Log inSign up
Jason Ma
1,053 posts
@JasonMa2020

Jason Ma

@JasonMa2020
Co-founder @DynaRobotics Prev: @GoogleDeepMind, @NVIDIAAI, @Penn, @Harvard.
jasonma2016.github.io
Joined August 2018
999
Following
124.6K
Followers
RepliesRepliesRepostsRepostsMediaMedia

Log in or sign up for X

See what’s happening and join the conversation

Continue with phone
or
Log in with username or email
Terms·Privacy·Cookies·Accessibility·Ads Info·© 2026 X Corp.
  • @JasonMa2020
    Jason Ma
    @JasonMa2020
    Aug 27
    This is our release that I am actually most excited about because it shows and proves what truly matters. Deployment is the ultimate prize and the only reliable eval for robotics. After having worked on so many research projects and models in my career, where I saw so many new
    @DynaRobotics
    Dyna Robotics
    @DynaRobotics
    Aug 27
    After so many demos, models, pilots, robots are still struggling to land real deployments with real customers. Until now. Today we’re excited to share that our robots have successfully crossed the ROI threshold, and Din Tai Fung, one of the highest revenue per location
    Image
    00:00
    10
  • @JasonMa2020
    Jason Ma
    @JasonMa2020
    Aug 18
    I will be at #Actuate26 giving a talk on the model and infrastructure behind dyna-2. Excited to chat with everyone about scaling robot foundation models and deploying them in the real world! Please reach out if you'd like to chat!
    Image
    7
  • @JasonMa2020
    Jason Ma
    @JasonMa2020
    Aug 17
    At million-hour scale, the bottleneck is not the robots or the GPUs. It's everything between the robots and GPUs. Our cracked infra team did an amazing job scaling up and revamping our ML infra to enable the dyna-2 breakthroughs! Read our new infra blog to catch a glimpse of the
    @DynaRobotics
    Dyna Robotics
    @DynaRobotics
    Aug 17
    When people talk about robotics, they usually talk about models, data, or hardware. Few people talk about the infrastructure that lets you iterate on all three quickly. Today we're publishing how we trained Dyna-2 on over 1,000,000 hours of egocentric video, repeatably. At this
    Image
    00:00
    12
  • @JasonMa2020
    Jason Ma
    @JasonMa2020
    Aug 14
    predicting the future is key for unlocking generalization, great work by Jesse and co!
    @Jesse_Y_Zhang
    Jesse Zhang
    @Jesse_Y_Zhang
    Aug 12
    Robot WAMs typically predict future RGB latents trained for pixel reconstruction...but why stop there? We propose Flex-π, a WAM that jointly denoises multiple visual streams jointly, enabling the WAM to inherit strong 3D, object-centric, AND spatio-temporal priors for action
  • @JasonMa2020
    Jason Ma
    @JasonMa2020
    Aug 14
    this is no longer true.. and there is a lot nuances for getting good signal from training curves
    @VilleKuosmanen
    Ville🤖
    @VilleKuosmanen
    Aug 14
    keep this in mind when you see loss curves in robotics youtube.com/watch?v=RXd47U…
    Image
    2
Advertisement
Advertisement