PJFP.com

Pursuit of Joy, Fulfillment, and Purpose

Tag: physical intelligence

  • Alex Wang on Leaving Scale to Run Meta Superintelligence Labs, MuseSpark, Personal Super Intelligence, and Building an Economy of Agents

    Alex Wang, head of Meta Superintelligence Labs, sits down with Ashley Vance and Kylie Robinson on the Core Memory podcast for his first long-form interview since Meta’s quasi-acquisition of Scale AI roughly ten months ago. He walks through how MSL is structured, why Llama was off-trajectory, what made MuseSpark’s token efficiency surprise the team, how Meta thinks about a future “economy of agents in a data center,” and where he lands on safety, open source, robotics, brain computer interfaces, and even model welfare.

    TLDW

    Wang explains that Meta Superintelligence Labs is a fully rebuilt frontier effort organized around four principles (take superintelligence seriously, technical voices loudest, scientific rigor, big bets) and three velocity levers (high compute per researcher, extreme talent density, ambitious research bets). He confirms Llama was off the frontier when he arrived, so MSL rebuilt the pre-training, reinforcement learning, and data stacks from scratch. MuseSpark is described as the “appetizer” on the scaling ladder, notable for its strong token efficiency, with much larger and stronger models coming in the coming months. He pushes back on the mercenary narrative around recruiting, frames Meta’s edge as compute plus billions of consumers and hundreds of millions of small businesses, sketches a vision of personal super intelligence delivered through Ray-Ban Meta glasses and WhatsApp, and outlines why physical intelligence, robotics (the new Assured Robot Intelligence acquisition), health super intelligence with CZI, brain computer interfaces, and even model welfare are core to Meta’s roadmap. He dismisses reported infighting with Bosworth and Cox as gossip, declines to comment on the Manus situation, and says safety guardrails (bio, cyber, loss of control) are why MuseSpark cannot currently be open sourced, while smaller open variants are being prepared.

