Brett Adcock: Humanoids Run on Neural Net, Autonomous Manufacturing, and $50 Trillion Market

Humanoid robots with full-body autonomy are rapidly advancing and are expected to create a $50 trillion market, transforming industries, economy, and daily life

 

Questions to inspire discussion

Neural Network Architecture & Control

🤖 Q: How does Figure 3's neural network control differ from traditional robotics? A: Figure 3 uses end-to-end neural networks for full-body control, manipulation, and room-scale planning, replacing the previous C++-based control stack entirely, with System Zero being a fully learned reinforcement learning controller running with no code on the robot.

🎯 Q: What enables Figure 3's high-frequency motor control for complex tasks? A: Palm cameras and onboard inference enable high-frequency torque control of 40+ motors for complex bimanual tasks, replanning, and error recovery in dynamic environments, representing a significant improvement over previous models.

🔄 Q: How does Figure's data-driven approach create competitive advantage? A: Data accumulation and neural net retraining provides competitive advantage over traditional C++ code, allowing rapid iteration and improvement, with positive transfer observed as diverse knowledge enables emergent generalization with larger pre-training datasets.

🧠 Q: Where is the robot's compute located and why? A: The brain-like compute unit is in the head for sensors and heat dissipation, while the torso contains the majority of onboard computation, with potential for latex or silicone face for human-like interaction.

⚡ Q: What hardware runs Figure's neural networks onboard? A: Figure runs fast, low-power inference fully onboard using cheap hardware (not H100/GB300), enabling real-time policy deployment without draining the robot's entire power supply.

Manufacturing & Scaling

🏭 Q: What is Figure's 2026 production target? A: In 2026, Figure aims to achieve robot every 30 minutes production in Baku, with robots in commercial workforce running 24/7, where one robot learns a task and every robot in the fleet knows it.

📍 Q: What is the scale of Figure's development facility? A: Figure's 300,000-400,000 sq ft facility houses hundreds of robots and a large team focused on neural net development, which is the dominant factor in achieving human-like motion and behavior.

🏢 Q: When does Figure's Grid facility open and what's its purpose? A: Figure's Grid facility opens in January 2026, expanding to hundreds of robots running 24/7 for both home and commercial workforce, with mission control monitoring their performance.

💰 Q: How does vertical integration reduce costs? A: Figure vertically integrates all hardware (actuators, sensors, compute) and software (neural nets, data) to achieve 90% cost reduction in Figure 3, enabling rapid iteration and scalable robotics.

💵 Q: What is Figure's target price for mass adoption? A: Figure targets $20,000 per robot, allowing for mass adoption and the potential for tens of billions of robots on Earth by 2035-2040.

Hardware Design & Capabilities

👁️ Q: What sensory capabilities does Figure 3 have? A: Figure 3 has palm cameras for grasping occluded items, integrated tactile sensors in every fingertip, and plans for 360° cameras for comprehensive vision, mirroring human perception with cameras in head and hands.

🔋 Q: What is Figure's battery life and charging solution? A: Figure robots have 2kWh batteries lasting 4-5 hours per full charge, with 1-hour wireless charging through their feet using thin charging mats placed anywhere, enabling opportunistic charging while working.

📡 Q: How do Figure robots maintain connectivity? A: Figure robots have 3 communication systems (Wi-Fi, 5G SIM, Bluetooth) for always-on connection, but can perform tasks offline with high onboard intelligence to avoid being bricked.

🦾 Q: How many degrees of freedom does Figure 3 have? A: Figure robots have 40+ degrees of freedom and use neural nets for high-level planning (e.g., kitchen tasks), enabling complex whole-body manipulation.

🦶 Q: What design features improve Figure's walking and safety? A: Figure uses passive compliant toes for improved walking and reduced joint requirements, soft materials to minimize injury risk, and focuses on reducing pinch points in the robot's design.

Task Capabilities & Applications

☕ Q: What complex bimanual tasks can Figure 3 perform? A: Figure's Helix 2 neural net platform enables bimanual manipulation and logistics tasks like picking up, opening, and using a Keurig coffee maker, demonstrating end-to-end neural net control capabilities.

🏠 Q: What home care capabilities will Figure robots provide? A: Figure robots will enable aging in place at home, providing elder care and health monitoring to help people stay healthy, with ability to remember things, navigate homes like a visitor, and perform tasks over days and weeks.

