No PriorsNo Priors Ep. 58 | The argument for humanoid robots with Brett Adcock from Figure
CHAPTERS
- 0:00 – 1:29
From Illinois farm to serial founder: software, Vettery exit, and the leap into hardware
Brett Adcock recounts growing up on a third-generation farm, learning to code early, and spending ~20 years building companies across software and hardware. He explains the arc from founding and selling Vettery to starting Archer Aviation, setting the stage for why starting Figure wasn’t a random jump.
- •Early background: farm upbringing and learning to build via coding
- •Career split: ~10 years software, ~10 years hardware/deep tech
- •Vettery sale (2017) as the catalyst to pursue harder hardware problems
- •Archer Aviation as the first major hardware bet before Figure
- 1:29 – 2:56
Why build an eVTOL company: traffic, sustainability, and “moving electric”
Sarah probes how someone decides to start an aircraft company. Brett explains the motivations—solving urban traffic and pushing sustainable transportation—and describes Archer’s electric VTOL vision and early execution steps.
- •Motivations: city traffic as an unsolved problem; belief in electric transport
- •eVTOL concept: helicopter-like, electric aircraft for short urban hops
- •Claimed value: major time savings (e.g., LA to SF in under 20 minutes)
- •Early execution: building a lab at University of Florida, then moving to California
- 2:56 – 4:09
Figure AI thesis: humanoids as the biggest market (human labor) + abundance + AGI impact
Brett lays out Figure’s core thesis: if humanoid robots are technically feasible, they can unlock an enormous market because human labor is roughly half of GDP. He also argues widespread robotic labor could drive down the cost of goods/services and potentially affect the AGI timeline through embodied data and capability.
- •Humanoid robots could be an order-of-magnitude larger business than transportation
- •Economic promise: robots doing daily work could create an “age of abundance”
- •Strategic goal: build a fleet + data engine to train robots for useful work
- •Belief that embodied robotics may influence the timeline to AGI
- 4:09 – 7:35
The case for humanoid form: the world is built for humans (not “optimal bodies”)
Sarah challenges the need for bipedal humanoids given instability and alternatives (more arms, different shapes). Brett argues the decisive point isn’t biological optimality but environmental compatibility: infrastructure, tools, and spaces are designed around human dimensions and capabilities, making humanoids a universal interface.
- •Two approaches: many specialized robots vs one general-purpose humanoid platform
- •Core argument: human-built environments favor human form factors
- •Humanoids as a general interface to the physical world (keyboard/mouse analogy)
- •Amortization: one platform can be reused across millions of tasks vs bespoke robots per task
- 7:35 – 12:33
Public demos and the “right time”: industrial autonomy, home manipulation, and speech-first UI
Brett describes Figure’s two demo tracks: industrial bin movement and consumer-style manipulation with speech-to-speech interaction. He then explains why now is the right decade—battery/motor improvements, better locomotion control, and modern AI/VLM capability enabling language as the default robot interface.
- •Demo 1: industrial bin movement—autonomous, end-to-end work in structured settings
- •Demo 2: home-style tasks—speech + vision to grasp objects, operate appliances, etc.
- •Architecture emphasis: neural nets from speech/video inputs to motion trajectories
- •Timing tailwinds: improved batteries/actuators, locomotion controllers, and AI/VLMs making speech a practical UI
- 12:33 – 15:11
Risk landscape: reliability, safety, huge action space, and the need for robot learning
Pressed on potential “walls,” Brett emphasizes the unprecedented difficulty: no one has commercially deployed humanoids at scale, and performance must be human-competitive and reliable over months/years. He highlights the combinatorial action space (30+ DoF) and argues scripting is impossible—robot learning is mandatory.
- •Commercialization risk: humans are productive; robots must match and sustain performance
- •Reliability and safety as gating factors for real-world deployment
- •High degrees of freedom create massive action/orientation space
- •Hand-coding behaviors doesn’t scale; learning-based approaches are required
- •Company mindset: “aggressively optimistic” + work intensity as mitigation
- 15:11 – 18:32
Building the team fast: mission/values, detailed org chart, and brute-force recruiting
Sarah asks how Figure assembled a multi-domain team quickly enough to achieve rapid progress. Brett explains his playbook: codify mission/values and a 10-year master plan, design the full org chart, then personally conduct hundreds of cold outreaches and close early technical leaders.
- •Foundational setup: mission/vision/values + culture doc + long-term master plan
- •Org design: explicit functions (controls, AI, actuation, batteries, kinematics, test, ID)
- •Recruiting tactic: identify “best in the world,” then cold email/call at scale (~300 calls)
- •Founder involvement: writes 30/60/90 plans, sits with engineers, attends engineering meetings
- 18:32 – 20:33
Deployment timeline: paid pilots, work cells, fleet management, and scaling manufacturing
Brett outlines near-term commercial deployment: customers paying now, robots delivered this year, and real facility pilots over the next 12–18 months. He frames the broader roadmap—reliability, fleet operations, scalable training and updates, and ultimately high-volume manufacturing.