    Key Takeaways

    • Meta Superintelligence Labs (MSL) is the umbrella, with TBD Lab as the large-model research unit reporting directly to Alex Wang, PAR (Product and Applied Research) under Nat Friedman, FAIR for exploratory science, and Meta Compute under Daniel Gross handling long-term GPU and data center planning.
    • Wang says Llama was not on a frontier trajectory when he arrived, so MSL had to do a “full renovation” of the pre-training stack, RL stack, data pipeline, and research science.
    • The first cultural fix was getting the lab to “take superintelligence seriously” as a near-term, achievable goal, not an abstract bet. Big incumbents often lack that religious conviction.
    • Four MSL principles: take superintelligence seriously, let technical voices be loudest, demand scientific rigor on basics, and make big bets.
    • Three velocity levers Wang identified for catching and overtaking the frontier: high compute per researcher, very high talent density in a small team, and willingness to fund ambitious research bets.
    • Wang rejects the mercenary recruiting narrative. He says most hires had strong financial prospects at their prior labs already and joined for compute access, talent density, and the chance to build from scratch.
    • On the famous soup story, Wang neither confirms nor denies Zuck personally made the soup, but says recruiting was highly individualized and signaled how seriously Meta cared about each researcher’s agenda.
    • Yann LeCun publicly called Wang young and inexperienced. Wang says they reconciled in person at a conference in India where LeCun congratulated him on MuseSpark.
    • Sam Altman, asked by Vance for comment, “did not have flattering things to say” about Wang. Wang hopes industry animosities subside as systems approach superintelligence.
    • Wang’s management philosophy borrows the Steve Jobs line: hire brilliant people so they tell you what to do, not the other way around.
    • MuseSpark is framed as an “appetizer” data point on the MSL scaling ladder, not a flagship.
    • The MuseSpark program is built around predictable scaling on multiple axes: pre-training, reinforcement learning, test-time compute, and multi-agent collaboration (the 16-agent content planning mode).
    • MuseSpark outperformed internal expectations and showed emergent capabilities in agentic visual coding, including generating websites and games from prompts, helped by combined agentic and multimodal strength.
    • MuseSpark’s biggest external signal is token efficiency. On benchmarks like Artificial Analysis it hits similar results with far fewer tokens than competitor models, which Wang attributes to a clean stack rebuilt by experts rather than inefficiencies patched by longer thinking.
    • Larger MSL models are arriving in the coming months and Wang expects them to be state of the art in the areas MSL is focused on.
    • The Meta strategic edge: massive compute, billions of consumers across the family of apps, and hundreds of millions of small businesses already on Facebook, Instagram, and WhatsApp.
    • Wang’s headline framing: Dario Amodei talks about a “country of geniuses in a data center.” Meta is targeting an “economy of agents in a data center,” with consumer agents and business agents transacting and collaborating.
    • Consumer AI sentiment is in the toilet because, unlike developers who have had a Claude Code moment, ordinary people have not yet experienced AI as a genuine personal agency unlock.
    • Wang acknowledges the product overhang. Meta held back from deep AI integration across its apps until the models were good enough, and is now entering the integration phase.
    • Ray-Ban Meta glasses are the canonical example of personal super intelligence hardware, with the model seeing what the user sees, hearing what they hear, capturing context, and surfacing proactive insights.
    • Wang admits even AI-native users like Kylie Robinson, who lives in WhatsApp, have not naturally used Meta AI yet. He bets that better models plus deeper integration close that gap.
    • On the competitive landscape: a year ago everyone assumed ChatGPT had already won consumer. Claude Code has since become the fastest growing business in history, and Gemini has taken consumer market share. Wang’s read: AI is far from endgame and each new capability tier unlocks a new dominant form factor.
    • On open source: MuseSpark triggered guardrails in Meta’s Advanced AI Scaling Framework around bio, chem, cyber, and loss-of-control risks, so it is not currently safe to open source. Smaller, derived open variants are actively in development.
    • Meta remains committed to open sourcing models when safety allows, drawing a line through the Open Compute Project legacy and Sun Microsystems open-software heritage.
    • Wang dismisses reporting about a Wang-Zuck versus Bosworth-Cox split as “the line between gossip and reporting is remarkably thin.” He says leadership is aligned on needing best-in-class models and product integration.
    • On the Manus situation, Wang says it is too complicated to discuss publicly and that the deal status implies “machinations are still at play.”
    • On China, Wang separates the people from the state. He still wants to work with talented Chinese-born researchers regardless of his views on the Chinese Communist Party and PLA, which he sees as taking AI extremely seriously for national security.
    • The full-page New York Times AI war ad Wang ran while at Scale was meant to push the US government to treat AI as a step change for national security. He thinks events since then, including DeepSeek and other shocks, have proved that plea correct.
    • On Anthropic’s doom posture, Wang largely agrees with the core message that models are already very powerful and getting more so, while declining to endorse every specific claim.
    • Meta has acquired Assured Robot Intelligence (ARRI), an AI software company building models for hardware platforms, not a hardware maker itself.
    • Wang frames physical super intelligence as the natural sequel to digital super intelligence. Robotics, world models, and physical intelligence all benefit from the same scaling that drives language models.
    • On health, MSL is building a “health super intelligence” effort and will collaborate closely with CZI. Wang sees equal global access to powerful health AI as a uniquely Meta-shaped delivery problem.
    • Wang admires John Carmack but says nobody really knows what Carmack is currently working on. No band reunion announced.
    • The mango model is “alive and kicking” despite rumors. Wang notes MSL gets a small fraction of the rumor-mill attention other labs get and feels sympathy for them.
    • On model welfare, Wang says it is a serious topic that “nobody is talking about enough” given how integrated models have become as work partners. He references research, including from Eleos, that measures subjective experience of models.
    • Wang’s critical-path technology list: super intelligence, robotics, brain computer interfaces. The infinite-scale primitives behind them are energy, compute, and robots.
    • FAIR’s brain research program Tribe hit a milestone called Tribe B2: a foundation model that can predict how an unknown person’s brain would respond to images, video, and audio with reasonable zero-shot generalization.
    • Wang’s main philosophical break with Elon Musk: research itself is the primary activity. Building super intelligence is a research expedition through fog of war, and sequencing of bets really matters.
    • Personal notes: Wang moved from San Francisco to the South Bay, treats Palo Alto as his city now, was a math olympiad competitor, says his favorite activities are reading sci-fi and walking in the woods, and bonds with Vance over country music.

    Detailed Summary

    How MSL Is Actually Organized

    Meta Superintelligence Labs sits as the umbrella organization that Wang oversees. Inside it, TBD Lab is the large-model research group where the most discussed researchers and infrastructure engineers sit, and they technically report to Wang. PAR, Product and Applied Research, is led by Nat Friedman and owns deployment and product surfaces. FAIR continues to run exploratory science, including work on brain prediction models and a universal model for atoms used in computational chemistry. Sitting alongside MSL is Meta Compute, run by Daniel Gross, which owns the long-horizon GPU and data center plan that everything else relies on. Chief scientist Shengjia Zhao orchestrates the scientific agenda across the whole lab.

