# What is true on Vision?

> Markdown twin of https://roboterminal.app/topics/vision · https://roboterminal.app/topics/vision.md

## On this layer

- org — [Gemini Robotics](https://roboterminal.app/companies/google) (7)
- org — World Labs (7)
- org — a16z (5)
- org — LimX Dynamics (2)
- org — 0xconglomerate (1)
- org — actuation cost (1)
- org — Aigen (1)
- org — Atlas (1)
- org — current humanoid BOM (1)
- org — dextrous hands cost (1)
- org — Embodied AI model (1)
- org — [Figure](https://roboterminal.app/companies/figure) (1)
- org — GalbotRobotics (1)
- org — [huggingface](https://roboterminal.app/companies/hugging-face) (1)
- org — ladder climb (1)
- org — LeggedRobot (1)
- org — LightOrigins (1)
- org — Linker Vision (1)
- org — [NVIDIA](https://roboterminal.app/companies/nvidia) (1)
- org — Perceptron AI (1)
- org — perceptroninc (1)
- org — PetoiCamp (1)
- org — pollenrobotics (1)
- org — rallies (1)
- org — ROBOTIS (1)
- org — sensing cost (1)
- org — structure cost (1)
- org — [Tesla](https://roboterminal.app/companies/tesla) (1)
- org — TRON 2 (1)
- org — [Unitree G1 humanoid robot](https://roboterminal.app/companies/unitree) (1)
- org — ZiNovaLabs (1)

## Who captures

- World Labs (5)
- Gemini Robotics (3)
- a16z (2)
- Tesla (1)
- @martin_casado (1)
- @ricci_nov (1)
- LimX Dynamics (1)
- ZiNovaLabs (1)
- Perceptron AI (1)
- 0xconglomerate (1)
- GalbotRobotics (1)
- Figure (1)
- @ParadisLabs (1)
- @gupta_abhinav_ (1)
- @spaceandtech_ (1)
- @RoboPapers (1)
- NVIDIA (1)

## Sourced numbers

- **TRON 2:** 1 units
  - _"TRON 2 serves as a modular and extensible embodied robotic platform for multi-tool, multi-step tasks across large workspaces."_
  - Source: [LimX Dynamics on X: "Together with ZINOVA's Tool Intelligence, TRON 2 takes on increasingly complex construction workflows.

@ZiNovaLabs builds on TRON 2 to explore an innovative robotic configuration for construction, demonstrating key tasks in a scaled-down tilt-up construction workflow, includi… / X](https://x.com/chris_j_paxton/status/2095114381818253768)
- **Atlas:** 1 units
  - _"Atlas is built to scale: its performance improves with increased training compute, and we expect this trend to hold as we continue scaling."_
  - Source: [Atlas: A World Model for Spatial Intelligence | World Labs](https://x.com/theworldlabs/status/2094839769616589051)
- **Embodied AI model:** 36,000,000,000 units
  - _"A 36B parameter model just blurred one of robotics’ biggest boundaries, the gap between seeing a scene and acting on it."_
  - Source: [@techniahqrobot on X](https://x.com/techniahqrobot/status/2094550793743708256)
- **@0xconglomerate 2094377490194288857:** $1million
  - _"$1 million"_
  - Source: [@0xconglomerate on X](https://x.com/0xconglomerate/status/2094377490194288857)
- **@0xconglomerate 2094377490194288857:** $649
  - _"$649"_
  - Source: [@0xconglomerate on X](https://x.com/0xconglomerate/status/2094377490194288857)
- **@0xconglomerate 2094377490194288857:** $500
  - _"$500"_
  - Source: [@0xconglomerate on X](https://x.com/0xconglomerate/status/2094377490194288857)
- **@0xconglomerate 2094377490194288857:** $319
  - _"$319"_
  - Source: [@0xconglomerate on X](https://x.com/0xconglomerate/status/2094377490194288857)
- **@0xconglomerate 2094377490194288857:** $399
  - _"$399"_
  - Source: [@0xconglomerate on X](https://x.com/0xconglomerate/status/2094377490194288857)
- **rallies:** 100 units
  - _"Galbot robots completed 100+ consecutive rallies at the World Humanoid Robot Games 2026 while tracking the ball, moving around the court and responding in real time."_
  - Source: [@ctorobotics on X](https://x.com/ctorobotics/status/2094107883365335067)
- **ladder climb:** 15 units
  - _"the company's humanoid climbing 15 feet up a ladder and 15 feet back down with no human input"_
  - Source: [@IntEngineering on X](https://x.com/IntEngineering/status/2093776043039416643)
- **Gemini Robotics 2 (VLA):** 1 units
  - _"three models: Gemini Robotics 2 (VLA) converts vision+language into motor control first DeepMind VLA to control a full humanoid feet-to-fingertips under one checkpoint."_
  - Source: [@csentropy on X](https://x.com/csentropy/status/2093431495277043724)
- **Gemini Robotics ER 2 (VLM):** 1 units
  - _"Gemini Robotics ER 2 (VLM) the embodied-reasoning brain; publicly available."_
  - Source: [@csentropy on X](https://x.com/csentropy/status/2093431495277043724)
- **Gemini Robotics On-Device 2 (VLA):** 1 units
  - _"Gemini Robotics On-Device 2 (VLA) runs locally, adapts to new embodiments in a few hours."_
  - Source: [@csentropy on X](https://x.com/csentropy/status/2093431495277043724)
- **Gemini Robotics ER:** 2 units
  - _"I believe Gemini Robotics ER 2 is a clear example here."_
  - Source: [@zhodonx on X](https://x.com/zhodonx/status/2092555239282843992)
- **current humanoid BOM:** ~$68k
  - _"Current humanoid BOM = ~$68k based on channel checks across 3 humanoid OEMs (UK and US). "_
  - Source: [@ParadisLabs on X](https://x.com/ParadisLabs/status/2092314785798828322)
- **actuation cost:** 33,800 units
  - _"- Actuation = 33.8% of current BOM"_
  - Source: [@ParadisLabs on X](https://x.com/ParadisLabs/status/2092314785798828322)
- **dextrous hands cost:** 30,300 units
  - _"- Dextrous hands = 30.3%"_
  - Source: [@ParadisLabs on X](https://x.com/ParadisLabs/status/2092314785798828322)
- **structure cost:** 7,600 units
  - _"- Structure = 7.6%"_
  - Source: [@ParadisLabs on X](https://x.com/ParadisLabs/status/2092314785798828322)
- **sensing cost:** 7,500 units
  - _"- Sensing = 7.5%"_
  - Source: [@ParadisLabs on X](https://x.com/ParadisLabs/status/2092314785798828322)
- **Unitree G1 humanoid robot:** 1 units
  - _"The Unitree G1 humanoid robot can drive a go-kart, controlling the steering wheel and pedals around an indoor track."_
  - Source: [@spaceandtech_ on X](https://x.com/spaceandtech_/status/2092234492266361018)
- **@RoboPapers 2091877173292356012:** 30000 hrs
  - _"30,000 hours"_
  - Source: [@RoboPapers on X](https://x.com/RoboPapers/status/2091877173292356012)