🔪 Q: How do Figure robots adapt to different work environments? A: Figure robots can be outfitted with specialized soft goods for different applications, such as cut-resistant jackets for handling sharp or dusty materials, showcasing versatility and adaptability.

🏥 Q: What medical capabilities are planned for 2026? A: By 2026, Figure aims to have humanoid robots with surgical capabilities comparable to human surgeons, enabled by teleoperation and AI systems working at the highest performance level.

🧠 Q: How do Figure robots learn new tasks without manuals? A: Figure robots will learn any task without instruction manual by researching the internet, using digital tools, reasoning, and talking to humans, with only neural net weights updated while hardware remains unchanged.

Development & Training

⏱️ Q: How long did it take to achieve closed-loop kitchen task control? A: Figure achieved closed-loop control of human-like manipulation (e.g., kitchen tasks) using neural nets and 2 years of full-time development, a level of autonomy not seen in Chinese competitors.

🎯 Q: What system architecture considerations are critical for neural net control? A: Figure's all-in neural net approach requires careful consideration of sensor selection, operating system, middleware, firmware, and embedded software to enable human-like work and capabilities.

📊 Q: What are the key requirements for general-purpose humanoid robots? A: Building general-purpose humanoid robots requires novel, cheap hardware, neural nets working at scale, reliable daily operation without human intervention, and iterative design for high-rate manufacturing.

🔬 Q: What are the table stakes for solving general-purpose robotics? A: Figure aims to master neural nets for scaling, pre-training, generalization, and robots building robots for manufacturing at scale, which are table stakes for success in general-purpose humanoid robotics.

Safety & Reliability

🛡️ Q: What safety architecture is required for home deployment? A: Figure robots require a fault-tolerant, redundant real-time safety architecture and a proven safety track record before widespread deployment, achieving a safety bar where the robot operates fully autonomously around kids.

👶 Q: How will Figure robots be safe around humans and pets? A: Figure is developing intrinsically safe robots with superhuman perception and always-on computing, making them safer than humans around humans, animals, and pets.

Market & Economics

💼 Q: How could Figure robots generate income for owners? A: Figure robots will work 24/7, earning enough to hire additional robots to work for their human owner, potentially tripling their owner's salary.

🌍 Q: What is the expected timeline for humanoid robot proliferation? A: General-purpose humanoid robots are expected to arrive rapidly, with dramatic yearly improvement in capabilities, aiming for a future where humanoids outnumber humans in cities like San Francisco.

📈 Q: Why will general-purpose humanoids dominate the market? A: Figure's humanoid robots will be general-purpose, dominating the majority of the robot market, while other niche robots will be expensive and siloed for specific tasks.

Form Factor & Design Philosophy

🤸 Q: Why did Figure choose the humanoid form factor? A: Figure's humanoid robots have a human-like form with two arms and two legs, designed to replicate human capabilities in the cheapest and lightest way for safety and manufacturability.

🔄 Q: How does Figure's approach differ from specialized robots? A: Figure uses neural nets for high-level planning while maintaining 40+ degrees of freedom, contrasting with Archer's aircraft that use pilots for similar decision-making, emphasizing the general-purpose nature over specialized solutions.

Launch Timeline & Milestones

📅 Q: When was Helix 2 launched and what are its key features? A: Figure's Helix 2 robot launched in 2025, runs neural networks end-to-end for long-horizon, full-body control in unseen environments, with integrated tactile sensors in every fingertip and palm cameras.

🎯 Q: What is Figure's 2026 deployment goal? A: Figure's Helix 2 robot has a full end-to-end neural net stack designed for scaling pre-training data, with the goal of deploying robots at scale in industrial and commercial use cases in 2026.

Operational Capabilities

🔄 Q: How does fleet learning work across Figure robots? A: In 2026, Figure aims for general robotics where one robot learns a task and every robot in the fleet knows it, enabling rapid capability distribution across all deployed units.

🏃 Q: What level of autonomy do Figure robots achieve in unseen environments? A: Helix 2 runs neural networks end-to-end for long-horizon, full-body control in unseen environments, with replanning and error recovery capabilities in dynamic environments.