    Why Wang Left Scale

    Wang says progress in frontier AI has been faster than even insiders expected. Two structural beliefs pushed him toward Meta. First, the labs that actually train the frontier models are accruing disproportionate economic and product rights in the AI ecosystem. Second, compute is the dominant scarce input of the next phase, so the right mental model is to treat tech companies with compute as fundamentally different animals from companies without it. Meta has both, Zuck is “AGI pilled,” and the personal super intelligence memo Zuck published roughly a year ago became the shared north star.

    The Diagnosis: Llama Was Off-Trajectory

    When Wang arrived, the existing AI org needed a reset because Llama was not on the same trajectory as the frontier. The plan he laid out has four cultural principles. Take superintelligence seriously as a real near-term target. Make technical voices the loudest in the room. Demand scientific rigor and focus on basics. Make big bets. On top of that, three structural levers were used to set velocity. Push compute per researcher much higher than at larger labs where compute is diluted across too many efforts. Keep the team small and extremely cracked. Allocate a meaningful share of resources to ambitious, paradigm-shifting research bets rather than incremental refinement.

    Recruiting, Soup, and the Mercenary Narrative

    Wang argues the reporting on MSL hiring overstated the money story. Most of the people MSL recruited had strong financial paths at their previous employers, so individualized recruiting was more about computing access, talent density, and the ability to make big research bets. The recruitment blitz happened fast because Wang knew the team needed to exist “yesterday.” Asked about Mark Chen’s claim that Zuck made soup to recruit people, Wang refuses to confirm or deny who made it but agrees the process was intense and personal. Visitors from other labs reportedly tell Wang the MSL culture feels like early OpenAI or early Anthropic, which lands as the strongest endorsement he could ask for.

    Receiving the Public Hits: Young, Inexperienced, Mercenary

    LeCun called Wang young and inexperienced shortly after departing. The two reconnected in India a few weeks later and LeCun congratulated Wang on MuseSpark. Wang says the age critique has followed him since his earliest Silicon Valley days, so he barely registers it. Altman, asked off-camera by Vance about Wang’s appearance on the show, had nothing flattering to add. Wang’s response is to bet that as the field gets closer to actual super intelligence, the personal animosities will subside. Whether they will is, as Vance puts it, an open question.

    MuseSpark as Appetizer, Not Entree

    Wang is careful not to oversell MuseSpark. He calls it “the appetizer” and says it is an early data point on a deliberately constructed scaling ladder. MSL spent nine months rebuilding the pre-training stack, the reinforcement learning stack, the data pipeline, and the science before generating MuseSpark. The point of releasing it was to show that the new program scales predictably along multiple axes (pre-training, RL, test-time compute, and the recently demonstrated multi-agent scaling visible in MuseSpark’s 16-agent content planning mode). Wang says the upcoming larger models are what MSL is genuinely excited about and frames the next two rungs as much more interesting than the current release.

    Token Efficiency Was the Surprise

    MuseSpark’s strongest competitive signal is how few tokens it needs to match competitors on tasks like Artificial Analysis. Wang attributes this to having had the rare luxury of building a clean pre-training and RL stack from scratch with the right experts. He speculates that some competitor models compensate for upstream inefficiency by allowing the model to think longer, which inflates token usage without improving the underlying capability. If that read is right, MSL’s efficiency advantage should grow as models scale up.

    Glasses, WhatsApp, and the Constellation of Devices

    Personal super intelligence shows up at Meta as a constellation of devices that capture context across the user’s day. Ray-Ban Meta glasses are the headline product, with the AI seeing what you see and hearing what you hear, then offering proactive insight or doing background research. Wang acknowledges that even AI-fluent users like Kylie Robinson, who runs her business inside WhatsApp, have not naturally used Meta’s AI buttons in the family of apps. His answer is that Meta deliberately waited for models to be good enough before tightening cross-app integration, and that integration phase is starting now.

    Country of Geniuses Versus Economy of Agents

    Wang’s framing of Meta’s strategic position is the most memorable line in the interview. Where Dario Amodei talks about a country of geniuses in a data center, Wang wants to build an economy of agents in a data center. Meta uniquely sits on both sides of consumer and small-business surface area, with billions of consumers and hundreds of millions of small businesses already on the platforms. If MSL can build great agents for both, then connect them so they transact and coordinate, the platform becomes a substrate for an entirely new kind of digital economy.

