What is manifested AGI? It is artificial general intelligence expressed in physical form — a machine that can see, walk, lift, and work in the real world at a level comparable to a human expert. The term was coined by Jeff Brown of Brownstone Research, but the concept it describes is real and is being built right now by at least seven companies with functional prototypes. This page explains the concept. It does not sell a newsletter.
The “manifested” part is the distinction that matters. AGI as a software system — a language model that passes every benchmark you throw at it — lives on a server. You interact with it through a keyboard and a screen. Manifested AGI is different. It has a body. It navigates physical space. It uses cameras to see, motors to move, and hands to manipulate objects. The intelligence is the same. The delivery mechanism is different.
The Difference Between AGI and Manifested AGI
Artificial general intelligence, in the technical sense, means an AI system that can perform any intellectual task a human can perform. It is a capability benchmark, not a physical one. You could theoretically have an AGI that exists only as a chatbot.
Manifested AGI takes that capability and puts it in a machine that can act on the physical world. The AGI is the brain. The robot is the body. The connection between them is the software that translates reasoning into physical action — picking up a box, walking across a room, driving a car through traffic.
The reason this distinction matters for investors is that the hardware is where the supply chain lives. An AGI that exists only in software requires compute infrastructure — GPUs, data centers, power. An AGI that is manifested into a robot requires all of that plus motors, sensors, cameras, power electronics, batteries, and the manufacturing capacity to assemble them at scale. The supply chain for manifested AGI is larger and more complex than the supply chain for software AI.
Who Is Building It
The humanoid robotics field has moved from research papers to functional hardware faster than most analysts predicted. As of 2026, at least seven companies have commercial humanoid robots in various stages of deployment:
Tesla — Optimus is the most visible. The Gen 3 version stands 5’8” and 160 pounds, lifts 150 pounds, and has a hand with 22 degrees of freedom (the human hand has 27). Tesla’s advantage is data. Every Tesla vehicle on the road collects real-world driving video that trains the same neural network architecture used in Optimus. The robot navigates using the same vision-based AI that drives the car.
Figure AI — A robotics startup that released video of its Figure 01 robot making coffee using neural networks. Figure has partnered with OpenAI, which brings language model capability to the robot’s reasoning layer. BMW has deployed Figure robots in its South Carolina factory for material handling.
Unitree — A Chinese robotics company producing the H1 and G1 humanoid robots at price points significantly below Western competitors. The G1 is priced around $16,000, which is notable because it approaches the consumer-affordable range before the market has matured.
Agility Robotics — Maker of Digit, a bipedal robot designed for warehouse and logistics work. Digit is already deployed in Amazon fulfillment centers under a pilot program. Agility has focused on a specific use case — moving boxes in a warehouse — rather than general-purpose humanoid form.
Apptronik — An Austin-based company whose Apollo robot is designed for industrial work. Apptronik has a partnership with Mercedes-Benz for factory deployment.
1X Technologies — A Norwegian company backed by OpenAI. Its Neo robot is designed for home assistance. 1X has focused on the consumer market from the beginning, which is a different bet than the industrial-first approach of most competitors.
Boston Dynamics — The oldest name in the field, now owned by Hyundai. The Atlas robot is the most technically advanced humanoid in terms of dynamic movement, but Boston Dynamics has been slower to commercialize than the newer entrants.
Seven companies with functional hardware means the market is real. It also means the supply chain for humanoid robots — motors, sensors, chips, power electronics — is a multi-customer market, not a single-company bet.
Why Vision Is the Key
Most of these robots use cameras as their primary sensory input. This is not accidental. Vision-based AI has a structural advantage over sensor-fusion approaches that rely on LiDAR and radar, and the advantage is data.
Tesla’s autonomous driving system runs on a neural network trained on billions of miles of real-world driving video. That dataset — 360-degree video from millions of cars, capturing every road condition, weather pattern, and traffic scenario — cannot be replicated by a competitor. It is the moat. When Tesla ported that same software to Optimus, the robot started with a pre-trained vision system that already understood how to navigate the physical world. It did not need to learn from scratch.
This is why the connection between autonomous driving and humanoid robotics is tighter than it appears. They are not two separate products. They are two applications of the same vision-based AI. The car is a robot that drives on roads. The humanoid is a robot that navigates buildings. The brain is the same. The body is different.
The Scale Question
Elon Musk has stated a production target of one million Optimus robots at approximately $25,000 each. One million units at $25,000 is a $25 billion market. That is significant but not transformative — it is roughly the size of the premium TV market.
The more aggressive estimate, from Jeff Brown, is one billion humanoid robots at $25,000 each — a $25 trillion market. That number assumes the market extends well beyond Tesla’s production to include competitors, industrial applications, and consumer adoption globally. For context, the global automotive market is roughly $3 trillion annually. A $25 trillion robotics market would be more than eight times the size of the current auto industry.
The gap between one million and one billion is three orders of magnitude. That is the distance between a product launch and a civilizational shift. Whether that gap closes depends on whether the $25,000 price point holds, whether the use cases expand beyond industrial work to household assistance, and whether manufacturing capacity can scale to meet demand.
What This Is Not
Manifested AGI is not the same thing as artificial general intelligence. You can have AGI without manifestation — a purely software system that matches human intelligence. You can have manifested AI without AGI — a robot that does useful work but does not match human intelligence across all domains. The term “manifested AGI” specifically means both conditions at once: human-level intelligence AND physical embodiment.
The current state of the art is manifested AI without full AGI. The robots work. They perform useful tasks. They do not yet match human intelligence across all cognitive domains. Whether they will, and when, is the prediction that newsletter editors like Jeff Brown are making — and that researchers in the field are debating with more uncertainty.
The newsletter pitch built around this concept is here. The concept itself is free. The stock picks that operationalize it cost $179 through The Near Future Report.