 

Key Insights

Neural Network Revolution

🤖 Figure's Helix 2 represents a fundamental shift from C++ code to an all-neural net approach for whole-body control, enabling rapid retraining and unexpected emergent behaviors that weren't possible with traditional programming.

🧠 Figure 3 integrates System Zero, a full-body reinforcement learning controller with no C code, enabling the robot to move its entire body using a learned controller for manipulation and perception as a world-first achievement.

⚡ Figure's robots run fast, low-power inference fully onboard using dedicated hardware instead of expensive GPUs, enabling real-time policy deployment at hundreds of Hz without consuming the robot's entire power budget.

🎯 The true challenge in robotics lies in achieving full end-to-end neural networks in unseen environments, generalizing to new places, and performing long-horizon work for days, not just pre-programmed open-loop behaviors.

📊 The key to scaling humanoid robots is accumulating large, diverse pre-training datasets to enable generalization, with the goal of building robots that perform tasks faster than humans with common-sense reasoning.

Hardware Integration and Performance

🔧 Figure vertically integrates all hardware and software, including actuators and neural nets, achieving 90% cost reduction in Figure 3 by avoiding reliance on low-readiness external supply chains.

🔌 Figure's robots charge wirelessly through their feet using 2kW inductive charging, providing 4-5 hours of runtime per full charge with the ability to charge while working through opportunistic charging and robot swarms.

📡 Figure's robots have multiple communication options (Wi-Fi, 5G, Bluetooth) for always-on connection, but can still perform tasks offline with onboard intelligence to avoid being bricked if internet is lost.

👁️ Helix 2 achieves room-scale autonomy with 40+ degrees of freedom, using cameras and palm sensors for onboard inference at hundreds of Hz, enabling complex tasks like grasping, planning, moving the body, and error recovery.

🎥 Figure's humanoid robots will have 360-degree vision with cameras in the head and on the body, enabling full situational awareness and the ability to interact with the environment.

Operational Capabilities

📦 Helix 2 can perform logistics tasks for days, using neural networks to grab, scan, and position packages at high speed and accuracy, with 67 hours of continuous operation across multiple robots.

🏠 Figure's robots can perform complex tasks like unloading dishwashers and folding laundry, which involve compliant materials and dynamic environments, but are still limited by the need for large-scale pre-training datasets.

🔄 Figure's humanoid robots achieve closed-loop control of human-like manipulation, a significant challenge not yet seen in Chinese competitors, requiring 2 years of full-time work to progress from tabletop to room-scale autonomy.

🛠️ Figure's humanoid robots have a modular design with interchangeable parts like hands and feet, allowing for task-specific adaptations such as grasping different materials or walking better, enabling rapid iteration and cost reduction.

Market and Competition

🌍 The humanoid robot industry is expected to consolidate globally to fewer than 10 serious players, with differentiation based on vertical applications and robot personalities, unlike the more straightforward differentiation in the automotive industry.

🚧 The shift to an all-neural net approach requires extensive data collection to create a barrier to entry for competitors, as building general-purpose humanoid robots requires novel, difficult hardware that is relatively cheap along with neural networks working at scale.

🎭 Most robot companies currently rely on teleoperation rather than full autonomy, with many videos showing humans controlling robots and putting out updates, which is not an impressive feat compared to true autonomy.

🤝 The partnership between Figure and OpenAI focused on using language models for semantic understanding in humanoid robots, but Figure's internal team outperformed them, leading to the decision to develop AI models in-house.

Deployment and Scale

🏭 Figure's general-purpose humanoid robots, developed over 3.5 years, will be deployed at scale in 2026 for both home and commercial workforce, with a facility running 250-300 robots 24/7 starting this month.

💰 Figure aims to produce millions of humanoid robots by 2026, with a target price of $20,000 per robot, enabling mass adoption and the potential for 5-10 billion robots in the commercial workforce by 2035-2040.

🎯 Figure's humanoid robots will have a mission control facility with real-time monitoring and data analysis of robot performance, similar to a military situation room, with video and telemetry from 250-300 robots.

📈 The humanoid robot market is expected to grow exponentially, with the potential for trillions in financing through credit card receivables and car leasing, as robots become personal assistants that can work multiple shifts and replicate.

