Overworld Raises $4.5 Million In Pre-Seed Funding Round

Listen to this article

Overworld, a startup specializing in AI driven world models, secured $4.5 million in pre-seed funding, signaling strong early investor confidence in local-first AI technologies for interactive simulations. The round was led by Kindred Ventures, with participation from several venture firms and notable angels, highlighting interest in AI that runs on consumer hardware to enable real time, playable worlds.

Overworld, formerly known as Wayfarer Labs, is a Providence, Rhode Island-based startup founded in 2025. It focuses on developing local-first diffusion world models, AI systems that generate interactive, adaptive digital environments running entirely on consumer devices like gaming PCs, Chromebooks, or consoles. Led by CEO Louis Castricato and Chief Science Officer Shahbuland Matiana, both former Stability AI researchers, the company emphasizes open science, real time interaction, and human directed creativity. Their flagship research preview, Waypoint-1, demonstrates real time world generation at up to 60 FPS with sub-20ms latency, licensed under Apache 2.0 and available on Hugging Face for experimentation.

The $4.5 million pre-seed round supports foundational development of AI native world models, aiming to eliminate data center dependency and reduce environmental impact. Investors include Kindred Ventures as lead, alongside Amplify.LA, Garage Capital, Northside Ventures, Vital Stage, East Sunshine, and angels such as Logan Kilpatrick (former OpenAI) and unnamed executives from Snowflake and Roblox. This marks Overworld’s first disclosed funding, with the company generating early revenue through its open source initiatives.

Evidence leans toward this funding accelerating Overworld’s role in the growing AI gaming sector, where local computation could democratize high fidelity simulations for robotics, entertainment, and beyond. However, as an early stage technology, it may encounter hurdles in achieving public ready stability, with broader adoption depending on community feedback from researchers and builders. The approach aligns with trends in edge AI, potentially influencing how interactive media evolves while respecting user privacy and control.

Overworld’s $4.5 million pre-seed funding round represents a pivotal moment for the nascent field of local-first AI world models, blending generative AI with real time interactivity to potentially transform gaming, simulation, and robotics. As a Providence, Rhode Island-based startup founded in 2025 and formerly operating as Wayfarer Labs, Overworld is spearheaded by co-founders Louis Castricato (CEO) and Shahbuland Matiana (Chief Science Officer), who bring expertise from prior roles at Stability AI. Their vision centers on creating “AI that dreams playable worlds,” emphasizing systems that run entirely on consumer hardware to deliver 60 FPS performance with sub-20ms latency, all while prioritizing openness, modifiability, and human agency. This funding, led by Kindred Ventures and supported by a mix of venture firms and high profile angels, underscores investor optimism in diffusion based technologies that could democratize advanced AI without relying on cloud infrastructure.

At its core, Overworld’s technology leverages diffusion world models, a subset of generative AI where data is progressively noised and then denoised to create coherent outputs. Unlike traditional diffusion models used for static image or video generation (such as those powering tools like DALL-E or Stable Diffusion), Overworld adapts this process for continuous, interactive environments. The model maintains a persistent world state in memory, updating it dynamically as users move or act, rather than producing isolated frames. This enables “playable” AI native worlds where user input influences every frame in real time. Key technical tenets include local execution on GPUs to minimize latency and environmental costs, broad accessibility across devices from Chromebooks to gaming consoles, and an open source approach to foster community driven innovation. The company’s research preview, dubbed Waypoint-1, exemplifies this: an experimental system released under Apache 2.0 license, hosted on Hugging Face, allowing researchers, engineers, and builders to experiment with real time world generation. Early demonstrations show capabilities like simulating adaptive environments at game level frame rates, though the preview acknowledges limitations in visual fidelity, stability, and depth compared to future public releases.

The funding round itself totals $4.5 million at the pre-seed stage, marking Overworld’s initial capital infusion and positioning it as a revenue generating entity through its open source and research oriented model. The capital is earmarked for advancing core technology, expanding the team, and building toward a comprehensive platform for AI native simulations. Kindred Ventures, known for backing AI pioneers in areas like human computer interaction and decentralized systems, led the round, aligning with their portfolio themes in AGI for good and new computing paradigms. Participating firms include Amplify.LA (focused on early stage LA tech), Garage Capital (Canadian seed investor in deep tech), Northside Ventures (early backers of innovative startups), Vital Stage (emerging AI and software investors), and East Sunshine (Asia-Pacific venture focused on tech frontiers). Angel investors add credibility: Logan Kilpatrick, a former OpenAI executive, brings AI scaling expertise, while unnamed senior leaders from Snowflake (data cloud) and Roblox (user generated gaming) signal relevance to data handling and interactive platforms. This investor mix reflects a bet on Overworld’s potential to bridge AI research with practical, consumer facing applications.

To contextualize the market, AI in gaming is exploding, with the global market estimated at $3.28 billion in 2024 and projected to reach $51.26 billion by 2033, growing at a 35.9% CAGR. Generative AI subsets, including world models, are key drivers, enabling dynamic content creation that reduces development costs and enhances player engagement. World models specifically address challenges in traditional gaming, such as static environments and high computational demands, by predicting and generating spatial dynamics in real time. Overworld’s local first approach differentiates it from cloud dependent competitors like OpenAI’s Sora or Google’s VideoPoet, which often incur latency and privacy issues. In robotics, the technology supports deployment for simulated training, while in entertainment, it could enable personalized, infinite worlds. However, the sector faces debates around AI’s environmental footprint and ethical use, Overworld’s on-device focus mitigates some concerns by reducing data center reliance. Broader industry trends show AI world models revolutionizing the $190 billion video game market through interactive 3D assets, with projections for AI generated 3D assets alone reaching $9.24 billion by 2032.

Community reception, as seen on platforms like X (formerly Twitter), has been enthusiastic. Overworld’s official account (@overworld_ai), with over 1,700 followers since joining in May 2025, announced the funding amid posts about the Waypoint-1 release, garnering likes, reposts, and integrations from developers (e.g., Daydream Scope plugin). Kindred Ventures publicly highlighted the investment, noting Overworld’s ex-Stability AI roots and applications in autonomy, robotics, and entertainment. Workshops and demos, such as those hosted by AI enthusiasts, indicate growing developer interest. Yet, early stage risks persist: the technology is experimental, and scaling to high fidelity public products will require overcoming computational bottlenecks on varied hardware.

Looking ahead, Overworld’s trajectory could influence how AI integrates into daily creativity. By inviting experimentation via Discord and open source tools, the company fosters a collaborative ecosystem. Blog posts on their site (e.g., on human evaluation in world models and optimizing diffusion decoders) provide deeper insights into their R&D, emphasizing balance between generation and reconstruction for stable outputs. If successful, this could lead to broader applications beyond gaming, such as virtual reality training or personalized education simulations. However, competition from established players and the need for robust evaluation metrics (as discussed in AI literature) will test their progress. Overall, this funding not only validates Overworld’s innovative stance but also contributes to the evolving dialogue on accessible, ethical AI.

Recommended: Cloudforce Raises $10M In Series A Funding Round

Please email us your feedback and news tips at hello(at)dailycompanynews.com