Originally published October 2025. Rec Room subsequently closed in June 2026 — read my post-mortem here.
I’m still thinking about the painfully sharp layoffs of Rec Room staff of August 2025, because I’m not sure people in the metaverse platform/virtual world industry are seriously considering the implications. If you’ve been to an XR / game development conference lately, you’ve noticed they’re awash with panel topics in the vein of, “How Gen AI prompts will revolutionize virtual worlds!”
But it just did the very opposite with Rec Room.
A leading metaverse/virtual world platform, which I’m tracking as having roughly 40 million monthly active users, Rec Room has impressive user activity and numbers, but the company is not yet profitable.
And Gen AI only put profitability farther away.
Indeed, Rec Room’s official cutback announcement partly attributed the company’s troubles to the recent introduction of Maker AI, a prompt-based creation tool which “frustrated our more impactful creators.”
How so? I reached out to a Rec Room staff member for more clarity on that point:
“The AI frustration for top creators is for the obvious reasons you’d imagine,” they told me. “If you spent years building up skills to make top Rec Room games, you don’t want to see easier tools coming out that help others skip some of the learning steps and start building games and rooms faster than you were able to when you were learning.”
What’s worse, a leading Rec Room influencer tells me Maker AI didn’t really help most veteran Rec Room creators — while at the same time, the existing tools they were working with to build and maintain popular Rec Room-based games and experiences had outstanding bugs and updates that needed addressing. (Causing still more frustration!)
Did all that prompt-based creation with Maker AI encourage newer users? Maybe. But it also generated quite a lot of low quality slop that was expensive for Rec Room the company to support. (Gen AI requires quite a lot of compute power and data!)
Rec Room management basically acknowledged this in an April 2025 blog post on Maker AI:
Here’s what the average player is doing right now:
Spending over 35 minutes with Maker AI
Issuing over 300 prompts
Generating more than 20 textures or canvases
With great usage comes… great costs. The level of engagement is helping us get a handle on the resource demands of supporting AI-driven tools for creators…
Right now, it looks like the money we’d spend to support a month of Maker AI for an individual is way higher than the RR+ [paid subscription] net revenue we get from that subscription.
Emphasis mine, because in other words, the more time a Rec Room user spent using Maker AI, the more money Rec Room the company lost. It got so bad, the company also then announced that Maker AI would not be supported outside a user’s Dorm (personal space): “Doing so,” they explained, “would leave us financially upside down on each player using Maker AI.”
The company then promised to spend the coming months to optimize and “bring down the costs of popular AI requests” — which again, meant spending more money and attention on an AI tool that could be going elsewhere. For example, increased support for the Rec Room creator community.
And so AI was causing Rec Room to lose money, even while the overall platform had a large user base and decent revenue. (”Our balance sheet is actually pretty strong,” as an August 28 Rec Room blog post explains, “but we were spending too much money. That’s why we had to do the layoff. “)
I’m reaching out to Rec Room’s lead developers, because I’d like even more clarity on what went wrong here. But in general, here’s a couple takeaways for all virtual world developers to consider:
Make noob creation tools easier and more social.
Prim/building block-based creation is already easy for new users to learn, and pleasurable in itself — clicking and dragging blocks into existence, connecting them with each other, setting them in the environment where other players can interact with them, is an experience already familiar in dozens of popular games/platforms. This rudimentary 3D creation could be made even easier (especially on mobile), but it remains core to the magic and appeal of virtual worlds. Not only is there no good reason to replace all this with prompt-based Gen AI — doing so tends to defeat the virtual world’s very purpose.
In virtual worlds, UGC is more than just 3D building
Not everyone enjoys playing with building blocks. But in a social virtual world built on user-generated content, all user activity is itself UGC. Emotes, chat, gestures — all forms of UGC which can and should be improved upon.
Text chat alone is in sore need of transformation, since it doesn’t just have to be words on a screen. (Veteran Second Life users will fondly remember Starax’s Wand, which would cause surprising objects — pianos, cows, and the like — to literally drop from the sky during user conversations.)
Or to put it another way:
If you really want to revolutionize virtual worlds, think of new and delightful ways to amplify human creativity. Not to automate it away.
Originally published on my Patreon.
Leave a Reply