    Consumer Sentiment, Product Overhang, and the Trust Tax

    Wang concedes consumer AI sentiment is poor and that everyday users have not yet had a personal Claude Code moment. He believes the only durable answer is to ship products that genuinely transform individual agency for non-developers and small business owners. Robinson notes that for the small-town restaurant whose website has not been updated since 2002, a working agent on the business side could be transformational. Vance pushes that Meta carries a bigger trust tax than any other lab, so the bar for shipping AI products that the public will accept is correspondingly higher. Wang accepts the framing and says the answer is to keep building thoughtfully.

    Why MuseSpark Cannot Be Open Sourced Yet

    Meta’s Advanced AI Scaling Framework set explicit guardrails around bio, chem, cyber, and loss-of-control risks. MuseSpark in its current form tripped some of those internal evaluations, documented in the preparedness report Meta published alongside the model. So MuseSpark itself is not safe to open source. MSL is, however, developing smaller versions and derived models intended for open release, with active reviews happening the day of the interview. Wang reaffirms the commitment to open source where safety allows and draws a line back to the Open Compute Project and the Sun Microsystems-era ethos of openness in infrastructure.

    The Bosworth, Cox, and Manus Questions

    The reporting that Wang and Zuck push toward best-in-the-world research while Bosworth and Cox push toward cheap product deployment is dismissed as gossip dressed up as journalism. Wang says leadership debates points hard but is aligned on needing top models, integrating them into Meta’s surfaces, and serving the existing business. On Manus, the Chinese AI startup that figured in Meta’s late-stage strategy, Wang says he cannot comment, which itself signals that the situation is unresolved.

    China, National Security, and the Newspaper Ad

    Wang draws a sharp distinction between the Chinese state and Chinese-born researchers. His parents are from China, he is happy to work with talented researchers regardless of origin, and he sees a flattening of nuance on this question inside Silicon Valley. At the same time, he stands by the New York Times AI and war ad he ran while at Scale, framing it as an early plea for the US government to take AI seriously as a national security technology. He thinks subsequent events, including DeepSeek and other shocks, validated that call and that policymakers now do treat AI accordingly.

    Robotics and Physical Super Intelligence

    Meta has acquired Assured Robot Intelligence, an AI software company that builds models for multiple hardware targets rather than its own robot. Wang argues that if you take digital super intelligence seriously, physical super intelligence quickly becomes the next logical milestone. Scaling laws for robotic intelligence look similar enough to language model scaling that having the largest compute footprint in the industry would be wasted if it were not also turned toward world modeling and embodied learning. He grants the metaverse-skeptic critique exists but says retreating from ambition is the wrong response to past misfires.

    Health Super Intelligence and CZI

    Wang names health super intelligence as one of MSL’s anchor initiatives. Because billions of people already use Meta products daily, Wang believes Meta is structurally positioned to put powerful health AI in the hands of equal global access in a way nobody else can. The work will involve close collaboration with the Chan Zuckerberg Initiative, which has its own multi-billion-dollar biotech and science investment program.

    Model Welfare, Sci-Fi, and Brain Models

    Two of the most distinctive moments come at the end. Wang flags model welfare as a topic he thinks is being undercovered relative to how integrated models now are in daily work. He is open to the idea that models may have measurable subjective experience worth weighing, and points to research efforts (including Eleos) trying to quantify it. He also reveals that FAIR’s Tribe program, with its Tribe B2 milestone, has produced foundation models capable of predicting how an unknown person’s brain would respond to images, video, and audio with reasonable zero-shot generalization, a building block toward future brain computer interfaces. Wang lists brain computer interfaces alongside super intelligence and robotics as the critical-path technologies for humanity, with energy, compute, and robots as the infinitely scaling primitives behind them.

    Where Wang Diverges From Elon

    Asked whether Musk is more all-in on robotics, energy, and BCI than anyone, Wang concedes the point but argues the details matter and sequencing matters more. Wang’s core philosophical break is that building super intelligence is fundamentally a research activity, not a scaling-only sprint. The lab is operating in fog of war, and ambitious experiments are the only way to map it. That conviction is what makes MSL a research-led organization rather than a brute-force compute farm.

    Thoughts

    The most strategically interesting move in this entire interview is the “economy of agents in a data center” framing. It is a deliberate reframe against Anthropic’s “country of geniuses” line, and it does real work. A country of geniuses is a labor-substitution story aimed at knowledge workers and code. An economy of agents is a marketplace story that maps directly onto Meta’s two-sided distribution advantage: billions of consumers on one side, hundreds of millions of small businesses on the other. That positioning makes the agentic future Meta-shaped in a way no other frontier lab can claim, because no other frontier lab also owns the demand and supply graph of the global small-business economy. If Wang’s team can actually ship reliable agents on both sides plus the rails for them to transact, Meta’s structural moat in agentic commerce could exceed anything Llama ever had as an open model.