Future Capabilities

🔬 In 2026, Figure's humanoid robots will have hardware capabilities comparable to surgeons, with teleoperation enabling learning of complex tasks, and the addition of infrared, ultraviolet, and tactile sensors boosting performance.

🧬 Figure's robots are fully integrated systems where the neural net stack is tightly coupled with the hardware, including sensors, compute, thermals, and low-level firmware, enabling advanced capabilities not possible with off-the-shelf robots.

🧠 Figure's humanoid robots have a high level of onboard computation, with the brain located in the head for efficient heat dissipation, and the potential for facial features or silicon overlays to enhance human-robot interaction.

Safety and Trust

🛡️ Safety is the top priority for deploying humanoid robots at scale, requiring solutions for semantic safety (understanding hazards like candles or boiling water), intrinsic safety around humans and pets, and superhuman perception for constant awareness.

👶 Figure robots are designed to be safe around children, with features like soft materials, no pinch points, and safety paddings on joints to prevent injury, similar to safety measures in automotive design.

🔒 Privacy concerns are addressed by being upfront about data collection, encryption, and cybersecurity, with a dedicated in-house team working on these issues for both the product and corporate side.

🏆 Figure's goal is to build general-purpose humanoid robots that can be trusted in homes, with a focus on safety, redundant real-time safety architectures, and a proven safety track record as the key barriers to entry.

Market Transformation

🌐 General-purpose humanoid robots will dominate the majority of the robotic market, with other robots being niche, expensive, and siloed for specific tasks, while humanoids will learn across a variety of tasks through transfer learning and rich data.

🏥 Figure's humanoid robots will enable aging in place at home, providing elder care and health monitoring to increase the value of in-home health services.

💼 The future of humanoid robots will involve massive abundance and universal high income, as individuals hire robots to work for them, potentially tripling their productivity by working multiple shifts and replicating.

⚡ The transition to general-purpose humanoid robots will happen rapidly, with the potential for more humanoids than humans in urban areas like San Francisco, as shown by dramatic changes in Figure's robots over the past two years.

Strategic Vision

🎯 Figure aims to solve general-purpose humanoid robotics by focusing on neural net scaling, pre-training, generalization, and robots building robots for mass production, with a goal of shipping billions of robots per year.

🔄 Building general-purpose humanoid robots requires neural networks working at scale for reliable, daily, human-free operation, which is a significant challenge that most companies haven't achieved.

🚀 Helix 2 represents a disruptive change in robotics, moving from a C++-based control system to an all-neural net approach, enabling a completely different future for humanoid robots with capabilities not possible before.

 

#SingularityNavigator #Abundance #StartupSocieties #AbundanceSociety

XMentions: @DigitalHabitats @Abundance360 @SalimIsmail @PeterDiamandis @SingularityU @NextBigFuture @adcock_brett @Figure_robot

