When Every Machine Becomes an AI Agent With Physical AI

Thank you to our Sponsor: Viture

For most of the AI boom, AI has lived behind a screen. We asked chatbots questions, generated images, summarized documents, wrote software, searched databases, and deployed AI agents to perform digital work.

Now AI is starting to leave the screen.

It is moving into robots, cars, factories, warehouses, medical equipment, drones, appliances, wearables, and industrial machinery. Instead of simply generating an answer, these systems can increasingly perceive what is happening around them, make a decision, and take an action.

That is the big shift behind physical AI.

The first phase of the AI boom was about putting intelligence into software. The next could be about putting agency into machines. And when that happens, almost every machine around us has the potential to become an AI agent.

AI Is Leaving the Screen

Generative AI taught machines to understand language, images, video, software, and increasingly complex instructions. Physical AI takes those capabilities and connects them to machines that can affect the world around them.

That changes what AI actually means.

An AI system inside a robot does not simply need to know what a box is. It needs to recognize the box, estimate where it is, understand whether it can safely pick it up, determine how much force to apply, move around obstacles, and adjust if the box unexpectedly shifts.

The same principle applies to autonomous vehicles, factory equipment, drones, medical machines, and household devices.

• AI is becoming connected to sensors. Cameras, microphones, radar, lidar, temperature sensors, pressure sensors, and other devices give machines continuous information about their surroundings.

• AI is becoming connected to movement. Robots and autonomous machines can translate AI decisions into physical actions such as moving, lifting, sorting, steering, inspecting, or repairing.

• AI is becoming more local. More intelligence is moving from distant cloud data centers to computers running directly inside vehicles, robots, machines, and devices.

• AI is becoming continuous. Instead of waiting for someone to type a prompt, physical AI systems can continuously observe their environment and respond to changing conditions.

• AI is gaining agency. The important change is not simply that machines are becoming smarter. They are increasingly able to decide what should happen next and then make it happen.

Thank you to our Sponsor: Pokee AI

The Factory May Be Where Physical AI Arrives First

Humanoid robots attract enormous attention, but the biggest early impact of physical AI may happen somewhere less glamorous: the factory floor.

Manufacturers already operate environments filled with machines, sensors, cameras, programmable equipment, industrial robots, and enormous streams of operational data. Adding AI turns those systems from relatively rigid automation into something potentially far more adaptive.

NIST's 2026 roadmap for AI in smart manufacturing identifies autonomous systems, robotics, advanced sensing, digital twins, supply-chain optimization, and industrial analytics as important areas where AI is already advancing manufacturing.

• Machines can become more aware. Equipment can continuously analyze vibration, temperature, pressure, visual information, and other signals rather than simply following preset commands.

• Factories can become more adaptive. Production systems could automatically respond to changing materials, equipment conditions, demand, schedules, and unexpected disruptions.

• Maintenance can become predictive. AI can identify patterns indicating that equipment is beginning to fail and trigger maintenance before a breakdown stops production.

• Robots can become more flexible. Instead of being programmed for one repetitive movement, robots can increasingly learn tasks and adjust when objects, conditions, or workflows change.

• Factories can become coordinated systems of agents. Robots, production equipment, inspection systems, warehouse systems, and scheduling software could eventually communicate and coordinate decisions across an entire facility.

This is where the idea of an AI agent becomes much bigger than the software agents we are talking about today.

The agent is no longer just working inside a browser.

It may be running the factory.

Thank you to our Sponsor: Omane Media

Robots Are Becoming Embodied AI

Robotics may be the most visible expression of physical AI because it gives AI something it has never really had before: a body.

Companies including Figure, Agility Robotics, 1X, Boston Dynamics, Apptronik, Unitree, and others are trying to create machines capable of performing increasingly general physical tasks. Nvidia, meanwhile, is building an extensive physical AI ecosystem around models, simulation, computing hardware, and robotics development tools. Its partners include industrial and robotics companies such as ABB Robotics, FANUC, Figure, KUKA, Universal Robots and Yaskawa.

• Robots are learning rather than simply being programmed. AI models can help machines generalize from demonstrations, simulations, video, and previous experience.

• Vision-language-action models connect understanding with movement. A robot can potentially interpret an instruction, understand what it sees, determine what actions are required, and execute them.

• World models help machines understand consequences. Instead of merely recognizing objects, AI can learn how objects behave and predict what could happen after an action.

• Simulation is becoming a training ground. Robots can practice millions of interactions inside simulated environments before encountering similar situations in the physical world.

• General-purpose robots become possible. The long-term objective is not a robot programmed for one task, but a machine that can learn many different tasks using a common intelligence system.

The breakthrough will not necessarily arrive when robots look completely human. It will arrive when machines become flexible enough to perform useful work that was previously too unpredictable for traditional automation.

Cars, Drones and Machines Become Agents Too

Physical AI is much bigger than humanoid robotics.

An autonomous vehicle is essentially an AI agent with wheels. A drone is an AI agent with wings. A warehouse robot is an AI agent moving inventory. An agricultural machine can become an AI agent managing crops.

Once AI can perceive, reason, plan, and act, almost any machine becomes a potential platform for autonomy.

• Vehicles can make increasingly complex driving decisions based on cameras, maps, traffic conditions, pedestrians, weather, and the behavior of surrounding vehicles.

• Drones can become autonomous inspection systems capable of examining pipelines, power infrastructure, construction sites, farms, warehouses, and difficult-to-reach locations.

• Agricultural equipment can become more precise, identifying individual plants, applying treatments selectively, monitoring crop health, and adjusting operations according to local conditions.

• Construction equipment can gain greater autonomy, using AI perception and planning to assist with excavation, surveying, material movement, inspection, and site management.

