When Google released a preview of Project Genie last week, the stock market reacted as if a new game engine that would replace all game engines had just been unveiled. Unity’s stock plunged by more than 20%, while Take-Two and Roblox both fell by over 10%.

The magnitude of the reaction was striking given what Project Genie actually is: an experimental AI demo that generates short, navigable 3D scenes. Google describes it as a “world model.” At a glance, Genie looks almost magical: you type a short prompt or upload an image, and the system generates a world that you can move in using basic controls.

In practice, Genie’s capabilities are quite limited. Each generated world only lasts for a minute before breaking down. Visual fidelity is modest, closer to a low-resolution video stream than a modern 3D game — running at 720p and 24 frames per second. Input responsiveness is noticeably laggy, and the world itself isn’t fully consistent: objects subtly change, geometry drifts, and the model “forgets” what was previously established.

What Is a World Model?

Despite Genie’s limited scope, the stock market reaction was anything but subtle. To understand the fuss around Genie, we need to clarify what a world model actually is.

Put simply, a world model is an AI that tries to guess what should happen next in a virtual space. In Genie’s case, instead of relying on hand-built levels, assets, and programmer-defined rules, Genie generates the world on the fly based on what it has seen before. When you look around and move, Genie makes educated guesses about what should be in front of you.

What’s still unclear is how an output from something like Genie would ever become a game in the traditional sense. There’s no obvious path from a probabilistic, generated world to an actual game with deterministic behavior, consistent mechanics, and a solid performance.

Moreover, games are not defined by the worlds they depict but by the rules they enforce. Players invest time in games because actions have predictable outcomes, and they care about their progress. Whether it’s learning enemy patterns, making damage numbers bigger, or competing on a ranked ladder, the appeal of games depends on stability and repeatability, which is exactly what world models don’t do.

Notably, world models aren’t entirely about games. Different teams are building world models for very different ends: some aim to simulate physical environments for robotics or autonomous agents, while others use games primarily as a rich source of behavioral data rather than something meant to be played. Genie sits at the games-adjacent end of this spectrum.

It’s also worth noting that attempts to generate games with AI aren’t new. Tools like Astrocade, Bitmagic, Pikoo, and Spawn are not world models, but they already create playable experiences from prompts, and the technical achievements behind them are genuinely impressive. In fact, these tools are already capable of producing games complete with mechanics, progression, and complex gameplay systems. But among the hundreds of games produced through these platforms, there hasn’t been a single breakout hit that meaningfully cuts through the noise. Much like Genie, these tools are useful for prototyping and exploration. However, any experienced developer knows how far a prototype is from a production release, let alone a continuously updated live service game.

Implications for Game Engines and Developers

There are reasons Unity could be concerned for its future, but competition from world models is not one of them. Game engines and world models solve fundamentally different problems. World models excel at fast, approximate generation. Game engines, by contrast, provide stable world representations, deterministic simulation, and efficient rendering at scale.

Indeed, both Epic and Unity are vocal about their views that these models will likely evolve into complementary tools, sitting alongside engines rather than replacing them. Epic CEO Tim Sweeney was quick to comment that we’re likely to see “constant leapfrogging between engine-centric AI and world-model-centric AI until they come together for maximum effect,” with each side playing to its strengths. Unity CEO Matthew Bromberg echoed the sentiment in a lengthy LinkedIn post, arguing that a world model “accelerates environment and asset generation, while Unity provides the execution layer that transforms generated content into reliable, monetizable experiences.”

It’s also unlikely that Google has ambitions to enter the game engine business directly, given how notoriously difficult the market is, with limited size, low margins, and long adoption cycles. Moreover, Google’s track record suggests it excels more at developing novel technology than at operating gaming products.

For developers, the most obvious use cases aren’t that different from today’s generative AI tools. World models could be useful for early visual exploration, mood setting, or rapid prototyping. While they are unlikely to produce commercially viable games on their own, they lower the barrier to expression. When anyone can sketch a world from imagination alone, the range of ideas that can be explored grows dramatically, even if only a fraction ever turn into real projects.

