Roblox Adds AI Founders to Boost Interactive Reality Vision

Roblox Corporation announced on June 5, 2026, that it has significantly bolstered its interactive AI capabilities by integrating the expertise of founders from Morpheus AI, Dynamics Lab, and Lucid AI. This strategic move aims to overcome key challenges in developing real-time, high-fidelity virtual worlds, including latency, consistency, and quality, bringing the company closer to realizing its ambitious Roblox Reality vision.

The integration brings aboard Xun Huang, the inventor of the "Self Forcing" technique and founder of Morpheus AI. Huang's breakthrough allows for the conversion of slow, offline video models into fast, autoregressive generation engines capable of real-time responsiveness. Morpheus AI's original goal of training egocentric video world models in pixel latent space to simulate real-world dynamics and physics is now finding its ultimate application within Roblox Reality. This technology is expected to enhance Roblox's existing Super Upsampler technology, adding visual components like higher-fidelity textures, realistic lighting, and fluid physics.

"For too long, video models have been trapped in slow, bidirectional offline generation," the company stated in its announcement. "Through years of post-training experimentation, researcher Xun Huang achieved a breakthrough and invented Self Forcing, a revolutionary recipe to convert sluggish offline video models into fast, autoregressive, interactive generation engines."

Convergence of Technologies

The acquisition is part of a larger effort to converge Morpheus AI's models with the Roblox Engine. While standalone world models offer powerful simulations, Roblox noted they often lack the long-term memory, consistent logic, and structured user input necessary for a massive multiplayer platform. The Roblox Engine, conversely, provides structured, symbolic logic and multiplayer state synchronization. The combined approach aims to create high-fidelity, interactive, multiplayer gaming experiences.

Further strengthening the team is Joe Chen, founder of Dynamics Lab, who joined Roblox in late 2025. Chen and his team specialized in building responsive, general-domain generative world engines that allowed users to upload images and transform them into live, interactive environments. Users could then prompt these worlds with text for surreal scenes or wild events, sharing links for instant multiplayer access.

Adding to the roster is Alberto Hojel, founder of Lucid AI, who joined Roblox in November 2025. Hojel's founding principle for Lucid AI was the necessity for AI-generated worlds to be reliable and consistent, balancing stunning visuals with deterministic game logic. He recognized that while video world models could produce impressive physics, they lacked the long-term state and predictable progression needed for true gaming.

Hojel has developed the Roblox game cartridge harness, which augments video world models with the deterministic logic of the Roblox Engine. This fusion is considered crucial for transforming explorable worlds into fully realized, high-fidelity multiplayer gaming platforms. The ambition for the Roblox Reality architecture targets 2K resolution at 60 Hz, representing a significant leap in video world model capabilities, requiring deep expertise in optimization and KV compression.

The company plans to move its Video Model (Super Upsampler) to adjacent edge data centers equipped with H200/B200-class GPUs. This infrastructure will support the complex computations required for real-time AI generation. The integration of game engines with video models is identified as the missing element for achieving consistent, photorealistic, multiuser worlds. This hybrid architecture seeks to leverage the structured simulation power of the Roblox Engine with the generative vision capabilities of models like those from Morpheus AI, ultimately creating a talent-dense environment to accelerate innovation in high-fidelity, interactive world creation.