• Infrastructure itself can become intelligent, with machines embedded throughout transportation, energy, utilities, logistics, and cities continuously responding to what is happening around them.

That means the physical AI economy could eventually become much larger than the robotics industry alone.

Thank you to our Sponsor: Partnerly

Even Everyday Objects Could Become AI Agents

The same shift will eventually reach consumers.

Today, calling a washing machine, refrigerator, pair of glasses, watch, or thermostat “smart” usually means that it has connectivity and some software features. AI could make the word smart considerably more meaningful.

Instead of waiting for explicit commands, devices could begin understanding context, predicting needs, coordinating with other systems, and taking limited actions independently.

• Wearables could become continuous personal assistants, understanding what users see, hear, and do throughout the day.

• AR glasses could give AI persistent visual context, allowing assistants to understand objects, locations, activities, and situations rather than relying solely on typed prompts.

• Home appliances could coordinate with one another, responding to household routines, energy prices, available supplies, schedules, and user preferences.

• Health devices could become more proactive, continuously interpreting sensor information and identifying situations that deserve attention.

• Homes could become collections of cooperating AI agents, with appliances, security systems, entertainment systems, robots, vehicles, and personal assistants sharing information and coordinating actions.

The smartphone put computing in our pockets.

Physical AI could put intelligent computing into nearly everything around us.

Thank you to our Sponsor: EezyCollab

Giving AI a Body Also Gives AI Consequences

There is one enormous difference between a chatbot and a physical AI system.

When a chatbot gets something wrong, it may produce an incorrect sentence. When an autonomous machine gets something wrong, it can move something, break something, damage equipment, disrupt production, or potentially hurt someone.

That makes safety central to the physical AI era.

Nvidia, for example, introduced Halos for Robotics in June 2026 as a safety architecture designed for robots and other physical AI systems that sense, decide, and act in the real world.

• Physical AI needs stronger guardrails. Machines must understand not only what they can do, but what they should never do.

• Human intervention must remain possible. Organizations need clear mechanisms for stopping, overriding, or limiting autonomous behavior.

• Actions need to be auditable. Companies will need records explaining what a machine observed, what it decided, and why it took a particular action.

• Cybersecurity becomes physical security. Hacking an AI system controlling machinery could have very different consequences from compromising an ordinary software application.

• Trust may determine adoption. The companies that win physical AI may not simply build the most intelligent machines. They may build the machines people trust enough to operate around humans.

When Every Machine Becomes an AI Agent

We are accustomed to thinking about AI agents as pieces of software.

That definition may soon feel remarkably narrow.

The much bigger opportunity is connecting AI to the enormous installed base of machines already operating throughout the economy and creating entirely new categories of autonomous machines alongside them.

• Factories could contain thousands of specialized AI agents coordinating production, inspection, maintenance, logistics, and energy consumption.

• Vehicles could become mobile AI platforms that perceive, plan, communicate, and increasingly operate autonomously.

• Robots could become physical coworkers, handling jobs that require movement, manipulation, navigation, and interaction with unpredictable environments.

• Consumer devices could become proactive agents, understanding context and acting without requiring a prompt every time.

• The physical world itself could become programmable, as AI gains the ability to connect digital intelligence with machines capable of taking action.

That may ultimately prove far more significant than putting a better chatbot on every computer.

The cloud gave AI intelligence at enormous scale. AI agents are giving that intelligence greater independence.

Physical AI gives it something more consequential:

the ability to act in the real world.

And once intelligence can see, decide, move, manipulate, and coordinate, the question may no longer be which devices contain AI.

It may be which machines don't.

New AI news site from Unaligned: 

We built an AI agent that reads 40,000 posts on X every day from the AI community and it finds the best info from all of that: https://alignednews.com/ai

We call it “Aligned News” and the same agent will start sending out its own newsletter every day. Please support it, it’s from our team. It will show you all the best discussion of news, AI papers, models, events, and much more.

Looking to sponsor our Newsletter and Scoble’s X audience?

By sponsoring our newsletter, your company gains exposure to a curated group of AI-focused subscribers which is an audience already engaged in the latest developments and opportunities within the industry. This creates a cost-effective and impactful way to grow awareness, build trust, and position your brand as a leader in AI.

Sponsorship packages include:

• Dedicated ad placements in the Unaligned newsletter

• Product highlights shared with Scoble’s 600,000+ X followers

• Curated video features and exclusive content opportunities

• Flexible formats for creative brand storytelling

📩 Interested? Contact [email protected], @samlevin on X, +1-415-827-3870

Just Three Things

According to Scoble and Cronin, the top three relevant and recent happenings

Nvidia’s $12.9 Billion Hugging Face Deal Expands Its AI Empire

Nvidia has reportedly agreed to acquire Hugging Face for $12.9 billion, a move that would give the chip giant control of one of the world’s most important platforms for sharing and developing open-source AI models. The deal would significantly expand Nvidia beyond chips and infrastructure, strengthening its push to become a vertically integrated AI company spanning computing, models, software, and applications. CNBC

AI Detects Heart Disease in Under Two Seconds

Researchers have developed an AI tool that analyzes routine ECGs to detect signs of heart failure and heart valve disease in under two seconds, identifying high-risk patients with up to 90% accuracy in trials. The technology could significantly speed diagnosis by prioritizing patients for follow-up heart scans and earlier treatment. The Guardian

Department of War Adds Grok AI to GenAI.mil

The Department of War has added Starshield AI's Grok for Government to its GenAI.mil platform, giving more than 1.7 million users access to secure, enterprise-grade generative AI for tasks such as planning, logistics, and acquisition. The deployment supports the Department's AI Acceleration Strategy by expanding mission-focused AI capabilities while promoting a broader U.S. AI ecosystem. U.S. Department of War

Scoble’s Top Five X Posts