In that sense, Project Genie may be one of the most impressive technical demonstrations of the modern AI era even if it’s not usable for serious game development work. Its output is approximate, it breaks down in edge cases, and it’s expensive to run. Meanwhile, the tools that are actually changing day-to-day game development are far less flashy: programmers pair-programming with Claude, artists accelerating bulk asset work with Scenario, and designers ideating with the tireless help of ChatGPT. Perhaps the biggest revolution we’ve seen so far is in mobile direct response marketing, where small, one-off, throwaway deliverables are exactly what’s needed (and in large quantities!).

At the same time, new creative tools have a habit of enabling entirely new kinds of content and new ways of consuming it. As world models improve, they may eventually be able to produce engaging interactive experiences on their own. Even if those experiences don’t resemble traditional games, they could find an audience as standalone interactive media, distributed more like videos or web experiences.

Why the Hardest Problems Remain Unsolved

Back in 2022, when ChatGPT launched, it truly felt like magic. Many of us were convinced that game development would be transformed within a year or two. That didn’t happen: AI didn’t revolutionize game development.

It did quietly make good developers better. Instead of an AI revolution, we got thousands of small use cases across programming, art, design, and writing, each delivering 10–15% productivity gains. Those gains are more than meaningful, but they’re not magic. The reality is that the AI tools that succeed tend to be boring, narrow, and practical improvements to existing workflows.

Fully leaning on prompt-based game generation would require rethinking how games are made end to end, and that’s hard to imagine given the fundamental advantages game engines provide: control, deterministic outcomes, and performance. With generative AI, it’s often quick to build something and feel 80% finished. Polishing and tuning the final 20%, however, is far more difficult.

By contrast, the tools that have actually taken off are the ones that slide neatly into existing processes. Claude can become a daily companion for developers; Genie, at least for now, feels more like a one-off experiment. More plausible near-term uses for world models look additive rather than disruptive: as part of creator toolkits on platforms like Roblox, or as a way to generate rough world layouts or visual references that are later rebuilt inside Unreal or Unity.

Seen in this light, the stock market’s reaction to Genie looks like a category error. World models expand what developers can do, without posing a direct threat to game engines. Even if the bull case for world models in game development assumes entirely new workflows for making games, that doesn’t mean that existing ones would disappear. World models could open up new forms of creation or slot into parts of today’s pipelines without fully replacing engine-driven development.

This doesn’t mean Genie serves no purpose. Genie, and other world models, should not be assessed as game engines, but as novel technology and an entirely new category of media. World models may find use as a new kind of storytelling engine, in industrial simulation, or as infrastructure for training other types of AI we haven’t seen yet. But even then, they are unlikely to replace games built on rules and deterministic systems. Whatever Genie becomes, it’s unlikely to ship the next breakout game on its own.

A Word from Our Sponsor: modl.ai

The New Era of Autonomous Game Testing

Most QA automation still depends on scripts and integrations that need updating every time your game changes. modl:QA+ takes a different path — a black-box, vision-based approach where AI agents literally play your game, while AI analysts watch, learn, and report issues. No SDKs, no Unity plug-ins, no manual setup.

Using a combination of computer vision, reasoning models, and feedback loops, modl:QA+ autonomously explores builds, detects bugs, and generates actionable reports that sync directly with your workflows. It’s the first scalable alternative to human testing — one that learns from every run and gets smarter with use. Test more. Ship faster. Let your QA evolve.

In Other News

💸 Funding & Acquisitions:

• Ares Interactive, a game developer and publisher, has raised $70M to build cross-platform, next-generation free-to-play games.
• Singapore-based games publisher 4Divinity secured a $10M strategic investment.
• Seattle startup Sound Games raises $6.5M to rethink how premium games are sold.
• Rewards platform Tesoro XP has raised $5.4M in seed funding to accelerate its core product and partnership roadmap.
• Ukraine-based social casino games developer Murka Games has completed a management buyout from US-based private equity firm Blackstone.
• Australian casino firm Aristocrat Leisure has divested assets from Big Fish Games following a strategic review of the company's games businesses.

📊 Business & Products:

• Epic Games Store sees record $400M in third-party player spending.
• Nintendo net sales up 99.3% as Switch 2 shifts 17.37M units.
• Stillfront records $253M non-cash impairment charge.
• South Korea to seek expanded Chinese game licence approvals in 2026 strategy.
• Playgama launches advertising solution for web gaming monetisation.

