Version 1

AI is rapidly expanding beyond digital screens into physical environments across three parallel tracks: humanoid robots gaining capability through proprioceptive feedback rather than improved vision[1][2], AI perception layers making physical retail spaces machine-readable[3], and consumer wearables enabling AI agents to see and act in the real world through human bodies and autonomous task pipelines[4][5]. The dominant analytical voice is Rohan Paul, who is framing all these developments under a single thesis: physical properties — strength, balance, sensory feedback, body surface — are what unlock real-world AI utility, not appearance or screen interaction.

If AI's next frontier is physical space rather than digital interfaces, the bottleneck shifts from language and reasoning to embodied sensing and actuation. This reframes which hardware and architectures matter most, and raises novel questions about autonomy — including whether AI should act through robots, through purpose-built perception layers, or directly through human bodies wearing AI-guided gear.

A cluster of demonstrations in May 2026 crystallized a broader shift: AI systems are moving off screens and into physical environments, relying less on language models alone and more on real-time sensory data from the world.

The most technically specific example comes from Boston Dynamics, whose Atlas robot was demonstrated lifting and carrying objects exceeding 100 lbs[1]. Analyst Rohan Paul's read on the demo focused not on the payload figure but on the underlying mechanism: the robot adapts to weight, grip, and balance through proprioceptive feedback — body-internal sensing — rather than improved visual recognition. His argument is that this architectural choice, not raw strength or visual AI, is what makes humanoid robots viable for hard physical labor[1]. He extended this into a broader framework: humanoid robot value derives from physical properties (body surface area, strength, balance, sensory feedback) that let robots convert unstructured, messy objects into manageable ones — not from looking human[2].

The same physical-world expansion is happening in commercial environments. Paul also highlighted Radar, a company building what he describes as a perception layer for retail: infrastructure that makes physical stores machine-readable so AI can identify, locate, and reason about objects, people, and shelves in real time[3]. This frames physical retail not as a setting for human-facing apps but as a structured environment AI can directly perceive and act within — analogous to how the web became machine-readable through HTML.

On the wearable side, two demos illustrated how AI is using humans and consumer hardware as physical interfaces. At the MIT Hard Mode 2026 hackathon, six students built 'Human Operator' in 48 hours: a head-mounted camera feeds what the wearer sees to an AI that then issues instructions directing the user's physical actions — effectively using a person as an autonomous robot's body[4]. Separately, a demo pairing Meta Ray-Ban glasses with Gemini Live and the OpenClaw agent showed a different model: egocentric vision from the glasses, voice-triggered scene interpretation, and autonomous task routing that culminated in a completed purchase without direct human action[5]. Both pipelines converge on the same input — a camera that sees what a human sees — but diverge sharply on whether the human remains in the loop as the actuator or steps aside entirely.