## Pulse

- [@TheHumanoidHub on X](https://x.com/TheHumanoidHub/status/2095976015818576359)
  Don't sleep on Tesla in humanoids just because they've gone quiet. Tesla bet on solving driving intelligence from vision alone, put a powerful inference computer in every car long before FSD was solve…
- [@a16z on X](https://x.com/a16z/status/2095972688200372368)
  >blogs about ray tracing for a CS project in 2012 >co-founds World Labs in 2024 >ships Atlas in 2026, a spatial intelligence model that turns a few photos into a whole 3D world curiosity comp…
- [@a16z on X](https://x.com/a16z/status/2095961376208646248)
  >be dr. fei-fei li
>born in beijing, raised in chengdu
>dad moves to new jersey, follow at age 16
>speak almost no english
>parents open a dry cleaner
>work the counter on weekends
>get into princeton…
- [@a16z on X](https://x.com/a16z/status/2095940012932215128)
  World Labs co-founder Justin Johnson says Atlas can recreate the famous "Bullet Time" shot from The Matrix with iPhones: "The way they did that shot is they had a ring of hundreds of cameras. [Neo] fe…
- [@drfeifei on X](https://x.com/drfeifei/status/2095926761305575826)
  Next view prediction is the key to Atlas, enabling us to unify pixel-level generation and reconstruction. @jcjohnss @BenMildenhall @martin_casado and I had a deeper discussion on some of the most exci…
- [@a16z on X](https://x.com/a16z/status/2095921217308086425)
  Old footage sitting in your camera roll could now be reconstructable as a 3D scene. World Labs co-founders Ben Mildenhall and Fei-Fei Li on how Atlas got there: Ben: "In a casual sense... I took three…
- [@martin_casado on X](https://x.com/martin_casado/status/2095905556242604308)
  Unbelievable 360 from a single image. When can we say NVS is solved? https://t.co/U9hbOsDcPN
- [@ricci_nov on X](https://x.com/ricci_nov/status/2095904540776346037)
  CPO is moving test upstream. Chroma says optical engines are now screened before final packaging: optical test is slow and downstream failure expensive. Suppliers are pushing detection to the wafer st…
- [@a16z on X](https://x.com/a16z/status/2095882301032828932)
  World Labs co-founders Fei-Fei Li, Justin Johnson, Ben Mildenhall, and a16z's Martin Casado on Atlas, a world model for spatial intelligence: LLMs are built on next token prediction. Video models are …
- [@LimX_Dynamics on X](https://x.com/LimX_Dynamics/status/2095758891892298066)
  One Navigation Brain, Two LimX Platforms: LightNav-0 × Oli / TRON 2 @LightOrigins’ open-source VLN model LightNav-0 was deployed zero-shot on LimX Dynamics’ Oli and TRON 2, enabling person following a…
- [LimX Dynamics on X: "Together with ZINOVA's Tool Intelligence, TRON 2 takes on increasingly complex construction workflows.

