AI Builds Browser Zombie Shooter in Hours, Entirely in Code

AI assistant Teammate generated a playable browser zombie shooter purely in code on Fable 5.2 within hours, demonstrating end-to-end building potential.
A developer shared that AI assistant Teammate built an endless zombie shooter playable directly in the browser in just a few hours on the Fable 5.2 platform. The game is implemented entirely in code with no 3D assets—all visuals and logic are code-generated, which the author calls "rawdogging it all in code." This contrasts with prior AI game-generation approaches that relied on pre-made assets and tests the AI's understanding of low-level graphics programming more thoroughly. The game runs directly in the browser without installation, lowering the barrier to entry. The case suggests AI is moving from assisted coding toward end-to-end application building, which can greatly shorten early game development iteration cycles. However, the information comes from a single tweet with no game link or detailed technical description, so the actual quality, AI's contribution ratio, and Fable 5.2's stability remain to be verified. Regardless, a playable game generated purely from code adds new imagination to AI's use in creative programming.
A developer shared on social media that he had AI assistant Teammate run on the Fable 5.2 platform for a few hours, and it built an endless zombie shooter playable directly in the browser. No 3D assets were used throughout the process—all visuals and logic were implemented purely in code, with the author describing this hardcore approach as "rawdogging it all in code."

Pure Code Implementation: A Hardcore Path Without 3D Assets
The most distinctive aspect of this game is that it completely avoids traditional 3D assets (such as models, textures, and animation files), relying solely on code to generate all graphics and interactions. This typically means the developer must write rendering logic, physics collisions, and the game loop themselves—yet the AI completed this work in just a few hours.
This approach contrasts with the current mainstream AI game-generation mindset: many tools rely on pre-made assets or simple assembly, while the pure-code path tests the AI's understanding of low-level graphics programming more thoroughly. Judging from the results, Fable 5.2 appears capable of handling such tasks, though the details have not been disclosed.
Procedural Generation refers to generating graphics, sound effects, or level content through algorithms rather than hand-crafted assets. In game development, it is often used for the automatic creation of terrain, textures, models, and even animations. In this article, "no 3D assets" means the developer used no pre-made 3D models, textures, or animation files—all visual elements (such as zombies, weapons, and environments) are computed and generated in real time by code. This typically involves mathematical functions (such as noise functions), combinations of geometric primitives, and shader programming. The advantage of procedural generation lies in small asset size and infinite variation, but it is difficult to implement and requires a deep understanding of graphics principles and performance optimization. The fact that AI can accomplish such tasks with code alone shows that it possesses a certain level of low-level graphics programming capability.
Fable 5.2: A Leap in Efficiency for AI-Assisted Development
The specific capability boundaries of Fable 5.2 remain unclear, but according to the tweet, it can produce a playable, complete game prototype within hours. This hints at progress in AI's code generation, logic orchestration, and real-time debugging.
The developer did not disclose the game's specific mechanics or complexity, but judging from the "endless zombie shooter" genre, it involves at least basic modules such as enemy spawning, shooting detection, a scoring system, and player controls. The automatic generation of these modules marks AI's advance from assisted coding toward end-to-end application building.
End-to-end application building refers to the full automation of the entire process—from requirement understanding and code writing to debugging and running—rather than merely assisting with a single step (such as code completion). Traditional AI programming tools mostly focus on generating functions or snippets, whereas end-to-end building requires AI to understand the overall architecture, module interactions, and the runtime environment. In the case described in this article, the AI produced a playable prototype within hours, involving the automatic orchestration of modules such as enemy spawning, shooting detection, a scoring system, and player controls. This requires AI to possess task decomposition, logical reasoning, and real-time error-fixing capabilities. Although this is currently only a single case, if it matures, it will significantly lower the development threshold and accelerate prototype iteration, making it especially suitable for small games or creative validation.
Playable in the Browser: Zero-Barrier Game Distribution
The game runs directly in the browser without installation or download. This distribution method lowers the barrier for users and aligns with the current trend of lightweight web applications. Pure-code implementation means the game's size may be extremely small and load quickly.
For developers, the browser platform means cross-device compatibility, but it also imposes requirements on performance optimization. Whether the AI-generated code can run stably on different hardware remains to be tested in practice.
Web game distribution relies on standard browser technologies (such as HTML5, WebGL, and WebAssembly) to run games without requiring users to download or install anything. WebGL handles 3D rendering, while WebAssembly allows high-performance code (such as C++ compiled output) to execute at near-native speed in the browser. Games generated purely from code are usually small in size, load quickly, and are cross-platform compatible. However, the browser environment imposes performance constraints, especially in graphics rendering and memory management. AI-generated code needs to be optimized for different hardware to avoid stuttering or compatibility issues. In addition, persistent storage, input handling, and multi-device adaptation for web games are also challenges. The "zero-barrier" experience in this article's case is precisely the core advantage of web distribution.
Significance for AI Game Development
Although this case is small, it points to a trend: AI is evolving from a code completion tool into a "collaborator" capable of independently completing small projects. Producing a playable prototype in a few hours can greatly shorten the iteration cycle in the early stages of game development.
However, the information currently comes from a single tweet, with no game link or detailed technical description. We cannot evaluate the game's actual quality, the specific proportion of AI's contribution, or the stability and limitations of Fable 5.2. All of these require more public cases for verification.
Regardless, the news of a playable game generated purely from code adds new imaginative space for AI's application in creative programming.
One-line Summary
AI assistant Teammate generated a playable browser zombie shooter purely in code on Fable 5.2 within hours, demonstrating the potential for end-to-end building.
Paragraph Summary
A developer shared that AI assistant Teammate built an endless zombie shooter playable directly in the browser in just a few hours on the Fable 5.2 platform. The game is implemented entirely in code, with no 3D assets used—all visuals and logic are generated by code, which the author calls "rawdogging it all in code." This contrasts with previous AI game-generation approaches that relied on pre-made assets, and it tests the AI's understanding of low-level graphics programming more thoroughly. The game runs directly in the browser without installation, lowering the barrier to entry. This case suggests that AI is moving from assisted coding toward end-to-end application building, which can greatly shorten early-stage game development iteration cycles. However, the information comes only from a single tweet, lacking a game link and detailed technical description; the actual quality, the proportion of AI's contribution, and the stability of Fable 5.2 remain to be verified. Regardless, a playable game generated purely from code adds new imagination to AI's application in creative programming.
Related articles

How Fast Can AI Become an Expert? Lessons from Napoleon and Rockefeller
From Napoleon's victories at 27 to Rockefeller running a major refinery at 25, we explore why AI can accumulate experience far faster than humans: reading a company's entire history in days, reaching top performance in months, and the real bottleneck of serial feedback.

US Air Traffic Control Keeps Breaking Down—Can Canada's Model Fix It?
A botched copper cable upgrade grounded thousands of US flights. Vox analysis: aging equipment, overworked controllers, and the FAA's annual funding battles are the root causes—Canada's 1996 nonprofit ATC model offers a way out.

Why Can't an AI with 20 PhDs Get Anything Done in the Enterprise?
OpenAI co-founder Greg Brockman uses the "20 PhDs with no data access" analogy to reveal the core bottleneck of enterprise AI: business context and permission guardrails are both essential. This article covers his conversation with the Databricks CEO and the Genie product's implementation approach.