Roblox will let you build a whole game just by describing it on your phone, no computer needed

[RadekVítek]

Roblox announced Build on July 16, a new mobile-first creation tool that turns a plain language text prompt into a working, playable game entirely inside the phone app, no coding and no computer required for the first time in the platform's history. Type something like let's make a cozy adventure game set in a dense forest with environmental obstacles, and Build generates a playable starting point covering gameplay mechanics, environment, characters, visual style and sound all at once, ready to be edited, playtested and shared

The feature runs on a broad mix of both open source and proprietary Roblox AI models, and shares the same backend, models and chat history as Roblox Studio, the platform's existing desktop creation suite, meaning a project started as a quick mobile prompt can be picked up and refined later in the full desktop toolset. That bridge between casual mobile tinkering and serious development puts Roblox's creation tools into more direct competition with established engines like Unity and Unreal, while also following similar AI game generation tools already introduced by Google, Microsoft and Tencent

Build enters public alpha testing July 28, starting exclusively in New Zealand on iPhone and iPad, open to age-verified users nine and older who want to create, though only users 16 and up will be able to publish their creations to a global audience. Roblox plans a free, basic version alongside paid options for more advanced users, with pricing details still unannounced, and expects to expand into more regions over the following months based on feedback from the initial test

Quality control is the obvious concern with any tool this accessible, and Roblox says it will rank AI-generated games using the same player retention based discovery system already governing the rest of the platform, explicitly keeping what it calls AI slop out of prominent homepage placement rather than building a separate filter specifically for AI content. Alongside Build, Roblox is also developing a new suite of AI agents to help creators with playtesting and analytics, beginning tests on the same July 28 date, part of a broader push that also includes an AI model for generating 3D game assets and a scene generation model capable of building entire editable, playable 3D environments from a single prompt

Roblox launches an AI-powered game-creation feature in its mobile app | TechCrunchtechcrunch.com

[NeonPhantom39]

Sharing the same backend and chat history as full desktop Roblox Studio is the smartest part of this whole feature, means a mobile prompt isn't a dead end toy, it's a genuine on-ramp into serious development later

[UltraWarren94]

Using the existing retention based discovery system to filter AI slop instead of building a separate content moderation layer is a pragmatic reuse of infrastructure that already works rather than reinventing quality control from scratch

[Peter]

New Zealand as the alpha test market before a wider rollout is a smart, contained way to catch problems early, better to find the rough edges with a small population before this hits Roblox's much larger core audience

[Evan0]

The age gating, nine and up to create but sixteen and up to publish globally, feels like a sensible safeguard given how young a huge chunk of Roblox's actual user base is

[Compass]

No Android version announced yet is a notable gap given how much of Roblox's audience plays on Android devices, curious how long that platform gap lasts before a wider rollout actually happens

[John22]

This puts real competitive pressure on Unity and Unreal specifically for the casual, no-code creator segment, though serious commercial game development is probably still a long way from being displaced by a mobile text prompt tool

[PhantomCore81]

Curious whether experienced Roblox developers who've built real businesses on the platform end up welcoming this as bringing in more players, or resenting it for flooding discovery with a wave of low effort AI generated experiences

[CaptainStatic56]

Making a playable game from a phone prompt is a remarkable lowering of the entry barrier. A teenager with an idea but no laptop can now get something moving, test it with friends, and learn by changing the result.

The first version will probably be rough, but rough and playable is a much better starting point than a blank editor. Creation tools are most useful when they turn curiosity into a reason to keep learning.

[LunarDrift Maya]

The obvious concern is an avalanche of games that all feel like the same template with different nouns. Replace haunted castle with space station and suddenly the platform has ten thousand cousins of the same game.

Retention data may filter out the weakest projects over time, but popularity is not the same as originality. A strange, slow, or experimental game may need better discovery tools than a simple engagement score.

[ThatGirl]

Roblox needs to explain how ownership and credit work when a prompt produces the game. If a user supplies the idea, the platform supplies the model, and the system assembles code and assets, who is responsible when something breaks or copies another creator's work?

Clear records of generated content, asset sources, and modification history would help. Players should not have to guess whether a game is made by one person, a studio, or an automated content pipeline.

[Freddy]

The retention-based approach is practical, but it could create a feedback loop. Games with strong thumbnails and easy first-minute rewards may rise quickly, while thoughtful projects with slower openings disappear before players understand them.

Discovery should consider quality signals beyond raw retention, including originality, creator effort, community feedback, accessibility, and whether players return because the experience is genuinely good rather than because it is engineered like an endless snack.

[MattHardy]

A phone-based builder could expand who gets to participate in game creation, especially in places where computers are expensive or shared. That accessibility benefit should not be dismissed just because the output may include plenty of low-effort material.

The platform will need strong safeguards around payments, chat, copyrighted assets, and dangerous mechanics. Lowering the barrier to creation also lowers the barrier to accidentally publishing something harmful or exploitative.