Foundation models for the physical world

AppliedMind is an artificial intelligence startup that emerged in 2025 to address the specific challenges of robotics and physical automation. While much of the AI industry focuses on large language models (LLMs) that process text or vision models that analyze static or temporal imagery, AppliedMind is building foundation models designed for physical tasks. This shift represents a move from digital reasoning to embodied intelligence. The core technical hurdle they address is the high-dimensional complexity of the real world—friction, gravity, and material properties—which are often lost in purely digital training environments.

Their mission is to bridge the gap between teaching machines to move and allowing those machines to communicate with humans. This is a departure from traditional industrial robotics, which typically relies on rigid programming for specific, repetitive actions. AppliedMind's approach suggests a model that can generalize across different types of physical labor, adapting to new tasks with less human intervention than traditional methods required. However, they explicitly maintain a "human-in-the-loop" philosophy, which indicates that their models are designed to augment human work rather than replace it entirely with autonomous systems.

Communication beyond text and video

A critical differentiator for AppliedMind is their focus on modalities beyond text and video. In the context of physical tasks, communication often requires more than visual instruction or textual description. It involves force feedback, spatial orientation, and real-time physical adjustment. By developing models that can interpret and generate signals in these dimensions, AppliedMind is targeting the next stage of human-robot collaboration. This allows a human operator to guide or correct a machine in ways that feel more intuitive than typing a prompt or reviewing a video stream.

This technical focus places the company in competition with other physical AI players such as Physical Intelligence or Covariant. While those firms often focus on the "brain" for general-purpose robotics, AppliedMind's stated goal of augmenting human intelligence suggests they are building the infrastructure for collaborative physical agents. These agents could be applied in settings ranging from laboratory automation to complex assembly lines where human expertise is still required but physical repetition can be offloaded.

Early stage and market position

Founded in 2025, AppliedMind is a young entity operating in a highly technical niche. Their public presence is currently minimal, consistent with an early-stage deep-tech firm focusing on research and development. The name sometimes invites confusion with the legacy firm Applied Minds, LLC—the Burbank-based innovation shop founded by Danny Hillis and Bran Ferren—but AppliedMind is a distinct startup specifically oriented toward the foundation model era of the mid-2020s. Their use of a GitHub-hosted landing page and restricted public data points to a company currently in a phase of intensive model training or stealth-to-public transition. As physical foundation models become the new frontier for agentic behavior, AppliedMind is positioned to provide the underlying intelligence that allows digital agents to interact with the physical world.