WatchUrl:https://www.youtube.com/watch?v=S_fXhVT67Uw

Clips

• 00:00 🤖 Humanoid robots with full-body autonomy are advancing rapidly, expected to create a $50 trillion market, and transform the economy and daily life.
  • The development of humanoid robots, such as Figure's, is rapidly advancing with full-body autonomy, large language models, and neural nets, potentially leading to a $50 trillion market with widespread impact on the economy and daily life.
  • Figure's humanoid robots, powered by neural networks, have achieved full body autonomy, performing complex tasks like household chores without pre-programming, and are being mass produced at a significantly reduced cost and weight.
  • Figure's humanoid robot has achieved full body autonomy using an all-neural net approach, replacing 109,000 lines of C++ code, enabling smooth motion and human-like control through sensor data and inference on board.
  • Figure's humanoid robots aim to achieve full-body autonomy, requiring advanced AI models that understand physics and can plan and reason at fast dynamic speeds, which current large language models (LLMs) lack when applied to physical tasks.
  • The humanoid robot market is expected to consolidate to a few major players globally, likely less than 10, with current estimates suggesting a triopoly, similar to how the car industry consolidated to a few major companies.
  • The development of truly autonomous humanoid robots requires solving complex challenges such as general-purpose neural networks, full-body autonomy, and scaled manufacturing to enable robots to perform long-horizon tasks in unseen environments without teleoperation.
• 23:35 🤖 Figure develops humanoid robots with full-body autonomy, aiming to deploy them in various industries and tasks, with a $50T market potential.
  • Figure's robotics developments in 2025 included achieving long-period neural network runs, enabling robots to perform tasks like package logistics with high accuracy and speed.
  • Figure's development of Helix 2, a humanoid robot with full-body autonomy, marks a significant step change, integrating a fully learned reinforcement control system, allowing for zero-code, self-controlled movement and manipulation.
  • Figure's Helix 2 robot achieves full body autonomy with a neural network stack, enabling scalable pre-training data accumulation and sharing of learned tasks across the robot fleet.
  • Figure aims to scale up robot production and deployment in 2026, with goals including putting robots on production lines, scaling out robots in industrial and commercial workforces, and delivering Figure 3 robots to clients.
  • Figure aims to develop a humanoid robot with full body autonomy, common sense, and conversational abilities, powered by a unified model that integrates speech, language, physics, and memory to enable robots to assist humans in various tasks.
  • A single neural net is used for Figure's robots, as teaching separate neural nets for each physical motion or task would be too storage-intensive and processing-expensive, but positive transfer of data allows for better performance with more information.
• 35:29 🤖 Humanoid robots, like Figure's, are being developed to achieve full body autonomy, perform tasks independently, and could lead to a $50T market with potential applications in various industries.
  • Figure's humanoid robots aim to achieve full body autonomy using affordable and efficient computing hardware, enabling fast inference and policy deployment while running fully on board without relying on extensive power or internet connectivity.
  • Figure's humanoid robots have advanced communication capabilities, including Wi-Fi, 5G, and Bluetooth, and are designed to perform tasks autonomously with onboard intelligence, even when disconnected from the internet, and have improved battery life with a unique foot-based charging mechanism.
  • The battery life of Figure's humanoid robot is expected to be ample for tasks, with 4-5 hours of operation per charge and opportunistic wireless charging capabilities reducing the need for long battery life.
  • Achieving real-time closed loop control of a robot doing human-like tasks, such as manipulation and movement, requires significant development, as evidenced by the two years of intense work to progress from tabletop manipulation to room-scale autonomy.
  • The development of humanoid robots required vertical integration of hardware and software due to the complexity of combining various technologies, and the company plans to move its supply chain out of China to focus on a collaborative effort to advance humanoid robotics for the benefit of humanity.
  • Figure's development of humanoid robots requires vertical integration, with in-house design, manufacturing, and assembly, due to the unavailability of reliable and cost-effective off-the-shelf components.
• 44:05 🤖 The humanoid robot market is expected to reach $50 trillion, driven by advancements in full body autonomy and neural net-integrated hardware, revolutionizing industries and transforming the economy.
  • Figure's robotics advantage lies in its in-house development of neural net-integrated hardware, allowing for full body autonomy, which can't be replicated with off-the-shelf robots or those retrofitted with added hardware.
  • Chinese robotics companies lack closed-loop AI control systems, instead relying on open-loop systems with local hand controllers, which is orthogonal work from designing systems for full autonomy.
  • The humanoid robot market is expected to reach $50 trillion, drawing in major tech companies like Apple, Google, and Microsoft, as it has the potential to revolutionize the economy, lead to an age of abundance, and create a sci-fi future.
  • Building humanoid robots is extremely challenging from an engineering perspective, comparable to rocket design, requiring mastery of six key layers: mechanical structures, actuators, sensors, control software, embedded systems, and compute integration.
  • Figure's robotics and similar companies are shifting towards entirely neural network-based systems, where high-level behaviors and decision-making are performed by neural nets, potentially replacing traditional human expertise and control.
