Now what?
Press enter or click to view image in full size
The Metaverse.
My problem child.
Such a glorious and thoroughly silly idea, one that seemed doomed to be trapped in the pages of science fiction forever. No matter how much money and hype have been thrown it, all attempts thus far have been overbuilt and underpopulated.
That’s because everybody’s been doing it wrong.
Until now.
The Thirty Year Itch
I ran screaming from the tech industry a few years ago, partly due to ennui, partly over the imminent incursion of AI into everything. But I am gradually becoming comfortable with the inevitability of it all. LLMs and generative art have become part of my daily toolbox, and I have even — holding my nose, mind you — made my first foray into generative AI music.
Until last week, though, I hadn’t used AI to write software. Then my pal Mark Pesce, always on top of all things new and awesome, sent me a free week of Claude credits. So I decided to take the plunge into vibe coding.
But what to build? I needed a starter project to get warmed up.
Think, Tony, think.
Oh, I know: I’m going to vibe code the Metaverse: client, server, wire protocol, load a scene, walk some avatars around in it. Just grey boxes, no fancy graphics. But it will be a working end-to-end system. And I’ll do it a week and put it on GitHub.
I really don’t know why my brain works this way. I might have picked a simpler starter project. But no.
I guess I needed to scratch an itch. A thirty year itch. Because, despite all the progress we have made in real-time graphics on the web, despite the international adoption of glTF as the lingua franca 3D file format, and despite us having literally all the pieces we need to make it easy and turnkey to build a 3D world completely based on open source and open standards, put it on a server and have people running around in it, creating and chatting, Second Life-style, right now, today in 2026… still, nobody has done it.
Well, I just did it. By myself. In a week.
An Army of One
Claude Code is a miracle. I’d been hearing stories, but I didn’t know what to expect.
I had a fair amount of trepidation. I hadn’t actually written any code myself for about five years. And when I do write code, I tend to be a control freak. I hate reviewing and debugging other peoples’ code. If my prior experience with generative AI was any indication, I was going to get a lot of unusable slop that I would have to rewrite.
I shared my concerns with Mark. He told me just to let it flow. Trust Claude Code, he said. Trust the process. There is no spoon.
Well, still being me, I waded in carefully, starting with several design discussions with Claude, before I started coding.
It was an absolute revelation.
I gave Claude the high level brief: hi, I’m Tony Parisi, thirty-year veteran blah blah, and we’re going to build an open Metaverse stack together. After Claude kissed the ring and gave me my props, we dove into the architecture. I knew the problem space well, so I gave Claude a quite thorough outline of what we needed to design: file format, wire protocol, client-side API, avatars, user-defined objects, and a browser to interact with it all.
Having built countless tools and systems in this domain before, I had a lot of the problems already solved in my head. But we had others to work out. Now, there’s quite a bit of detail to this. So I was simply shocked at the level of… understanding?… Claude was demonstrating. He… it… turned out to be an amazing design partner on everything. Claude knew the latest and greatest libraries to use, and was up to date on best practices. Before sending instructional briefs to Claude Code to generate a single line, we worked things out in excruciating detail.
I can’t explain the feeling when Claude would unknowingly stumble onto papers that I had written and asked if I was aware of this or that protocol or product. It wasn’t always obvious from the web links, so I would then upload a document to Claude and get a facepalm in response. Of course you wrote that white paper. Yeah, that’s still weird.
Anyway, on to coding. Again, I’m generally not a fan of other peoples’ code, so I was terrified of whatever Claude Code was going to produce. Overall, it was flawless and quite human-readable. I was absolutely blown away. And Claude Code is great at writing and executing tests harnesses, building Node packages and doing all that other schmutz that most of us just hate dealing with.
As a result, in about four days of work I was able to build the first proof of concept for Project Atrium: an end-to-end online virtual world system with a a well-defined communications protocol, a client browser that loads and renders 3D files and tracks avatar positions and broadcasts changes to world state, and a multiplayer server to keep everything synchronized across multiple clients. All built in JavaScript, on top of open source libraries and open standards.
Me, an army of one.
It would have taken half a dozen people three months to do this the old way. Playing this forward, I believe that with vibe coding, a very small team and I could build an infrastructure to rival Second Life or Roblox in just a few months.
Which is part of what has been holding the Metaverse back. We’ve never been able to square the circle of the huge investment required. We’ve had to package it into contrived business ideas— social avatar chat, art galleries, interactive music videos, spinning shoe e-commerce, industrial digital twins, on and on it goes — that often weren’t businesses at all but just individual use cases. The value proposition for real-time 3D needed to be distilled into something simple enough to explain it to venture capitalists like they were five years old, even though the Metaverse is so much bigger of an idea than that.
Well now, we don’t need to go through any of those gyrations. Because the cost of building the Metaverse is rapidly approaching zero. A thousand flowers can bloom. The applications and use cases will soon be limited only by the imagination.
So, hooray for the Metaverse! Maybe it’s finally time. Keep an eye on Project Atrium, and if you’re into this kind of thing, please share far and wide, and get involved! (I could still use help. Human help.)
But that’s not really what is blowing my mind. The Metaverse will Metaverse… or it won’t. As before. As always. We shall see.
No, what’s blowing my mind is what this means for everything tech.
The Idea Economy…?
Let’s zoom out from my tinkering and take a look at the bigger picture. As my friend Reuben Steiger likes to ask:
What do we do when execution cost goes to zero?
Mind you, Claude Code isn’t doing everything for us, and it won’t replace all software jobs (though it will replace many). We’re essentially pair-programming with the machines, and they still need us. For now at least, vibe coding is not a replacement for software engineering; it’s a new way of doing it. We still need system design and architecture, human taste and decision making. But it is getting so much cheaper and faster, very quickly.
So if execution is no longer the long pole, the gating factor for which ideas will make it, then what is? I think Reuben would say that it’s the ideas themselves.
I have known so many brilliant and talented people in the Bay Area with amazing product ideas across numerous human endeavors, from health care to finance, sports betting to video editing, matchmaking to skills training. They may have had the vision, but not necessarily the technical chops or the bandwidth to realize it. The old regime required massive investment, high-demand skills and talent, and huge compromises in vision and scope to get something to market. Now, one person can vibe code a prototype, get their first few hundred users (real or simulated) and rapidly test their concept.
So, maybe ideas are the new seed rounds? Or something. As a self-styled idea guy, that’s an awesome thing to contemplate. If great ideas now have a much better chance of rising to the top, is that going what we build an economy around? Or will some other, not so glamorous motive force replace capital as the engine of innovation?
It’s a big thought.
As to my humble little Metaverse: I am launching Project Atrium into the wild today. With no business model. No goal. No expectations. Just for the fuck of it.
Because I can.
So… now what?