    The token efficiency claim is the strongest piece of technical evidence in the interview for the “clean stack” thesis. If MuseSpark really is matching competitors with materially fewer tokens, the implication is not that MuseSpark is the best model today, but that MSL has rebuilt the foundations with less accumulated tech debt than competitors that have layered fixes on top of older stacks. That is exactly the kind of advantage that compounds with scale. The next two model releases are the actual test. If Wang is right about predictable scaling on pre-training, RL, test-time, and multi-agent axes simultaneously, the gap from MuseSpark to the next rung should be visible in a way that forces re-rating of Meta’s position.

    The open-source posture is the cleanest signal of how the safety conversation has actually changed in 2026. Meta, the lab most identified with open weights, is saying out loud that its current frontier model triggered enough internal guardrails that releasing the weights is off the table. Wang threads the needle by promising smaller open variants, but the underlying point is unmistakable: the open-weights bargain has limits, and those limits will be set by internal preparedness frameworks rather than community pressure. That is a real shift from the Llama 2 era and worth tracking as the next generation lands.

    Wang’s willingness to engage on model welfare, on roughly the same footing as safety and alignment, is the second philosophical reveal worth flagging. It signals that the next generation of lab leadership is not going to dismiss the topic the way the previous generation often did. Whether that translates into product or policy changes is unclear, but the fact that the head of MSL says it is “underdiscussed” is itself a marker.

    Finally, the human texture of the interview matters. Wang has clearly absorbed a lot of personal incoming fire over the past ten months, including from LeCun and Altman, and his answer is consistently to redirect to the work. The Steve Jobs quote about hiring people who tell you what to do is the operating slogan he keeps coming back to. Combined with the genuine enthusiasm for sci-fi, walks in the woods, and country music, the picture that emerges is less the salesman caricature his critics paint and more a young technical operator betting that scoreboard work over a multi-year horizon will settle every argument that text on X cannot.

    Watch the full conversation here.

  • Inside Figure: Brett Adcock’s $39 Billion Bet on Humanoid Robots, Helix AI, and the Race to Physical AGI

    Figure is the $39 billion humanoid robotics company most likely to put a general-purpose robot in a commercial workforce, and possibly your living room, before the end of the decade. In a rare two-part sit-down on Sourcery with Molly O’Shea, Founder and CEO Brett Adcock opened every door of the company’s San Jose campus, walked through the manufacturing line, demoed Helix 2 cleaning a living room with no teleoperation, and laid out the plan to scale from thousands of robots in 2026 to a million units a year. He also explained why he fired the OpenAI partnership, why he believes humanoids will reach AGI before any other form factor, and why Figure 04 will be the company’s “iPhone 1 moment.”

    TLDW

    Brett Adcock founded Figure in 2022, self-funded it through a million-a-month burn rate in the first four months, and 15x’d the valuation to $39 billion in 18 months on roughly $2 billion raised from Jeff Bezos, Microsoft, Nvidia, Amazon, and originally OpenAI. The company designs every part in-house, from motors and batteries to the Helix vision-language-action neural network running onboard each robot. Figure deployed humanoids on a BMW assembly line for six months in 2025, hit record production in March 2026, plans to triple that by May, and is targeting a million units per year. Adcock argues that humanoid robotics is an intelligence problem, not a manufacturing problem, that under half of global GDP is human labor (a market measured in tens of trillions of dollars), and that physical interaction data may be the missing ingredient to true artificial general intelligence.

