Excited to announce that we have raised $120M in our Series A to advance the frontier of general-purpose high-performance robots. 🤖 The new funding will accelerate progress towards our mission of bringing foundation-model powered robots to everyone, everywhere. Read more 👇

At our Robotics Day in SF, we sat down with @ArmenAgha,@JasonMa2020, and @philippswu, who shared 8 sharp takes on what the field is missing: → The 'data pyramid' weighs quality vs. scale → Data quality isn't static → Build in-house where the frontier is unsolved → RL works best when robots can match humans → Pixel-level world models may be a dead end → Physics understanding is being neglected → Data attribution is "dark magic” → Know where open source fits in your training stack Learn more from the full conversation 👉 bessemervp.team/4dFdgos

Robotics is moving from impressive demos to scalable learning systems. 🤖DYNA’s Cofounder & Chief Scientist @JasonMa2020 will be speaking at the #CVPR workshop on Scalable Robot Learning Systems. 🙌If you’re working on embodied AI, data engines, VLM/VLA, sim2real, MLsys, or eval, come join the conversation! We will share how we build robots that don’t just perform one task — but keep learning, adapting, and scaling. scalebot-workshop.github.io/

Everyone at Red Bull Mirage brought a DJ. We brought a robot. The robot worked out of the box, no additional training onsite. 8 hours. 700 Red Bulls served. 99% success rate. No humans, no breaks, no resets. Shifting desert light. Cans handed in at every angle. Thousands of people are moving around it. The robot just kept serving. 🤖🥫

A must read. True lesson. Honest hot takes from @YorkYang5050 , a second time founder and cofounder of @DynaRobotics

There is a massive misconception in the AI and robotics space right now. People think that once a model is "smart" enough in the lab, deploying it into the real world is the easy part. It’s not. The real world is chaotic, unpredictable, and entirely unforgiving. At Dyna Robotics, we are taking breakthrough embodied AI out of the lab and putting it into commercial environments, today already. To do that successfully, we realized we had to completely rethink what a Forward Deployment Engineer actually does. If you look around the industry, most "forward deployment" roles are essentially field operations or customer integration. You're handed a model and told to make it work on-site. That is NOT what this role is. I’m looking for a hardcore engineer who can straddle the line between our AI research and our systems architecture. I need someone who wants to build the connective tissue that makes general-purpose robots a reality at scale. Rather than just acting as a consumer of our models, you’ll be actively participating in applied research—building out model evaluation infrastructure, ensuring stability, and solving complex edge cases. At the same time, you'll be architecting the end-to-end infrastructure that bridges off-board computing operations all the way down to the application-level software running natively on the robots. We don't need someone to just string APIs together. We need someone with a deep, first-principles understanding of computer science who is excited to build the data pipelines, tooling, and fleet management systems that allow our robots to operate autonomously in the wild. If you want to push beyond traditional software boundaries and be the critical bridge between state-of-the-art AI research and the physical world, I want to talk to you. Link to apply is below, or feel free to DM me directly. My DMs are open. jobs.ashbyhq.com/dyna-roboti…

Proud to share Dyna’s first customer story. Monster Laundry in Sacramento is running Dyna's fully autonomous robots in real production. Over the past 3 months since it was deployed, DYNA robot has folded over 200,000 towels from 10 different customers, saving hundreds of hours of attendant time. Our customer calls it "Sophy Swiftfold". 👉See how we bring research to reality: dyna.co/blog/monster-laundry

DYNA's Cofounder & Chief Scientist @JasonMa2020 will be speaking at the @HumanoidsSummit on Dec 11 &12! Jason will break down how @DynaRobotics is building robust robot foundation models for real-world manipulation and what it takes to scale.🦾 From research labs to production deployment—don't miss this! Come say hi if you are around!  #HumanoidsSummit #Robot #AI

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

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 results you see in the video. Task 2: cup stacking 🎉 A test of high-precision control, which requires precise and delicate positioning at every step. Mistakes at any step are catastrophic! The arms we use have high control error, but the model makes up for it. These are quite distinct tasks that are difficult in different ways than some of our earlier demos like laundry/napkin folding. Very glad to see by injecting dextereous control into pre-training in an integrated way, we see substantial boost in post-training robustness and efficiency!

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 to ~100% success on challenging new tasks with as little as 1 hour of data! 🤯 Watch it master two of them: cup stacking & celery chopping on repeat, no failures. 👇

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, or (2) are somewhat unrealistic and do not represent tasks in real-world scenarios. So we decided to use object counting and manipulation as a real benchmark; it's quite common and realistic, but I haven't seen much work in this area. End-to-end Imitation learning would fail because of combinatorially many permutations you can ask to the robot. Our VLA model can count and follow language commands fairly robustly -- all in an end-to-end architecture without external memory modules or counting logic. The model also robustly handles external disturbances to the scene (like shuffling the candy baskets). It's a small cute experiment we did to benchmark reasoning, but it's pretty fun so thought we'd share!