Intelligence enters the physical world
Know what
moves next.
Independent, source-conscious guides to physical AI, humanoid robotics, embodied intelligence, robot models, and the companies building them.
Three lenses
Featured topics
Embodied AI
How bodies, environments, perception, and action shape intelligent behavior in robots and simulated agents.
02 / TopicHumanoids
A grounded guide to human-shaped robots, their design tradeoffs, applications, and the evidence needed to evaluate progress.
03 / TopicFoundation Models
An introduction to broadly trained models for robot perception, planning, language grounding, and action.
Latest guides
New · Evidence audit
Can Wording Change a Robot's Reward?
A practical audit of paraphrase fragility in robot reward models, including score crossing, flip rate, selection regret, mitigation, and deployment limits.
Read the guide →Living tracker
Which humanoids are actually available?
Separate announcements and demonstrations from pilots, sales, and recurring work.
Open the evidence tracker →Latest developments
Robot rewards need language invariance.
ROBORMBENCH holds robot behavior fixed and changes only the goal wording. Many evaluated models crossed from failure-level to success-level scoring on identical trajectories.
The new audit separates average score accuracy from paraphrase robustness, selection regret, mitigation, and physical deployment evidence.
Read the robot reward-language audit →Field notes
Understand what moves next.
A concise briefing on consequential physical-AI research, products, and companies. Subscribe to receive new editions when they are published.
Guides
Explanations built from first principles.
Explore →Topics
Navigate the field by concept.
Explore →Companies
A conservative starting directory.
Explore →Glossary
Decode robotics vocabulary.
Explore →Physical AI Guide is a new independent publication with a long horizon. We are starting with durable explanations and transparent editorial standards, then expanding carefully as the field develops.