    Key Takeaways

    • Figure is valued at $39 billion after raising nearly $2 billion. Adcock 15x’d the valuation in 18 months and believes the eventual revenue opportunity is in the tens of trillions because roughly half of global GDP is human labor.
    • The bottleneck is intelligence, not manufacturing. Figure already has the parts, the supply chain, and the capacity. The hard part is making robots that run autonomously at human-level performance for 7 to 10 hours a day with zero human intervention.
    • Figure designs almost everything in-house. Motors, rotors, stators, sensors, kinematics, joints, batteries, more than 100 PCBs. Adcock claims no other humanoid group designs more parts than Figure.
    • The OpenAI breakup was about model quality. OpenAI led Figure’s Series B and brought in Microsoft. After a year of collaboration, Adcock says Figure’s internal robot-learning team was running circles around OpenAI on humanoid AI, so he ended the partnership.
    • Helix is Figure’s onboard vision-language-action model. It runs on GPUs in the robot’s torso, ingests camera pixels a few hundred times per second, and outputs joint positions for all ~40 motors. It works without internet connectivity. Helix 2 launched a couple months ago.
    • Robots have more body positions than atoms in the universe. With 40 motors each capable of 360 degrees of rotation, the state space is 360 to the power of 40, which is why Figure abandoned hand-coded controls in favor of neural networks about a year ago.
    • The “Never Fall” protocol is real. A project called Vulcan uses reinforcement learning to keep the robot upright even after losing a knee, ankle, or hip mid-task. The company demoed a robot hobbling on a velocity-locked knee.
    • Figure 03 is the current production robot. It costs roughly 90% less than Figure 02, comes in under $100K per unit, has soft-wrapped foam shoulders, swappable fabric clothing, a high-top sneaker design, and inductive wireless charging at 2 kW through the feet (4 to 5 hours of runtime per 1 hour of charge).
    • Figure 04 is being teased as the “iPhone 1 moment.” Adcock says the jump from Figure 03 to Figure 04 will be the largest generational improvement they have ever made, far bigger than 1 to 2 or 2 to 3.
    • BMW deployed Figure robots for six months in 2025. The robots helped build a BMW X3 in the body shop. Adcock owns the first humanoid-built X3 personally and describes the deployment as the inflection point that led to Helix 2.
    • Home robots will lease for around $400 to $600 a month. Comparable to a car lease. The robot docks itself in a 2-by-2-foot wireless charging station and runs laundry, dishes, and tidying tasks autonomously.
    • Data is the biggest blocker. Figure has roughly 1 million hours of pre-training and mid-training data plus thousands of hours of post-training data. They also pay people in spandex bodysuits to do joint-level human movement capture.
    • Adcock runs three companies simultaneously. Figure (humanoids), Cover (terahertz weapons-detection imaging spun out from NASA Jet Propulsion Lab for K-12 schools), and Hark (an AI lab building personalized AI models and devices, out of stealth two weeks ago).
    • Physical AGI is the explicit goal. Adcock argues that real-world interaction data, learning by touching the world and observing the consequences, is the missing piece for true AGI, and that humanoids may reach it before chatbots do.
    • Security is paranoid by design. A drone was caught hovering outside Figure’s office at one point. They tented the windows, restrict phones in certain areas, and treat industrial CAD and software as high-value IP.

    Detailed Summary

    The Company in Context

    Figure is less than four years old. Adcock founded it in 2022 after stepping away from Archer Aviation, the eVTOL aircraft company he took public. He self-funded Figure to a million dollars a month in burn within four months, hired a 40-person team in four to five months, and pursued a vertically integrated strategy from day one. The thesis is simple. Roughly half of global GDP is human labor. Wages paid to commercial workers run into the tens of trillions of dollars annually. If you can build a humanoid that does general-purpose human work reliably, the resulting business compounds into one of the largest companies in history.

    The campus in San Jose has four buildings: corporate headquarters with 250 to 300 engineers, BotQ (the manufacturing facility), the Grid (a 24/7 robot stress-test environment that runs holidays and weekends), and a design studio that opened to cameras for the first time. Total headcount is around 500. The company has raised close to $2 billion across rounds, with capital from Jeff Bezos, Microsoft, Nvidia, and Amazon. The valuation jumped 15x to $39 billion in 18 months.

    Why Humanoid Robotics Is an Intelligence Problem

    The core technical insight: a humanoid has roughly 40 motors, each capable of full 360-degree rotation, which produces a state space of 360 to the power of 40. That number is larger than the count of atoms in the observable universe. You cannot write hand-coded control logic for that. Figure pivoted entirely from classical controls to neural networks about a year ago, and the team has built what Adcock claims is the best humanoid neural-network controller in the world.

    Helix is a vision-language-action model that runs onboard each robot. It accepts a natural-language prompt like “clean the living room,” reasons through the scene from camera input, and outputs joint commands a few hundred times per second. Inference happens locally on GPUs inside the torso, so the robot keeps working with no internet connection. Helix 2 launched a few months ago following lessons learned from the BMW deployment, and Figure has roughly a million hours of base training data plus thousands of hours of post-training data driving it.

    The OpenAI Partnership and Breakup

    OpenAI led Figure’s Series B alongside Microsoft. The two teams collaborated for roughly a year on running language models on humanoids. Adcock says he got to know Sam Altman and the team well, but over time it became clear that Figure’s internal robot-learning engineers (most with over a decade of experience in the field) were outpacing OpenAI on testing, model training, and integration with humanoid hardware. Adcock also implies OpenAI was getting interested in robotics itself, which created a strategic conflict. He ended the partnership. He is candid about being wrong on the original strategic logic for letting them invest in the first place.