@ZiNovaLabs builds on TRON 2 to explore an innovative robotic configuration for construction, demonstrating key tasks in a scaled-down tilt-up construction workflow, includi… / X](https://x.com/chris_j_paxton/status/2095114381818253768)
  Post Log in Sign up Post LimX Dynamics on X: "Together with ZINOVA's Tool Intelligence, TRON 2 takes on increasingly complex construction workflows. @ZiNovaLabs builds on TRON 2 to explore an innovati…
- [Atlas: A World Model for Spatial Intelligence | World Labs](https://x.com/theworldlabs/status/2094839769616589051)
  September 1, 2026Introducing Atlas, our new omni world model for spatial intelligence. Atlas: A World Model for Spatial Intelligence World models generate, reconstruct, and simulate any possible world…
- [大仏ヨヨ on X: "Gemini Robotics ER 2を試してみたけど結構いいね。リアルタイムでOpenDuckMiniを追跡させてみたけどレスポンスは想定していたより悪くはなかった。どういうロボットかまで判別できているのが良かった。これ以外の機能もあるのでもう少し深堀しようと思う。
#OpenDuckMini,#microduckと同じ作者" / X](https://x.com/U6kcXA6yRN47345/status/2094603818340434044)
  Post Log in Sign up Post 大仏ヨヨ on X: "Gemini Robotics ER 2を試してみたけど結構いいね。リアルタイムでOpenDuckMiniを追跡させてみたけどレスポンスは想定していたより悪くはなかった。どういうロボットかまで判別できているのが良かった。これ以外の機能もあるのでもう少し深堀しようと思う。 \#OpenDuckMini,\#microduckと…
- [@techniahqrobot on X](https://x.com/techniahqrobot/status/2094550793743708256)
  A 36B parameter model just blurred one of robotics’ biggest boundaries, the gap between seeing a scene and acting on it. Perceptron AI has released Isaac 0.5, an open-weight Embodied AI model designed…
- [@0xconglomerate on X](https://x.com/0xconglomerate/status/2094377490194288857)
  $1 million in sales in under SEVEN hours, is the Microduck really that impressive? Here are other affordable open-source robots for context (+ my thoughts): ▸ @LeggedRobot Mini Pupper 2 (~$649): ROS2,…
- [@ctorobotics on X](https://x.com/ctorobotics/status/2094107883365335067)
  Humanoid robots just took tennis to another level. 🎾🤖 Galbot robots completed 100+ consecutive rallies at the World Humanoid Robot Games 2026 while tracking the ball, moving around the court and res…
- [@IntEngineering on X](https://x.com/IntEngineering/status/2093776043039416643)
  Figure founder Brett Adcock has posted video of the company's humanoid climbing 15 feet up a ladder and 15 feet back down with no human input, three weeks after its first autonomous ladder climb. Figu…
- [@csentropy on X](https://x.com/csentropy/status/2093431495277043724)
  Announced July 30, 2026 (author Carolina Parada) as part of Gemini Robotics 2, three models: Gemini Robotics 2 (VLA) converts vision+language into motor control first DeepMind VLA to control a full hu…
- [@zhodonx on X](https://x.com/zhodonx/status/2092555239282843992)
  I’ve said before that I think Google & DeepMind are playing a longer game with models. This is still a firm stance. Because a lot of the time, i think people look at Google’s model strategy too narrow…
- [@ParadisLabs on X](https://x.com/ParadisLabs/status/2092314785798828322)
  Humanoid BOM model: Current humanoid BOM = ~$68k based on channel checks across 3 humanoid OEMs (UK and US). - Actuation = 33.8% of current BOM - Dextrous hands = 30.3% - Structure = 7.6% - Sensing = …
- [@gupta_abhinav_ on X](https://x.com/gupta_abhinav_/status/2092309532500463836)
  @0fir0z @deepakpathak In the silicon valley world the race for robotics companies is to be similar to LLMs. But actually robotics needs many more concepts from computer vision. We are proud to say S1 …
- [@spaceandtech_ on X](https://x.com/spaceandtech_/status/2092234492266361018)
  The Unitree G1 humanoid robot can drive a go-kart, controlling the steering wheel and pedals around an indoor track. It can handle tight corners and even perform controlled drifts. Developed with Symb…
- [@RoboPapers on X](https://x.com/RoboPapers/status/2091877173292356012)
  Instead of choosing between training a world model and training a language conditioned robot policy, why not do both? LDA-1B s a new foundation model that is trained on 30,000 hours of human and robot…
- [Cosmos 3 Post-Training in Action With Aigen and Linker Vision | Cosmos Labs - Live Broadcast by NVIDIA Robotics / X](https://x.com/NVIDIARobotics/status/2090469297667920292)
  \\ \\ REPLAY NVIDIA Robotics @NVIDIARobotics Cosmos 3 Post-Training in Action With Aigen and Linker Vision \| Cosmos Labs Scan to get the app

---

Source: Roboterminal · https://roboterminal.app/topics/vision.md