Dexterity, Inc.

This week: debating whether humanoid production volumes are real or CGI. Next week: another robot launch. Missing in all this is the real differentiator. Speaking to CNBC, Dexterity's CEO Samir breaks down why the flexibility of a Physical AI stack is the real competitive moat. To become useful for anything outside manufacturing—logistics warehouses, retail floors, pharmaceutical facilities, airport operations—robots need more than precision programming. Real-world applications require robots that can perceive, reason, and adapt in real-time. That's not a hardware problem. It's an AI problem. This is where Physical AI creates the unlock: Moving industries from rigid automation to dynamic flexibility at scale, regardless of the robot form factor. That's the real opportunity. Watch the full segment 👇 #PhysicalAI #Robotics

Physical AI at scale requires world-class automation infrastructure. Today we're announcing our partnership with Beckhoff Automation USA for Mech superhumanoid robot development. The technical challenge: Dual-arm manipulation at superhuman performance, operating safely in human workspaces, with real-time control at 1kHz—all in an exceptionally compact mobile form factor. Beckhoff's automation technology stack provides the foundation: → Real-time control for advanced kinematics → Safety over EtherCAT® (FSoE) for mission-critical environments → Compact hardware designed for space-constrained mobile robots → Enterprise-grade reliability our logistics partners demand Their Special Projects Team worked directly with our engineers to implement the control architecture enabling Mech's breakthrough capabilities. Building robots that supercharge people requires partners who share our commitment to performance, reliability, and safety. Grateful to Beckhoff for being that partner. Read more here - https://lnkd.in/e-cyyjXV #PhysicalAI #Robotics #Automation

In production, robots get 600 watts in a 55°C truck. You struggle running 3x in these types of industrial environments, making it near impossible to force every decision—from simple motion to complex manipulation—through billions of parameters. This is why we built a Transactable World Model for multi-agent coordination. Specialized agents. Shared world state. Real-time consistency. Optimized for the constraints of operating in the real world. Rather than one monolithic model trying to do everything and overheating, we use multiple agents transacting against reality as it actually exists. Watch our Founding Engineer Robert Sun break down why this architecture matters when Physical AI leaves the lab. Full presentation in the comments ↓ Thanks Foxglove for the opportunity to share our approach at #Actuate2025. #PhysicalAI #WorldModels #Robotics

Physical AI's hardest problem isn't what the robot sees. It's how multiple agents maintain shared understanding when reality constantly changes. Multiple agents sensing, controlling, and planning at different rates—all transacting against the same world state while objects shift and sensors occlude. Enterprise-grade Physical AI demands an interpretable system of record for spatial intelligence. Not end-to-end black boxes. Not foundation models that hallucinate. The video shows this in action: robot + world model + interpretable code + multi-agent timeline maintaining physical consistency in real-time. We introduced this concept a few weeks ago. Today's blog by our Founding Engineer, Robert Sun, goes much deeper into our technical architecture for a shared world model, the operator paradigm, and why interpretability is non-negotiable for Physical AI. Learn why this production-hardened Physical AI approach is trusted by industrial and logistics leaders across the globe 👉 https://lnkd.in/eJ3-Ti5s #PhysicalAI #WorldModels #Robotics

Three challenges define production Physical AI: 1. Superhuman performance—human-level isn't enough 2. The long tail—wet boxes, high altitude, 95% humidity, extreme cold 3. Reliability + safety at viable economics Solve one? Impressive. Solve two? Remarkable. Solve all three simultaneously? That's what changes an industry. This conversation between Eric Choi (The Dong-A Ilbo (동아일보) / Palo Alto Capital) and Dexterity's CEO, Samir Menon goes deep on what it takes to get robots and Physical AI to scale in production. Most importantly: why Physical AI must empower people, not replace them. Full interview 👉 [English Translation] https://lnkd.in/gDT6KCnr [Original Korean] https://lnkd.in/gWDunwVi #PhysicalAI #Robotics #Automation #Enterprise

Going into production, robots need to do more than see the world - they must understand it. Where exactly is each box face? What's the precise orientation of the object being grabbed? Which surfaces are accessible? How do objects relate spatially? The gap between vision and understanding is what Dexterity has been solving for years. Introducing Dexterity's Transactable World Model: the spatial intelligence layer that lets robots manipulate the physical world with >96% one-shot success in production. Swipe through to see how we did it. #PhysicalAI #WorldModels #Robotics #SpatialReasoning

What does a robot actually see when it loads a truck? Not what you think. Tomorrow: the world model that makes superhuman manipulation possible. #physicalai #robotics #worldmodels #spatialintelligence

We're honored to be named one of Bloomberg's Top 24 AI Startups to Watch in 2026. The inflection point we've been building toward is here: Physical AI moving from labs into production, from experiments to sustained enterprise-grade safety and performance. With robots that pick, pack, load, unload, palletize, and move in the chaos of real industrial environments, we're delivering transformational impact for customers today. This recognition marks a milestone, but it's just the beginning. Stay tuned. More to come soon. https://lnkd.in/e28jmyje #PhysicalAI #Robotics #AI #Innovation

Fragile. Oversized. Heavy. Awkward. Expensive. Everything robots are bad at figuring out how to move. Everything Dexterity was built for. Next week: we show you how. #PhysicalAI #Robotics #Automation #AI