👾 Miscellaneous:

• Nintendo’s Fire Emblem Heroes celebrates ninth anniversary and over $1.2B in revenue.
• Esports World Cup Foundation commits $45m to Esports Nations Cup 2026.
• India's Union Budget commits $27.3M to talent development in creative sectors.
• Indie game film adaptation Iron Lung earns $21.5M on opening weekend.
• Candy Crush Crushable is a new daily puzzle game exclusive to Yahoo Games.

Content Worth Consuming

"You Don't Have to Tell Me Pyramid Head's Backstory" — Poppy Playtime's George Krstic on Why Less is More in Horror (gamesindustry.biz): “Poppy Playtime was released on Steam in 2021, tasking players with exploring an abandoned factory stuffed with terrifying toy-like creatures, including the iconic, blue-haired Huggy Wuggy. It quickly became a phenomenon: according to Video Game Insights, the game has sold around 3.2 million units on Steam alone, with millions more players on console and mobile. The story of the game has been continued over DLC, with Chapter 5 set to be released on February 18. Ahead of its launch, GamesIndustry.biz spoke with George Krstic, senior director of creative development at the game's developer, US indie studio Mob Entertainment.”

Interview: World of Warcraft Bosses on Ageing Players, Midnight and Expanding the Warcraft IP (The Game Business YouTube Channel): “In this week’s edition of The Game Business Show, we sit down with World of Warcraft leaders Holly Longdale (VP) Ion Hazzikostas (game director), where we discuss the game’s ageing audience, the impact of World of Warcraft Classic, working with Microsoft and the future of the Warcraft IP.”

State of AI in 2026: LLMs, Coding, Scaling Laws, China, Agents, GPUs, AGI (Lex Fridman Podcast): “Nathan Lambert and Sebastian Raschka are machine learning researchers, engineers, and educators. Nathan is the post-training lead at the Allen Institute for AI (Ai2) and the author of The RLHF Book. Sebastian Raschka is the author of Build a Large Language Model (From Scratch) and Build a Reasoning Model (From Scratch).”

Abiotic Factor with Geoff “Zag” Keene and Henry Feltham (The AIAS Game Maker’s Notebook): “Corey May chats with Abiotic Factor developers Geoff “Zag” Keene and Henry Feltham from Deep Field Games. Together they discuss the challenges and benefits of working in small teams; the importance of relationships and communication; narrative design as a production tool; balancing seriousness and comedy; and what they didn't like about the survival genre that they hoped to remedy with Abiotic Factor.”

Palmer Luckey on Hardware, Building, and the Next Frontiers of Innovation (The a16z Show): “Recorded live at our Founders Summit, a16z general partner Chris Dixon speaks with Palmer Luckey, founder of Anduril and Oculus VR. They talk about what it takes to build hardware at scale, where the biggest technological bottlenecks are today, and why optimism is still warranted despite geopolitical turmoil and regulatory constraints. They also cover crypto, stablecoins, modern warfare, the U.S.–China technology race, AI and manufacturing, and frontiers like fusion and quantum computing—plus lessons from Oculus, the founding of Anduril, and how to build mission-driven teams.”

Ads in ChatGPT (with Rishabh Jain) (Mobile Dev Memo Podcast): “My guest on this week's episode of the podcast is Rishabh Jain, the CEO and co-founder of FERMÀT Commerce, an eCommerce advertising optimization platform. I invited Rishabh back to the podcast to discuss OpenAI's recent announcement that it will bring advertising to ChatGPT.”

More About Naavik

Naavik's team of experts has helped over 300 companies — publishers, studios, tech companies, and investors — better succeed across the video game industry. We'd love to work with you too! Here's what we offer, spanning all platforms, genres, and regions:

• Strategy Consulting: Projects covering market research, corporate strategy, game & economy design, gamification, live ops strategy, AI strategy, product management, brand & performance marketing, and more.
• M&A and Investment Advisory: Expert commercial due diligence for buyers, fundraising support for sellers, and fractional CFO/CSO services.
• Fractional Talent: The one-stop shop for top-tier fractional talent covering dozens of game industry roles (analytics, design, marketing, art, QA, and more).

Check out the links above for more details. And if you'd like to chat about how Naavik can serve your team, click the box below or send us a note at [email protected].