Robots Digest 🤖's banner
Robots Digest 🤖's profile picture

Robots Digest 🤖

@robotsdigest • 4,897 subscribers

Follow @RobotsDigest for latest in Robotics, Humanoids, and Hardware + AI.

Videos

LIVE
1.2k

Anya Rossi

sweetdream.ai

SweetDream.aiSponsored•Livecam
Streaming Now

Watch Anya Live

Anya is streaming live right now! Join her private show and enjoy exclusive content.

HD live stream
Exclusive private shows
1.2k viewers online

Free preview available • Premium content

Physical intelligence needs touch, not just vision. eFlesh is a fully 3D-printable magnetic tactile sensor that enables reliable slip detection and contact awareness on real robots, using low-cost hardware and scalable fabrication
1:00
robotsdigest's profile picture

Physical intelligence needs touch, not just vision. eFlesh is a fully 3D-printable magnetic tactile sensor that enables reliable slip detection and contact awareness on real robots, using low-cost hardware and scalable fabrication

Robots Digest 🤖

243,286 views • 4 months ago

Qwen-VLA feels like one of the first real robotics foundation models. A single system trained across robot manipulation, navigation, egocentric human video, simulation, and vision-language reasoning instead of isolated robot policies.
2:52
robotsdigest's profile picture

Qwen-VLA feels like one of the first real robotics foundation models. A single system trained across robot manipulation, navigation, egocentric human video, simulation, and vision-language reasoning instead of isolated robot policies.

Robots Digest 🤖

14,422 views • 10 days ago

Realtime-VLA FLASH tackles one of the biggest deployment bottlenecks for diffusion-based VLAs: inference latency. The key idea is speculative inference for flow-matching VLAs. A lightweight draft model predicts an action chunk, while the main model’s Action Expert verifies it in parallel using flow-consistency checks instead of running full denoising every replanning round. This lets the system replace many expensive 58 ms full inference rounds with speculative rounds as fast as 7.8 ms, reducing average latency to 19.1 ms and achieving a 3.04× speedup on LIBERO while largely preserving success rate. Interesting systems insight: they profile π0 and show VLM prefill is compute-bound, while Action Denoise is memory-bound. FLASH exploits this by reusing KV cache and parallelizing verification instead of repeatedly running sequential denoising.
1:54
robotsdigest's profile picture

Realtime-VLA FLASH tackles one of the biggest deployment bottlenecks for diffusion-based VLAs: inference latency. The key idea is speculative inference for flow-matching VLAs. A lightweight draft model predicts an action chunk, while the main model’s Action Expert verifies it in parallel using flow-consistency checks instead of running full denoising every replanning round. This lets the system replace many expensive 58 ms full inference rounds with speculative rounds as fast as 7.8 ms, reducing average latency to 19.1 ms and achieving a 3.04× speedup on LIBERO while largely preserving success rate. Interesting systems insight: they profile π0 and show VLM prefill is compute-bound, while Action Denoise is memory-bound. FLASH exploits this by reusing KV cache and parallelizing verification instead of repeatedly running sequential denoising.

Robots Digest 🤖

14,066 views • 25 days ago

Robots forget because vision is expensive. AstraNav-Memory shows you can compress vision 20x and still remember hundreds of past frames, unlocking true lifelong navigation.
0:45
robotsdigest's profile picture

Robots forget because vision is expensive. AstraNav-Memory shows you can compress vision 20x and still remember hundreds of past frames, unlocking true lifelong navigation.

Robots Digest 🤖

26,667 views • 5 months ago

Cosmos Policy turns a pretrained video diffusion model into a robot controller. Instead of redesigning the architecture, it injects robot state, actions, and values directly as latent frames inside the video model
1:15
robotsdigest's profile picture

Cosmos Policy turns a pretrained video diffusion model into a robot controller. Instead of redesigning the architecture, it injects robot state, actions, and values directly as latent frames inside the video model

Robots Digest 🤖

22,933 views • 4 months ago

Are all robots equally UMI-able? New research from CMU Robotics Institute + Stanford AI Lab : UMI-on-Air proves that they can be .
0:20
robotsdigest's profile picture

Are all robots equally UMI-able? New research from CMU Robotics Institute + Stanford AI Lab : UMI-on-Air proves that they can be .

Robots Digest 🤖

31,330 views • 7 months ago

Boston Dynamics Atlas Robot Powered by AI "Large Behavior Models" Boston Dynamics and Toyota just showed Atlas doing something wild, packing boxes using a Large Behavior Model. One AI brain controls walking, crouching, lifting, everything. Just learned from human demos.
2:24
robotsdigest's profile picture

Boston Dynamics Atlas Robot Powered by AI "Large Behavior Models" Boston Dynamics and Toyota just showed Atlas doing something wild, packing boxes using a Large Behavior Model. One AI brain controls walking, crouching, lifting, everything. Just learned from human demos.

Robots Digest 🤖

25,305 views • 6 months ago

Touch was the missing sense in robotics But FlexiTac is here to change that. An open-source, low-cost, scalable tactile sensor that’s: • ~3 min to fabricate • ~$2.5 per unit • Real-time + ML-ready This is what making hardware actually accessible looks like.
3:36
robotsdigest's profile picture

Touch was the missing sense in robotics But FlexiTac is here to change that. An open-source, low-cost, scalable tactile sensor that’s: • ~3 min to fabricate • ~$2.5 per unit • Real-time + ML-ready This is what making hardware actually accessible looks like.

Robots Digest 🤖

12,744 views • 2 months ago

Everyone is scaling VLAs with more robot data. TiPToP shows another path. No robot training, no policy learning. Just RGB + language → 3D scene → GPU TAMP planner → trajectory. Foundation models + planning alone can run real manipulation tasks.
1:38
robotsdigest's profile picture

Everyone is scaling VLAs with more robot data. TiPToP shows another path. No robot training, no policy learning. Just RGB + language → 3D scene → GPU TAMP planner → trajectory. Foundation models + planning alone can run real manipulation tasks.

Robots Digest 🤖

10,362 views • 2 months ago

Robot learning is starved of data not only because it is hard, but because it is outright BORING! RoboCade: “what if we made it feel like Clash Royale instead of manual labor?”
3:13
robotsdigest's profile picture

Robot learning is starved of data not only because it is hard, but because it is outright BORING! RoboCade: “what if we made it feel like Clash Royale instead of manual labor?”

Robots Digest 🤖

10,835 views • 5 months ago

No more content to load