Jason Ma
@JasonMa2020
Followers
6K
Following
7K
Media
88
Statuses
788
Co-founder @DynaRobotics Prev: @GoogleDeepMind, @NVIDIAAI, @MetaAI, @Penn, @Harvard.
Joined August 2018
We just did World’s first on-stage autonomous demo of long-horizon dexterous VLA 🚨 No training. No setup. Performance out of the box. Live demo is hard and unpredictable, but we felt great about our model’s generalization, and it went pretty well! 💯 Zero-shot. 100% success.
15
58
504
Agreed; it's my first time attending a developer conference instead of an academic conference. Really got me thinking deeper about deployment-in-the-loop model iteration. Also thanks for the photo feature :D
Actuate 2025 was a very cool developer conference for the new era of robotics, a short blog post:
1
5
43
Most companies chase no-code. These guys did the opposite. Instead of replacing engineers, @timgdelisle & @chriscrane, Co-Founder Fiveonefour, built tools that make devs 10x faster. Here’s what we discussed: • Why “developer-first” beats “no-code” • How simplifying code
14
3
25
yep our Dynasaur does way more than folding the field's moving fast but we’re focused on making robots actually useful - performant and reliable
Excited to share our latest progress on DYNA-1 pre-training! 🤖 The base model now can perform diverse, dexterous tasks (laundry folding, package sorting, …) without any post-training, even in unseen environments. This powerful base also allows extremely efficient fine-tuning
0
3
19
@DynaRobotics Amazing progress! Dexterity from pre-training is a huge green flag.
1
1
9
This is pretty incredible! @DynaRobotics continues to be at the forefront - true usefulness with complex manipulation tasks as opposed to fancy robot dancing demos!
We have been stress testing our model flywheel and seeing strong results in many challenging tasks! Task 1: chopping veggie🧑🍳 A test of coordinated tool use that involves asymmetric task and dynamic feedback to make consistent cuts. We used just 70 trajectories to get the
0
1
2
HUGE opportunity. RedBalloon is now hiring for these three high-level positions at The Daily Wire: 1. VP of Social Media 2. Producer 3. Counsel, Litigation & News If you’ve got the drive, discipline, and values to match then this is your shot. The Daily Wire isn’t
13
6
29
DYNA is leveling up — from folding to a wide range of dexterous tasks 👐All powered by one model, one set of weights. Everyone talks about generality, but few achieve it without post-training. DYNA actually scales — real generalization through scaling laws.
Excited to share our latest progress on DYNA-1 pre-training! 🤖 The base model now can perform diverse, dexterous tasks (laundry folding, package sorting, …) without any post-training, even in unseen environments. This powerful base also allows extremely efficient fine-tuning
1
4
24
We have been stress testing our model flywheel and seeing strong results in many challenging tasks! Task 1: chopping veggie🧑🍳 A test of coordinated tool use that involves asymmetric task and dynamic feedback to make consistent cuts. We used just 70 trajectories to get the
Excited to share our latest progress on DYNA-1 pre-training! 🤖 The base model now can perform diverse, dexterous tasks (laundry folding, package sorting, …) without any post-training, even in unseen environments. This powerful base also allows extremely efficient fine-tuning
2
10
122
Excited to share our latest progress on DYNA-1 pre-training! 🤖 The base model now can perform diverse, dexterous tasks (laundry folding, package sorting, …) without any post-training, even in unseen environments. This powerful base also allows extremely efficient fine-tuning
5
35
245
End-to-end architecture without external memory modules or counting logic. Unlocking more intelligence to production-ready embodied AI.