- •Current status: companies paying; deliveries starting this year
- •12–18 month goal: robots doing real work in real work cells (task-specific stations)
- •Next hurdles: reliability/safety, fleet management, training and updates at scale
- •Manufacturing belief: once performance is solved, scaling production should be achievable
- 20:33 – 22:49
Build vs buy and vertical integration: forced to build without a mature supply chain
Sarah notes Figure’s extreme vertical integration (actuators, OS, etc.) and asks why, given the already huge problem. Brett explains the default preference is to buy, but the supply chain for humanoids is immature; integration and reliability needs force Figure to design and sometimes manufacture key systems themselves.
- •Build-vs-buy evaluated component-by-component; default preference is buying
- •Reality: humanoid supply chain doesn’t exist at required capability/performance
- •Integration pain: bring-up is messy; many parts break when systems meet
- •Result: building actuators, software stack, and other critical components becomes necessary
- 22:49 – 24:42
Product management in hardware+software: iterative design, requirements, and design gates
Brett describes how Figure manages product development: move fast via iterative build-and-test, grounded in clear requirements derived from customer needs. He details structured engineering gates (conceptual, preliminary, critical design reviews) and emphasizes continuous hardware/software iteration rather than long upfront research.
- •Iterative design as philosophy: build/test/learn continuously
- •Start from customer needs → translate into measurable requirements (payload, runtime, safety, IP rating)
- •Formal design reviews/gates to manage complexity across the organization
- •Continuous updates: hardware and software evolve indefinitely; “never good enough” mindset
- 24:42 – 28:26
Testing and rapid prototyping: objective trade studies and compressing long cycles
Comparing to software sprints, Brett explains hardware requires more objective decision-making and methodical trade studies because iteration cycles are 10–100x longer. The competitive advantage is compressing prototype-test loops; if you discover mistakes faster, you survive.
- •Software-like feedback loops exist, but hardware timelines are far longer
- •Top-down trade studies: e.g., actuator approaches (hydraulic/pneumatic/electromechanical; linear vs rotary)
- •Objective decision-making is critical because mistakes are expensive
- •Speed as a hiring/value focus: faster prototyping leaves room to recover from wrong choices
- 28:26 – 31:49
Corporate partnerships: BMW as first customer and OpenAI for robot “brain” reasoning
Brett discusses BMW as the first announced commercial customer, with planned deployments in Spartanburg, SC and a ranked set of initial use cases. He then explains the OpenAI partnership: OpenAI provides top-level vision-language reasoning and dialogue, while Figure focuses on low-level motor/hand control via neural networks.
- •BMW: robots shipped into manufacturing; target is real work within 12–18 months
- •Operational planning: five initial use cases selected and prioritized
- •OpenAI: contributes VLM + language reasoning; Figure handles embodied control/execution
- •Two-layer model: high-level task planning and dialogue + low-level control policies
- 31:49 – 33:36
Humans at home: chores, affordability via scale, and the path from industrial volume to consumer
Sarah asks what people will do with home robots; Brett focuses on practical physical labor: laundry, cooking, tidying, and daily chores. He argues costs will fall through design-for-manufacturing and experience curves, and that industrial deployments can bootstrap data and manufacturing volume before broad consumer adoption.
- •Primary home value: repetitive physical chores (laundry, dinner, cleaning toys)
- •Long-term vision: humanoids as general labor, making “labor optional”
- •Cost thesis: expensive today but driven down by DFM + scale/experience curves
- •Industrial rollout as the bridge to consumer: volume + data collection
- 33:36 – 35:26
Social acceptance: trust earned through safety record, gradual rollout, and sci-fi stigma
Brett argues social acceptance won’t happen overnight; it must be proven via safe, gradual deployment in industrial settings before entering homes. He notes branding and trust-building matter, especially given cultural fears shaped by sci-fi narratives about humanoid robots.
- •Acceptance depends on demonstrated safety and reliability, not persuasion
- •Gradual exposure: industrial environments first, then broader contexts
- •Brand and trust-building must be considered in every demo and deployment
- •Acknowledges sci-fi-driven stigma and the need to counter it with real-world performance
- 35:26 – 38:00
AGI vs robots: embodied actuation, avoiding a dystopian “humans as actuators,” and physical reasoning gaps
Sarah raises whether robots accelerate AGI by providing actuation for powerful models. Brett says he wants humanoids solved pre-AGI to avoid a future where humans are effectively coerced into being the actuators for software intelligence, and he argues current LLMs still struggle with physical planning—robot data may help close that gap.
- •Concern: without robots, humans remain the execution layer for increasingly capable systems
- •Goal: solve humanoid robotics before AGI to shift physical labor away from people
- •Observation: LLMs struggle with physical-world reasoning and action planning
- •Belief: embodied robotics could materially contribute to progress toward AGI over 5–10 years