  • Humanoid robots are expected to play a significant role in various settings, including homes, elder care, and healthcare, enabling people to stay healthy at home.
• 53:17 🤖 Humanoid robots are being developed to achieve full-body autonomy, aiming to disrupt the $50 trillion market by providing personalized assistance in various settings, including senior care and homes.
  • Figure aims to deploy fully autonomous humanoid robots in various settings, including senior care and homes, to help people age in place and provide personalized assistance, potentially disrupting the $50 trillion market.
  • Figure is building a massive robotics facility to deploy hundreds of humanoid robots, aiming to achieve full-body autonomy and general-purpose capabilities, potentially leading to a $50 trillion market.
  • The next 12-18 months will see the largest AI transformation ever, with advancements in multimodal systems, synthetic humans, and robots that can reason, understand, and perform tasks like humans, potentially revolutionizing industries like healthcare.
  • Figure's robots are nearing full body autonomy, enabling advanced capabilities like real surgery and dextrous tasks, with the right data and hardware allowing them to learn and perform complex actions.
  • Figure's humanoid robots aim to match human capabilities at the lowest cost and weight possible, currently having headroom to increase speed by 3-5x with their actuators but are limited by software.
  • Industrial robots are expected to decrease in cost to $10-20,000 over time, making them more viable for home use, whereas currently, expensive robots are being considered for industrial use cases where safety and control can be better managed.
• 01:08:05 🤖 The humanoid robot market could reach $50 trillion with mass-produced robots like Figure's, designed for full-body autonomy and general-purpose tasks, potentially dominating the robotics market.
  • The market for humanoid robots could reach $50 trillion with tens of billions of robots potentially being built, with estimates suggesting 5-10 billion robots in the commercial workforce and a possible price point of $20,000 per robot.
  • A general-purpose humanoid robot with neural network-based system can be mass-produced and scaled to a $50 trillion market if two key challenges are solved: scaling neural nets to generalize and enabling robots to build other robots.
  • Figure's robots are being designed for full-body autonomy, with plans to ensure safety and privacy through intrinsic safety features, superhuman perception, and robust cybersecurity measures to enable widespread deployment in homes.
  • The company aims to mass-produce humanoid robots, with a focus on safe and general-purpose machines that can learn across various tasks, and predicts that humanoid robots will dominate the robotics market, making up the plurality of all robots.
  • The conversation includes a discussion about a robotics company's experiences with their humanoid robot, Figure, including a unique event where robots performed on stage with a DJ, Dead Mouse.
  • Figure aims to launch a general-purpose robot capable of long-horizon work in a home environment, with potential deployment as early as this year or next, but with a focus on ensuring reliability and safety.
• 01:27:30 🤖 Humanoid robots are expected to become ubiquitous, potentially outnumbering humans, with a potential $50 trillion market, and achieving full body autonomy with neural networks trained on diverse data.
  • The speaker trusts Figure's robot to be safe for general use when he feels safe enough to leave it fully autonomously in his home with his kids without supervision.
  • General purpose robots, particularly humanoid ones, are expected to become ubiquitous soon, with the transition happening rapidly, potentially leading to a future where robots outnumber humans in certain areas.
  • The CEO of Figure discusses the rapid development of autonomous robots, citing the example of Waymo, a self-driving taxi service that required 16-17 years of engineering work to create a seamless and safe user experience.
  • To meet the demand for a potential $50 trillion market for humanoid robots, humans could leverage robotic and engineering solutions to overcome challenges like mining in difficult environments such as Greenland or even asteroids.
  • Achieving robotic autonomy through neural networks trained on diverse data allows robots to generalize tasks, bounded only by the availability of data, not hardware limitations.
  • The company Figure developed a humanoid robot, iterating from Figure 1 to Figure 3, with significant design improvements, reduced costs, and advanced features such as cameras, actuators, and compute capabilities.
• 01:38:51 🤖 Humanoid robots are advancing with improved designs, autonomy, and capabilities, potentially transforming industries and even operating in space.
  • The conversation appears to be a casual discussion about the design and safety features of a humanoid robot, specifically its ventilation system and protective padding on joints.
  • Figure's new humanoid robot features a slimmer design, improved tactile sensors, reduced pinch points, and enhanced compute and thermal systems, with capabilities including 20kg weight carrying and advanced mobility.
  • The robot's design features three screens, cameras, and sensors in its head, allowing for various functions, onboard computation, and heat ventilation, while also considering aesthetics and potential uses.
  • Figure's robots can be outfitted with various soft goods, such as silicon faces, hair, and different outfits, including cut-resistant jackets and specialized gloves.
  • Figure's robots may operate in space, including assembling data centers, and eventually utilize materials from the moon and asteroid belt.
  • The host invites viewers to subscribe and join his weekly newsletter, Metatrends, which provides a two-minute read on meta trends impacting various aspects of life.

-------------------------------------

Duration: 1:43:48

Publication Date: 2026-02-12T08:01:27Z

-------------------------------------