🎃 Halloween is coming. Our hardworking team is lining up for sweet treats, of course, served by Dynasaur! DYNA VLA model now has robust agentic reasoning capability, allowing it to serve arbitrary combinations and counts of candies! Pure imitation learning can’t work given the
6
3
47
.@JohnFetterman @SenFettermanPA @giselefetterman This is a deeply disturbing story. I was not aware of it. I don't think many Democrats are aware of it. But it's a story that must be told and re-investigated. Fetterman's story makes no sense IMO. https://t.co/fFOYYV0Opw
9
39
95
Another video of our VLA's robust reasoning capabilities against real-time disturbances:
1
0
13
We did a fun and timely halloween experiment benchmarking our VLA models' robust reasoning capabilities! 🎃 There's a lot of interest in reasoning for VLA models, but I personally felt most tasks the community benchmark on (1) do not require meaningful reasoning capabilities,
🎃 Halloween is coming. Our hardworking team is lining up for sweet treats, of course, served by Dynasaur! DYNA VLA model now has robust agentic reasoning capability, allowing it to serve arbitrary combinations and counts of candies! Pure imitation learning can’t work given the
1
8
102
🎃 Halloween is coming. Our hardworking team is lining up for sweet treats, of course, served by Dynasaur! DYNA VLA model now has robust agentic reasoning capability, allowing it to serve arbitrary combinations and counts of candies! Pure imitation learning can’t work given the
2
10
77
🎉Honored to be named to the 2025 AI #Disruptors60 list by @TechCrunch - and featured on the @Nasdaq tower! This list spotlights the most groundbreaking startups turning AI into real-world impact. At DYNA, our mission is to power the future of the physical economy with
3
6
52
The appearance of glass in this woman suggests a fragility we women know to be real. But Jesus can and does pick up the pieces and put us back together again.
1
2
8
It's really unfortunate to hear the FAIR layoff news.. Meta friends: if you are interested in working on frontier real-world robotics, @DynaRobotics is hiring! Please DM me!
2
19
97
Putting @OpenAI agent builder to the test against @bubblelab_ai Building simple sales analysis workflow
has anyone acc built a useful flow w OpenAI agent builder cuz I spent an hour on it and could not even get a relatively simple flow working….😳
5
1
21
built an open-sourced, agentic workflows (with full observability and exportability!!) check it out and pls dm me with feedback if you try it out!! https://t.co/3rOJNH3jGi
@bubblelab_ai
github.com
Open source workflow automation platform built for developers - full observability and code exportability! - bubblelabai/BubbleLab
0
3
6
POV: it’s Sunday afternoon and you STILL haven’t planned your @Techweek_ schedule 😭 Don’t panic I got you!!! Here’s a thread with 👇 🗓️ Full SF Tech Week event spreadsheet 💌 Personalized daily schedule to your inbox powered by @bubblelab_ai (free!)
7
4
37
College campuses are waking up. Whistleblowers are rising. I first saw this moment in a dream 10 years ago—months before the famous businessman from New York announced his run for President. Now the puzzle pieces make sense. Go to my page to see full prophetic word.
0
7
69
Very useful tool to plan your @Techweek_ schedule by @Selinaliyy, check it out!
POV: it’s Sunday afternoon and you STILL haven’t planned your @Techweek_ schedule 😭 Don’t panic I got you!!! Here’s a thread with 👇 🗓️ Full SF Tech Week event spreadsheet 💌 Personalized daily schedule to your inbox powered by @bubblelab_ai (free!)
0
0
5
Great to see people already using video rewinding (proposed in https://t.co/oE9qpYUeDV) to improve their reward models! This paper adds stage prediction to better reward long-horizon tasks. Excited to see how much better reward models can get!
🚀 Introducing SARM: Stage-Aware Reward Modeling for Long-Horizon Robot Manipulation Robots struggle with tasks like folding a crumpled T-shirt—long, contact-rich, and hard to label. We propose a scalable reward modeling framework to fix that. 1/n
1
13
131
@MacSchwager @pabbeel @YideShentu @philippswu @JasonMa2020's previous reward model works (VIP, LIV etc.) as well as @DynaRobotics 24-hour folding towel blog were definately inspirations to SARM. PLease checkout Dyna's new folding T-shirt demo at #CoRL2025
https://t.co/GCKFUZhpfT 6/n
We just did World’s first on-stage autonomous demo of long-horizon dexterous VLA 🚨 No training. No setup. Performance out of the box. Live demo is hard and unpredictable, but we felt great about our model’s generalization, and it went pretty well! 💯 Zero-shot. 100% success.
2
3
20