Markov Robotics (@markovrobotics) header
Markov Robotics (@markovrobotics) profile picture

Markov Robotics

@markovrobotics

0-shot dexterity

mr.inc
Joined December 2025
  • Tweets21
  • Following4
  • Followers310
7Media
Photo by Markov Robotics (@markovrobotics)Photo by Markov Robotics (@markovrobotics)Photo by Markov Robotics (@markovrobotics)
Markov Robotics retweeted
Josh Kaplowitz (@jjkaplowitz) profile picture
Josh Kaplowitz
@jjkaplowitzNov 6
Absolutely overwhelming interest in my upcoming SF Bay Area event. We鈥檙e still looking at people but about to start waitlisting we are just about at capacity! luma.com/ak5cch9g Thank you to our sponsors @jase (ready.net), @FoundryRobotics, @sintersys and @markovrobotics for making this happen!
Image from Josh Kaplowitz (@jjkaplowitz)
1
1
10
622
Great writeup from @modal on how we鈥檙e building general-purpose robotic dexterity. Being able to scale from zero to hundreds of GPUs on demand has made a huge difference in how quickly we can iterate on our world models. Appreciate the Modal team!
Modal (@modal) profile pictureModal
@modalOct 26
.@markovrobotics is teaching robots to tackle complex manipulation tasks with fewer than ten hours of robot data. Their bet: video already contains most of the physical reality a robot needs to act. They use Modal for video processing, world model training, and remote inference.
1
2
11
595
Hand tracking through full occlusions!
atharva (@_gundawar) profile pictureatharva
@_gundawarOct 18
鈥o world models have x-ray vision?
7
918
shake all you want!
1
1
11
3,684
Markov Robotics retweeted
atharva (@_gundawar) profile picture
atharva
@_gundawarOct 4
Astra is good at controlling the fx3 and Halo 300x, gets you decent recordings
Image from atharva (@_gundawar)
atharva (@_gundawar) profile pictureatharva
@_gundawarOct 1
馃憖
1
14
1,456
Markov Robotics retweeted
Greg Tarr (@gregtarr) profile picture
Greg Tarr
@gregtarrOct 1
Oh no
7
5
1
35
10,290
Even finger protectors can鈥檛 stop our dexterity!
1
2
1
9
1,082
Sub-millimeter object pick and place!
atharva (@_gundawar) profile pictureatharva
@_gundawarSep 1
World models are very precise on out-of-distribution tasks! Here鈥檚 an example of non-compliant arms running at low Hz, picking up a fake $100 bill. The policy was never trained on this or any remotely similar objects Zero-shot generalization is the only way to physical AGI
1
2
12
6,058
Markov Robotics retweeted
atharva (@_gundawar) profile picture
atharva
@_gundawarAug 22
We made this game to hire cracked teleoperators (upto $100k and equity) Play here: telebench.vercel.app/ Here is our friend Charis's run:
2
3
2
63
58,589
Fast continuous motion learned efficiently
atharva (@_gundawar) profile pictureatharva
@_gundawarAug 17
Real time chunking world models: 1x, autonomous, 200ms per generation!!
10
3,138
Markov Robotics retweeted
atharva (@_gundawar) profile picture
atharva
@_gundawarAug 12
world model powered toaster!
atharva (@_gundawar) profile pictureatharva
@_gundawarAug 11
super excited to partner with @ltx_io for a Day 0 launch with them!
8
11
1
87
18,824
LTX-2.5 is changing how we train robots. Years of hand-built training data, replaced by a world model that already understands physical space. Trained on LTX. @ltx_io
6
6
4
58
683,508
Markov Robotics retweeted
atharva (@_gundawar) profile picture
atharva
@_gundawarAug 8
$50,000 hand hanging off a $3,000 arm bolted to a $100 stand
Image from atharva (@_gundawar)
7
5
2
125
9,068
Markov Robotics retweeted
atharva (@_gundawar) profile picture
atharva
@_gundawarAug 3
sneakpeak
Image from atharva (@_gundawar)
1
1
15
1,900
Markov Robotics retweeted
atharva (@_gundawar) profile picture
atharva
@_gundawarJul 31
this means we can handle fragile things
atharva (@_gundawar) profile pictureatharva
@_gundawarJul 29
modeling force using video diffusion
6
10
3
71
23,722
Markov Robotics retweeted
atharva (@_gundawar) profile picture
atharva
@_gundawarJul 29
modeling force using video diffusion
1
3
1
28
13,970
Markov Robotics retweeted
atharva (@_gundawar) profile picture
atharva
@_gundawarJul 25
new embodiment unlocked
Image from atharva (@_gundawar)
1
1
11
535
0-shot pick and place on unseen objects with less than an hour of teleoperation data
1
3
1
20
3,604
Toward general dexterity We are training robot policies that learn a rich and coherent physical representation of the world by conditioning on multimodal observations Here鈥檚 a small glimpse of what we鈥檝e been building: Task: Pick the ramen cup and place it in the box Given the current world state, the model generates a multimodal trajectory; here we show the decoded video and the corresponding actions executed on the humanoid
2
6
3
27
3,119
Load more