Robohouse ’26 Library
Contents

Chapter 1

The State of the Field in One Page

3 sections · about 5 minutes

Robotics in 2026 is in the middle of a bet that has not yet paid off and has not yet failed.

The bet is that the recipe that worked for language — a single large model, pretrained on a very large and diverse corpus, then adapted cheaply to specific tasks — will work for physical action. The models are called VLAs (vision-language-action models). The corpus is the problem: there is no internet of robot behaviour, and there is no obvious way to make one.

Everything else follows from that. The academic labs have split into those trying to manufacture the corpus (Stanford's IRIS and REAL, MIT's Improbable AI, Toyota Research Institute), those arguing the corpus is the wrong idea and structure should replace scale (MIT's LIS group, the RAI Institute, most of the control-theory establishment), and those building the sensors and mechanisms everyone else will eventually need (GelSight, BDML, CHARM, the Movement Lab). The companies have split into those selling bodies (Figure, Unitree, Agility, UBTech), those selling brains (Physical Intelligence, Skild, Generalist, Google DeepMind), those selling the picks and shovels (NVIDIA, Hugging Face, the data-labelling industry), and those quietly selling actual working robots into actual warehouses (Dexterity, Ambi, Path, Chef, RightHand).

Four numbers frame the whole field.

  • ~3,000 hours. The size of the largest purpose-built real-robot manipulation dataset in existence, AgiBot World — 1,001,552 trajectories, 2,976.4 hours, collected by 100+ robots in a purpose-built 4,000 m² facility.
  • ~100,000 years. Ken Goldberg's estimate, in Science Robotics, August 2025, of the human-equivalent time represented by the text and image corpora used to train frontier VLMs. The ratio between these first two numbers is the entire problem.
  • 1,250 hours. The runtime of the most-cited real humanoid deployment in the West — Figure 02 at BMW Spartanburg — over eleven months, loading 90,000+ parts at one station. That is roughly seven months of one human's working time.
  • 73.6%. The share of Unitree's humanoid revenue that came from research and education buyers in the first nine months of 2025, per its IPO prospectus. The only company with audited humanoid revenue is mostly selling to people who study robots, not people who use them.

Hold those four numbers in mind and most of the rest of this book is commentary.

1.1 A compressed timeline, 2022–2026

DateEventWhy it mattered
Dec 2022Google RT-1First serious "transformer for robot actions" at scale
Mar 2023RSS: Diffusion Policy (Chi, Song, Tedrake et al.)Generative modelling of action sequences becomes the default architecture
Jul 2023RT-2A VLM fine-tuned to emit actions — the VLA template
Jul 2023ALOHA (Zhao, Finn)A <$20k bimanual teleop rig; democratises data collection
Oct 2023Open X-Embodiment21 institutions pool 1M+ trajectories, 22 embodiments; positive cross-embodiment transfer demonstrated
Jan 2024Mobile ALOHAAdds a mobile base; the viral moment for low-cost imitation
Feb 2024UMI (Chi, Song et al.)Handheld gripper decouples data collection from robot ownership
Aug 2024Amazon hires Covariant's foundersThe first "reverse acquihire" in robot foundation models
Oct 2024Physical Intelligence π0The first credible "one policy, many robots" commercial model
Nov 2024Stanford Robotics Center opensAcademic robotics gets a shared physical plant
Mar 2025NVIDIA Isaac GR00T N1An open humanoid foundation model; NVIDIA positions as the platform
Mar 2025Gemini RoboticsDeepMind folds robotics into its frontier model line
Apr 2025π0.5Open-world generalisation to unseen homes; diminishing returns above ~100 homes
Jul 2025TRI's Large Behavior Models paperThe field's first rigorous, blind, statistically-controlled evaluation — and a sobering one
Aug 2025Goldberg's "100,000 Year Data Gap"The scaling-skeptic case gets its canonical citation
Sep 2025Figure Series C, $39BPeak valuation for a pre-revenue humanoid company
Sep 2025Brooks: "Why Today's Humanoids Won't Learn Dexterity"The tactile-first critique gets its canonical citation
Nov 2025π*0.6 + RECAP; Generalist GEN-0RL-from-experience, and the first published embodied scaling-law claim
Jan 2026Skild AI, $1.4B at $14B+The "omni-bodied" thesis gets funded at scale
Feb 2026Intrinsic absorbed into GoogleAlphabet consolidates physical AI
Mar 2026Unitree STAR Market IPO filingFirst audited humanoid P&L
Apr 2026π0.7; Generalist GEN-1Compositional generalisation claims; 500k hours of human wearable pretraining with zero robot data
Jun 2026Agility SPAC ($2.5B); Neura Series C (up to $1.4B); Sanctuary pivots hardware-agnosticLate-cycle financing signals
Jul 2026Walden Robotics launches (Tedrake, $1.1B); Gemini Robotics 2; FCC adds mobile robots >2kg to the Covered ListGeopolitics arrives

1.2 The stack, from silicon to task

Understanding any lab or company in this book means locating it on this stack. Almost nobody spans more than three layers well.

  1. Actuators and transmissions — 40–60% of humanoid bill-of-materials cost per McKinsey. Quasi-direct-drive, harmonic reducers, tendon drives. Sangbae Kim's lab, Steve Collins' lab, LG Axium, Hyundai Mobis, Unitree.
  2. Hands and end effectors — parallel jaws, suction, multi-fingered, tendon-driven. 1X (25-DOF), Apptronik SharpaWave (22-DOF), Sanctuary hydraulics, RightHand hybrids.
  3. Sensing — cameras, depth, proprioception, and the missing modality, touch. GelSight/Adelson, Kennedy's DenseTact, Meta Digit 360, Amazon Vulcan.
  4. Low-level control — whole-body control, MPC, operational-space formulation, balance at 1 kHz. Khatib, Boston Dynamics, Biomimetic Robotics, Figure's S0.
  5. Policy — the learned visuomotor layer. Diffusion policies, flow matching, action chunking. Tedrake/TRI, Song, Finn, Physical Intelligence.
  6. Reasoning and task planning — VLMs, embodied reasoning, TAMP. Gemini Robotics ER, Figure's S2, MIT LIS, Jiajun Wu.
  7. Data and evaluation — collection interfaces, datasets, simulators, benchmarks. Everyone claims this layer; almost nobody has solved it.

1.3 Glossary of terms used throughout

  • VLA — Vision-Language-Action model. A network mapping images plus a language instruction to robot actions.
  • Action chunking — predicting a short sequence of future actions rather than one step, which reduces compounding error. Introduced at scale by ACT (ALOHA).
  • Diffusion policy — modelling the action distribution with a denoising diffusion or flow-matching process; now the default for multimodal manipulation behaviour.
  • Cross-embodiment — training one model on data from multiple different robot bodies.
  • Embodiment gap — the mismatch between the body that generated data (a human, another robot) and the body executing it.
  • Sim-to-real gap — the divergence between simulated and real physics, worst for contact, friction, and deformables.
  • Teleoperation — a human directly driving the robot, usually the source of demonstration data.
  • Perioperation — Improbable AI's term for collecting data with a passive exoskeleton mechanically coupled to the robot's hand, so a human's demonstration is already in the robot's action space.
  • Data flywheel — the claim that deployed robots generate the data that improves the next model, which improves deployment.
  • BOM — bill of materials; the hardware cost of a robot.
  • RaaS — robots-as-a-service; renting the robot per month rather than selling it.
  • MTBF — mean time between failures. The number almost nobody in humanoid robotics reports.
  • Whole-body control / loco-manipulation — coordinating locomotion and manipulation as one problem rather than two.