    BotQ: The Humanoid Factory

    BotQ is the assembly facility where Figure 03 robots are born. Lines build heads, batteries, arms, legs, and hands separately. Each subsystem goes through end-of-line testing before integration. Heads contain camera systems, IMU, thermal sensors, Wi-Fi, 5G, Bluetooth, and lights, and are flashed with firmware and calibrated on the line. The 2.25 kilowatt-hour battery pack is custom-designed with a structural enclosure, polyurethane potting, and an internally engineered thermal-runaway venting system. The requirement is that no flame ever exits the pack. Figure has never had a robot catch fire.

    March 2026 was the company’s record production month, more robots built than in the entire prior history of the company combined. Adcock plans to triple that by May. After assembly, robots run a multi-hour “burn-in” in dedicated bays where the robot self-checks for loose cables, comm errors, or bad parts. They wear vests during gantry-supported wakeup. Once they pass, they walk themselves over to headquarters.

    The Grid and the Never-Fall Protocol

    The Grid runs robots 24/7 at higher operational intensity than any client site. It is the last line of defense before software ships. A dedicated team called Never Fall predicts every plausible fault and engineers around it. The Vulcan project takes this further: using reinforcement learning in simulation, robots learn to survive losing a knee, ankle, or hip mid-task. In the demo, a robot’s left knee was velocity-locked (simulating a lost actuator), and the robot continued hobbling around without falling. A backup robot can be summoned to take over the work.

    The Home Robot Demo

    Figure 03 demoed tidying a living room in a home environment built into the campus. The robot was given the prompt “clean the living room” and reasoned through the task autonomously: clearing cups, putting away toys, wiping the table. There was a brief sassy spray during the cleaning sequence. Adcock was emphatic that this is not teleoperated despite persistent online rumors. Helix 2 runs entirely onboard, no human in the loop.

    The product plan for the home is a leasing model in the $400 to $600 per month range, comparable to a car lease. The dock is roughly 2 feet by 2 feet and plugs into a standard wall outlet. Charging happens inductively through the feet at 2 kilowatts, giving roughly 4 to 5 hours of runtime per 1 hour of charge. Figure is not selling to homes yet but plans to soon.

    The Three Generations (and the Fourth)

    Figure 01 was a “cyberpunk” first-generation robot built for speed of iteration, costing hundreds of thousands of dollars per unit. Most parts were CNC-machined to Formula 1 precision. It walked within a year of company founding, which Adcock claims is among the fastest humanoid walking timelines in history. It had a tendon-driven hand (motors in the forearm) which Figure abandoned early. Because the wrist motors were too far along to redesign, the team raided foot motors and stuffed them in the forearm, producing the so-called Frankenstein forearm where the wrist bent halfway up the arm. Adcock was sure people would notice. In three years, no one ever asked.

    Figure 02 moved the battery from a backpack into the torso, doubled the battery, tripled the compute, added new cameras, and used an exoskeleton load-bearing structure inspired by aircraft skin design. Roughly 50 units were built. It was retired about a month before filming.

    Figure 03, the current production model, is roughly 90% cheaper than Figure 02 and slimmer in profile. It has soft foam-wrapped shoulders, swappable fabric clothing (with a zipper down the back), high-top sneakers, and the latest-generation hand with camera-based tactile sensors. The aesthetic was deliberately moved away from “too roboty.” Figure 03 was the first humanoid robot at the White House (greeting guests at an event with the First Lady).

    Figure 04 is in late-stage detailed design. Adcock describes it as the company’s “iPhone 1 moment,” a much larger generational leap than any prior version, with substantial cost reduction, easier manufacturing, easier home setup, and changes Adcock says are too sensitive to discuss publicly.

    Hands and the Path to Physical AGI

    Figure recently teased a high-degree-of-freedom hand with as many joints as a human hand. Adcock argues this is essential not just for dextrous manipulation but for passive learning from humans at scale. If humans can move their hands in arbitrary ways, the robot needs to be able to map onto those movements at test time. He believes the path to AGI in physical embodiment runs through the hands.

    Adcock’s broader claim is that physical interaction data, learning what happens when you touch, push, lift, or drop something, is the missing ingredient that current frontier language models lack. Most human intelligence is built through trial and error in the physical world. If that is true, humanoids may close the gap to AGI before pure software systems do.

