally well, but enjoy constructing fake URLs more just to see how it interprets the query strings and params to create a visual output from it.
To be very meta in this explainer, I took all the text from this essay so far and just pasted it into the address bar:
And Websim then produced a page that looks a lot like the one you are reading this essay on now:
Which is simple enough, but what if I told you I also made this very glitchy cube using it?
Wild huh? Once I discovered this unlock, it was deep down the rabbit hole to see just what it was capable of rendering.
And now the possibility of Websim starts to come into focus. This isn't just a modern day website builder, some piece of SaaS with an AI strapped onto it. No, this is the ability to summon and conjure up vector space. It is the ability to travel into the depths of a large language model and transform the information within it into net new experiences rendered in networked address space.
We have not seen anything like this before and it is very exciting. The community in their Discord is doing astounding work, creating games, virtual spaces, art, simulating models and really only being limited by imagination. It's truly inspiring to see how people are exploring this new functionality.
In my last essay I introduced the concept of Complex Adaptive Media Systems (CAMS), which is an organizing model for how to think about future media once we have the ability to scale our creations past objects and websites and into worlds and simulations. Thanks to generative AI lowering production barriers and massive scaling in crypto's throughput and capabilities, we are on the cusp of a revolution where media will become programmatic, heavily networked and operate on orders of magnitude larger than it does today. Think of how your parent's landline telephone was used in 1990, now think of how you use your iPhone today. That sea change in scale, scope and use cases is coming to media over the next decade.
A critical aspect of that change will be in how we interact with media. Rather than passive, scripted media we snack on from our couch, these virtual worlds and simulations will be hands on and exploratory. We'll need tools and new modalities for summoning what is inside them in ways that work for us. In short, we'll need AI to render user interfaces on the fly based on input from us. We will tell the AI where and what we want to investigate, and it will dynamically create liminal spaces that transform all the objects, information and relationships within a virtual world into an interaction layer for us to explore.
This is the new world that awaits us, and Websim is the first time I've interacted with a project that approximates what is coming. It is a proto-CAMS, a glimpse 20 minutes into the future.
The internet is fun again thanks to Websim, give it a try and see for yourself.
I had a lot of fun playing with Websim over the last week and wanted to document what it was capable of in its first month of existence. My ten favorite outputs are assembled into a NFT collection on Zora as editions of 12 (120 piece set total). Each piece is a video capture of a Websim simulation, and metadata includes a link to render the piece in browser on Websim.
Wallets can mint up to 10 pieces in case anyone wants to collect the full set. Enjoy!
Editor's Note: I am an idiot and messed up the release date. The drop will go live Monday, April 29th at 3:45 ET.
