Altitude Development Partners
A few months ago, building anything real in Roblox Studio meant learning Lua, fighting with the physics engine, and spending nights figuring out why your leaderboard kept resetting. Now you connect Claude or ChatGPT to Studio through MCP, type what you want, and watch a playable experience appear in around ten prompts, with no scripting knowledge required.
If you haven't seen MCP yet, it works as a bridge that lets an AI model talk directly to Roblox Studio — spawning parts, writing scripts, setting up game logic, and wiring the whole thing together. You describe the game, the model builds it inside your project, and then you test it and ask for changes while it edits in place. For someone who has never opened a scripting tutorial, this feels like magic. And you can already see the results all over the platform.
AI-built games have a look. When you give a model a vague, low-detail prompt like "make me an obby" or "build a tycoon," it falls back on its defaults, and those defaults repeat from game to game: the gray baseplate, the floating text UI in the corner nobody bothered to move, the starter shop button nobody styled, the generic particle effects on generic parts. Once you've seen a few of these, you can't unsee the pattern. You load into an experience and within two seconds you know a person didn't build this, a prompt did. It's the digital version of walking into a house furnished entirely from the first page of the catalog, where nothing is wrong with any single piece but the whole room screams template.
Real games look different because someone made a hundred small decisions the AI never thought to make — the custom health bar, the sound that plays when you almost die but survive, the economy that punishes you just enough to keep you playing. Those choices come from a developer who sat with the thing and cared how it felt, and you can feel that difference before you can explain it..png)
From what I'm seeing on the platform right now, something like 30% of upcoming games are AI slop — built fast, built cheap, and pulled straight from those defaults. That number is my own estimate rather than a Roblox statistic, but if you spend a week browsing new releases you'll probably land somewhere close. And the share is climbing, because the barrier to shipping a game just dropped to almost zero, and when something gets that easy a flood always follows.
And this flood is landing on top of an already enormous publishing volume. In October 2022, Roblox said creators were publishing more than 60,000 new experiences every day , more than 420,000 a week, or roughly 22 million a year if that pace continued. The figure is dated, and “published” does not mean every experience became discoverable or attracted players, but it shows the scale AI creation tools are entering.
Roblox is pouring fuel on it too: the company is rolling out its own text-to-game tool, Build, that generates a playable experience from a single prompt on your phone, part of the same AI creation push that brought native 3D generation into Studio.
Here's where a lot of people expect me to complain, and I won't. Roblox is not a museum, it's a place to have fun and burn a few hours, so if someone loads into an AI-built game and enjoys it, the game did its job. The player didn't ask who wrote the Lua — they asked to be entertained for twenty minutes, and they were.
I love Roblox for the same reason a kid does: the entertainment, the time you lose in it, the small economies people invent, and the way a developer takes one simple idea and quietly builds a huge hidden machine of features underneath it. Some of that magic comes from craft and some of it comes from a fast, dumb, fun game somebody threw together in an afternoon, and both are valid. A game doesn't owe you effort, it owes you a good time. So I don't think AI slop is ruining Roblox. More games means more chances that someone finds the thing they love, and if players show up and stay, that's the only review that counts.
The fear I hear most is that if anyone can spit out a game in ten prompts, nobody will bother building a real one and the good stuff will drown. I don't buy it, and the reason is simple: players aren't stupid.
AI slop works today partly because it's new and people haven't learned the patterns yet, but that won't hold. Give it a few weeks or a couple of months and the average player will recognize the defaults as fast as you and I do, and once people can smell it they'll bounce off it, because nobody actually likes AI slop once they know that's what they're looking at. Novelty is doing most of the heavy lifting right now, and novelty fades. What's left afterward is what was always true on Roblox — the games people return to are the ones somebody built with care. Effort still wins over the long run because effort is what makes a game feel like a place instead of a template. If anything, the flood makes good work stand out more, the way a real restaurant looks better on a street full of vending machines.
Here's what actually excites me, and it's the reason I wanted to write this. There's a niche opening up in the Roblox ecosystem that almost nobody is standing in yet. Call them Roblox experience prompt engineers — people who know both sides, who understand how these models build and also understand Roblox itself: what makes an experience sticky, how a good economy is tuned, and why one UI feels premium while another feels like a template.
MCP doesn't change the one thing that matters here. The AI will build whatever you ask for, but most people don't know what to ask for. They type "make a tycoon" and get the default tycoon, while someone who knows Roblox writes twenty specific prompts and pulls out something that actually feels designed, with the right progression curve, the right feedback loops, and a UI that looks like a human touched it. That skill of knowing what to ask and how to ask it is going to be worth real money. Studios and solo creators will want the help, especially the ones who can't script and never will but have a strong idea and a budget, and they'll need someone who can sit between the vision and the model and pull a real game out of it.
The people who combine prompt skill with genuine Roblox experience are in a rare spot. Pure prompt people don't know what makes a game good, and many pure Roblox veterans are still stuck in the old, slow way of building, so the ones who sit in the middle — fast with the tools and sharp on design — are about to be in demand while there are still very few of them. Clients looking for help with design or scripting will go looking, and the field is wide open.
MCP made building a Roblox game as easy as describing one, which is going to fill the platform with AI slop, and roughly a third of new releases already fit the description. That's fine, because players decide what's good by playing it, and a game that keeps them entertained has succeeded. The slop won't last, since people learn fast and nobody loves a template once they can see it, so quality games aren't going anywhere. And in the gap between "anyone can build a game" and "almost nobody knows how to build a good one," there's a real opportunity for the people who understand both the AI and Roblox — which is exactly the space I'd want to be in.