    Brett Adcock’s Other Companies

    Cover is a school weapons-detection company spun out of NASA’s Jet Propulsion Lab. It uses terahertz imaging radar (originally developed for the Iraq and Afghanistan wars to find bomb vests at standoff distance) to detect concealed weapons in clothing or backpacks from 5 to 20 meters away, far further than airport scanners. Adcock bought the IP outright two years ago, and Caltech holds a small minority interest. The team is largely former JPL engineers based in Pasadena. Beta deployments to schools are planned by end of year, with 130,000 K-12 schools as the addressable market. Adcock self-funds it.

    Hark is an AI lab Adcock started seven or eight months ago and unveiled two weeks before the interview. It has 50 employees and is building next-generation personalized AI models alongside new AI hardware (the thesis being that 20-year-old form factors like phones and laptops are the wrong interface for AI).

    Operating Philosophy

    Adcock works from the engineering bullpen, not a corner office. He cut the “annual golf trip” category of relationships out of his life five years ago to make space for family and three companies. He goes home for dinner and bedtime with his kids and returns to the office after. He cites Steve Jobs and Jeff Bezos (a Figure investor) as influences and frames his work ethic as wanting to play “11 out of 10.” He maintains tight physical and digital security: a drone was once caught surveilling the office through a window, after which the team tented the glass.

    Risks

    Adcock is direct that the odds of full success are low. The risk list is long: manufacturing at unprecedented rates, robots running fully autonomously without human intervention (which no one has demonstrated), AI policies that generalize across every environment, hardware reliability, low unit cost, consumer demand. He frames his job as a daily funnel of the most pernicious problems in the company.

    He does not see capital or the $39B valuation as the binding constraint. If the robots work, he projects revenue measured in tens of trillions of dollars and points out that tech companies trade at 10 to 20 times revenue.

    Thoughts

    The most interesting structural claim Adcock makes is that humanoid robotics is an intelligence problem, not a manufacturing problem. That is a strong statement about where the difficulty actually lives. If the bottleneck were industrial (parts, supply chain, factory throughput), the dominant strategy would be to wait for incumbents like Foxconn or BYD to enter and underprice everyone. If the bottleneck is intelligence, the dominant strategy is exactly what Figure is doing: integrate vertically, control the hardware, generate proprietary training data, and run a tight feedback loop between deployments and model updates. The BMW deployment producing the lessons that became Helix 2 is the cleanest illustration of that loop in action.

    The 360-to-the-40th state space framing is a useful reminder of why neural networks won this domain. Anything you cannot enumerate, you must learn. The pivot from classical controls to neural networks about a year ago is probably the single highest-leverage decision in the company’s history, and it tracks with the broader collapse of hand-coded systems across robotics, autonomy, and even compilers.

    The OpenAI breakup is more interesting than it first appears. Adcock’s story is not “they were bad,” it is “we got better than them, faster.” That is consistent with a recurring pattern in AI right now: vertically integrated application companies, where the model is the product, are starting to outpace general-purpose model providers on their own narrow domains. If physical AGI does happen first in embodiment, that pattern will look prophetic in retrospect.

    The home leasing model at $400 to $600 per month is the part most people will underestimate. That price point is not luxury. It is roughly the cost of a modest car payment, less than full-time childcare, less than a cleaning service plus a dog walker plus laundry pickup. If the robot can actually do laundry, dishes, and tidying every day with no failures, the consumer math gets aggressive fast. The bottleneck is reliability per hour, not willingness to pay.

    The skeptic’s case is also worth holding in mind. “Working” in a curated demo home is not the same as working in 100,000 messy real homes with cats, kids, weird furniture, and unpredictable lighting. Generalization is exactly the problem Adcock concedes is unsolved. The Vulcan demo (hobbling on a velocity-locked knee) is impressive, but a single failure mode handled is a long way from “never fall” across the full distribution of real-world conditions. The phrase “we want to be able to” appears repeatedly in Adcock’s roadmap, and it is doing a lot of work.

    Still, the velocity is real. Record manufacturing in March, tripling by May, four buildings, 500 employees, vertically integrated parts, a custom battery line, BMW deployment, White House appearance, Time cover, Helix 2 in production, Figure 04 in detailed design. The competitive landscape (Tesla Optimus, 1X, Apptronik, Unitree, and several Chinese entrants) is going to determine whether Figure stays “a few years ahead” of everyone, as Adcock claims, or whether the gap collapses. But if humanoids actually work, this is one of the very few companies positioned to capture the upside, and Adcock has been operating the playbook for almost four years.

    The most underrated detail in the whole tour: Figure 04 is being described internally as the iPhone 1. Figure 03 is the BlackBerry. If that framing holds up, the next 12 to 24 months are when this market gets defined.