Moth Quantum's browser maze uses qubits and their correlations to generate shifting walls on quantum hardware, without claiming a speed advantage over classical games
A browser game from Moth Quantum turns a quantum processor into a maze generator. In Quantum Backrooms, each tile represents a qubit, and the presence of a wall between tiles is determined by correlations between neighboring qubits. The result is not a faster game or a demonstration of quantum advantage. It is a deliberately accessible experiment in what creative software might do with present-day quantum hardware. Independent confirmation of the project's public demonstrations was not identified in the available search results, so the description here reflects the company's account rather than an independently audited benchmark.
From Qubits to Mazes
The game grew from work on quantum imaginary-time evolution and quantum dynamics. Moth's researchers adapted a process more commonly associated with optimization and quantum chemistry into a game mechanic: a quantum system evolves toward an ideal end state that defines the maze. Players can navigate a space whose configuration changes as that process unfolds.
The company has said that visitors at Gamescom could set their own desired maze configuration and run the process on hardware including IBM Fez and IBM Miami. However, the available search results contain no independent Gamescom documentation, major-media report or official IBM record confirming that those specific processors were used in Quantum Backrooms. What can be assessed scientifically is the proposed interface: a quantum processing unit is being presented as part of the game's generative workflow rather than as a conventional maze decorated with quantum terminology.
The underlying physical system still produces noisy and probabilistic outputs, and the experience does not establish that the processor can solve a classically difficult problem. In quantum computing, a physical qubit is vulnerable to control errors, environmental noise and measurement uncertainty. A logical qubit, by contrast, would encode information across multiple physical qubits and use error-correction procedures to suppress those faults. The distinction is central to interpreting any creative demonstration built on current hardware.
That limitation is central rather than incidental. Moth's Harry Kumar explicitly separates the creative value of a quantum-controlled environment from computational performance. The game runs no faster than a classical equivalent, and the present hardware does not provide the fault-tolerant operation required for the large speedups associated with future quantum algorithms. Research programs at MIT and other leading laboratories likewise distinguish proof-of-principle control experiments from scalable, error-corrected computation.
Atlas Is the Larger Bet
Quantum Backrooms is presented as an example built with Atlas, Moth's platform for people who may have no quantum-computing background. The available material describes Atlas as an alpha-stage platform for creative users, with tasks that could be run on an emulator or through a user's own access key to a quantum service. The company is developing tools and an application programming interface that can connect creative work to emulators or real quantum processors without requiring users to write quantum code.
Atlas is aimed beyond games. Moth describes possible uses involving image processing, music tools, 3D shaders, video and experimental data sequences. A user could upload an image to an engine, adjust parameters and run a process through emulation or hardware. The platform is closer in spirit to a generalized creative engine than to a conventional games studio, although Moth uses comparisons with Unreal Engine and Epic Games to explain the model.
The most technically grounded examples remain exploratory. Moth has developed an entanglement shader for rendering iridescent surfaces in Blender and is considering how similar quantum processes might be adapted to music production. The suggestion of turning that shader into a reverb plugin captures Atlas's purpose: not to prove that quantum hardware is already superior, but to let creators make combinations that a specialist engineering team might not have designed in advance.
For perspective, a quantum circuit is not simply a faster version of a classical graphics or audio pipeline. Its output is described by probabilities until measurement, and the useful behavior depends on state preparation, gate fidelity, connectivity, readout quality and the algorithm's structure. A peer-reviewed Nature quantum-computing study illustrates why claims about computational performance require a clearly specified task, baseline, hardware configuration and reproducible measurements rather than qubit count alone.
Randomness Is Not Advantage
Atlas is also intended to expose users to quantum random-number generation and entropy engines. Quantum measurements can provide randomness rooted in the behavior of a quantum system, whereas ordinary software generators produce sequences that appear random but are generated deterministically. Other physical systems can also supply entropy, so the distinction alone does not make a music tool or game more useful.
Kumar's comparison with generative AI is revealing. He places quantum creativity today near the stage of AI development associated with Google Deep Dream and early DALL-E rather than with mature commercial software. In that period, unusual outputs attracted artists before the systems became broadly useful. Moth's thesis is that creative experimentation could help reveal applications for quantum computers that the current industry's focus on chemistry, cryptography and optimization does not yet expose.
The analogy has limits. Early generative AI could be evaluated through rapidly improving models and visible output quality, while quantum creative tools must also contend with hardware access, calibration, noise and limited processor capability. As discussed in an earlier hardware analysis, qubit count alone says little about whether a system can deliver useful computation; connectivity, error rates and control quality matter as well.
Moth's discussion of quantum randomness acknowledges another constraint: a listener or viewer generally cannot identify whether a particular output came from a quantum random source or a classical one without being told. That does not make the experiment meaningless, but it places the value in the creative process and the interaction with a natural computational system rather than in a perceptible technical signature.
The First Draft Problem
The conversation also turned to artificial intelligence and authorship. Kumar said he avoids using AI to replace the first draft because drafting is where much of the reasoning takes place. He uses it for administrative work and has begun experimenting with Claude Code for software development, while treating the initial act of forming and testing an idea as a human responsibility.
That position gives Atlas a useful counterpoint to automated creative production. Moth is not promising that quantum processors will make artists more creative by themselves. It is building an interface through which artists, musicians and game makers can encounter quantum dynamics directly and decide what those encounters are worth.
The platform was in an alpha phase on October 9, 2026. Moth said it had opened Atlas to a wider group of alpha users, offered emulator access and planned free compute credits for some users, alongside the Moth Hack and an open call for creative projects. The podcast itself was divided into 14 chapters, while Moth reported that six or seven quantum music tools had been created by early users in the preceding week. Those figures describe an early community and an active experiment, not evidence of industrial adoption. The available results do not independently verify the dates, user totals or subsequent platform development.
IBM-related search results available for this report concern other subjects, including IBM z17 processors and a DARPA initiative on quantum benchmarks. They do not confirm Quantum Backrooms, Atlas or the use of IBM Fez or IBM Miami in the project. That distinction matters because technical reporting should not treat a company's association with a hardware provider as evidence that a particular experiment was formally endorsed, benchmarked or documented by that provider.
Quantum Backrooms therefore demonstrates something narrower and more credible than a quantum advantage claim. It shows, according to Moth's description, how quantum hardware could be placed inside a browser-based creative workflow and made legible through a game whose geometry responds to qubit correlations. The important test for Moth is whether Atlas can turn that novelty into repeatable creative practice without hiding the hardware's limits.
A physical qubit is an individual controllable quantum system, not a logical qubit protected by error correction. In this project, the qubits supply evolving states and correlations that influence the maze, but the system is not performing a fault-tolerant computation or correcting errors to preserve a useful algorithm. That distinction matters because a compelling quantum-controlled experience can be scientifically and artistically interesting while remaining no faster and no more capable than a classical implementation. CERN and NASA demonstrate, in different fields, why complex instruments are evaluated through calibrated procedures and reproducible measurements rather than spectacle alone. Moth's strongest contribution is to make that boundary visible: quantum hardware can already be a medium for experimentation, but Atlas does not yet turn experimentation into practical quantum advantage.
