Applied Reality is an Applied AI Lab building multimodal World Models that understand, simulate, and generate interactive environments.

We've been backed by Long Journey, Zuckerberg Family and Streamlined, early investors behind DeepMind, ByteDance, Anduril, and TogetherAI, and core researchers from OpenAI, Google Deepmind, Anthropic, and xAI to build the future scaling layer of AI.

Powered by a hybrid architecture of advanced coding agents with real-time diffusion, we enable infinite, accurate, and scalable environment generation- making the creation of games, robotics simulations, and interactive 3D worlds as effortless as generating an image.

Growth

Acqui-hired two gaming & product studios with 11.5M+ downloads and 100 games made for Voodoo and Supersonic.

Customers: Partnered with 100s of gaming and livestream creators (100K–10M followers each) for upcoming product launch Roam. We've achieved 50M+ views over last 14 days.

Team: Grew from 3 to 13 people in 9 months. HQ'd in SOMA, San Francisco, + Europe, 996.

Backing: Raised a round from investors behind DeepMind, ByteDance, Anduril, Together AI, Polymarket, Arielle Zuckerberg, staff from OpenAI, Anthropic, xAI, Deepmind, etc. Recently turned down an acquisition from a frontier lab.

World Engine: The Golden Ratio

Our stack delivers accurate worlds in any setting or functionality, with persistent state and no collapse, at ~50x lower cost than other approaches and supports 1M+ user concurrent scale.

Advanced multi-agent coding agents generate the complete structural and logical foundation, objectives, UI, sounds, VFX, and persistent world state.

ANIMA-0 (in-progress), our lightweight, real-time, 3D-conditioned diffusion model, acts purely as visual layer while preserving perfect temporal and multi-view consistency.

Product

Our first product Roam is a mobile world-generation platform that lets users make and share studio quality games from natural language, images, or multimodal prompts.

Games are the best across all engagement metrics, yet complex creation excludes 99.9% of people. The market is starving for a creation solution, seen in companies like Astrocade with 3M+ MAUs with terrible 2D game outputs. With studio comparable 3D game outputs we aim to hit 5M+ MAUs by end of summer.

Opportunity: Dual Use

While launching as a gaming powerhouse, every engineering hour spent on it contributes to us building the essential simulation layer for the physical AI era.

The Simulation Flywheel: Our causal 3D construction solves the 3D data scarcity problem, the most expensive bottleneck in robotics today. This has already secured high-level interest from several of the top 5 robotics companies, that we plan to aggressively pursue.

Interactive Entertainment: We are unlocking creation for the 99.9% of the population currently excluded from game development. By building tools with the correct UX, we're on track to bring world gen as the next interactive medium to everyone, far beyond the current game developer audience.

Roam is building the engine for the two greatest shifts of our time: new age of entertainment and physical AI. By capturing the convergence of environment generation, we're going after $100B+ market cap outcome.

Team

Team of 13 people, composed of four core divisions:

Applied AI: Ex-Moonshot (Kimi 2.5 post-training & agents), second time technical founders, Worked on multi-agent workflows, diffusion model training, RL-based world-simulation systems, on-premise Hugging Face-style platforms and OS projects with 10K+ downloads.

3D & Gameplay: Studio owners made 100+ mobile titles for Voodoo & Supersonic, games with 11.5M+ downloads, made games for Sidemen and several other global creators.

Infrastructure & Mobile: Ex-Trading View, Yandex (YaML) and senior iOS engineers who scaled consumer apps from 0 to 2 Million DAUs, built high-reliability distributed systems, and shipped large-scale SwiftUI and real-time multiplayer products across mobile and backend.

Growth & Marketing: 150M views in AI gen media for Wombo.AI and took [redacted] Series A company from 0 to 160M views 1BN+ impressions.

Technical Challenges

Hybrid Multi-Agent Orchestration: Coordinating distributed coding agents to create persistent structural logic, objectives, terrain, interactions, and multiplayer state.

Engineering multi-view and temporally consistent style-transfer models (ANIMA-0) that are aggressively distilled for mobile while maintaining perfect 3D grounding.

Building reinforcement learning pipelines for open-ended environment generation that feature hierarchical action spaces and dynamic, real-time difficulty adjustment.

Lightweight model training, that can run at top performance on consumer mobile hardware.

Company Details

Founded: late 2024

Raised: Seed round closed, Series A in progress

Team Size: 13 (all engineering)

Headquarters: San Francisco, SOMA