I know it feels like a really competitive market, but Physical AI is not a zero-sum game.

@AppliedInt CTO Peter Ludwig looked out at 200 founders at Dirty Jobs this year and said every one of them can win, because the markets we build in are really THAT big.

He explained why on in interest across autonomy "Reinforcement Learning only got fast enough to matter in the last 12 to 18 months"

The time is now. EP 104 of CLIMB 👇

[New Blog] How does labeled data become a model back on the road? 🔄 At @AppliedInt, our physical AI data flywheel runs over hundreds of petabytes. Here's how: appliedintuition.com/engineering-bl…

🎤 At @lancedb's #Reverie2026 on Nov. 5th? Catch my talk on the main stage, 11:30 AM PT.

How does labeled data become a model back on the road?

@_xushiyan breaks down each stage: curation, scenarios, data loading, training, and evaluation, across hundreds of petabytes of multimodal data.

appliedintuition.com/engineering-bl…

Catch him on the main stage at LanceDB's at 11:30 AM PT.

Can we achieve state-of-the-art for mainstream E2E and vectorized driving benchmarks without using any human demonstration data?

Terra series at @AppliedInt now tops the NAVSIM, HUGSim, nuPlan val14 and InterPlan benchmarks, powered by self-play RL without imitation, scaling in ents being reactive.
🚘 With TerraZero released a few months ago, we set the SOTA on vectorized benchmarks like nuPlan val14 and InterPlan, powered by our super high-throughput self-play RL framework experiencing 25 centuries of challenging driving scenarios.
📽️ In June this year, TerraTransfer was released as the first of its kind to bring the SOTA vectorized policy from self play (TerraZero) to an E2E policy via image-to-vector encoder alignment, which topped the HUGSim closed-loop score.
🔚 This week, we are thrilled to release TerraTransfer++, a demonstration-free E2E planner setting the SOTA for NAVSIM and HUGSim, by adding planning capability from self-play rollout experience to TerraTransfer.

Setting SOTA for public benchmarks is just a beginning, as we have seen the journey from pioneering work like UniAD to successful mass production of E2E ADAS in the industry, but what makes me quite excited is to see the strong signal from a new paradigm with potential to refactorize how the next generation autonomy can be developed, and bring the scalability, safety and performance to the next level.

We are the believer of data and scaling, but not just human demonstration data and open-loop scaling like imitation or open-loop RL (AV2.0) - we have the bet on closed-loop scaling with both synthetic and real-world data (AV3.0), and the Terra series is the starting point of the journey towards real product with our bet.

Great work from the team @caffe_tt, Weixin Li, Zhouchonghao Wu, @AkshayRangesh, Saarth Bonde, Grantland Hall, @ChenTangMark, Yihan Hu, for creating TerraTransfer and TerraTransfer++!

TerraTransfer++ report:  arxiv.org/abs/2610.30319
TerraTransfer++ website: terra-applied.github.io/TerraTransferP…
TerraTransfer report: arxiv.org/abs/2606.17386

E2E driving is bottlenecked by data. Imitation needs fleet-scale logs that barely cover the safety-critical long tail. Closed-loop RL in photoreal sim pays per-step rendering and backbone cost.

TerraTransfer++ removes human demonstrations from every stage and achieves g performance across open-loop and closed-loop benchmarks. 🧵

Paper: arxiv.org/abs/2610.30319
Website: terra-applied.github.io/TerraTransferP…

Then we teach it to see. Adapters on a frozen DINOv3 backbone align camera features to the frozen teacher’s latent space.

Every target traces back to self-play, so training needs only paired (image, scene-state) frames. No expert trajectories, no online rendering.

Our policies set the new SOTAs of navtest, navhard, and HUGSim.

Applied Intuition has been awarded a contract by the @USNavy to scale Data Engine across its fleet.

As the infrastructure layer for maritime autonomy, Data Engine collects operational data through passive navigation capabilities installed on Navy ships.

Here’s how the loop from ships at sea
🔵 Curate it and train perception models
🔵 Evaluate the models
🔵 Deploy them back to vessels over the air

The result is what one ship learns now reaches the entire fleet – a threat encountered by a single vessel is turned into an improved model and deployed over the air, so every ship can detect it.

Our goal is for each military service to run its own Data Engine, tailored to its domain and feeding into a common infrastructure.

Learn more about our work with the U.S. Navy here:
appliedintuition.com/blog/navy-data…

Qasar Younis (Applied Intuition): “The fantasy that these machines that exist around us are going to be replaced by new machines just doesn’t exist”

Live from Day 2 of #LongHorizon #PhysicalAI

Where does physical AI go from here? @qasar shares his takes with @eostudi0.

Watch the full episode: youtu.be/REsrgNeuZkU

Everyone’s been talking about Mission S vs flying Roadsters and the physical part of their design BUT the human-machine interaction matters just as much, if not more.

Like you see and interact with your car from the inside most of the time. So then why are most in cabin experiences far from magical, seamless AI driven interactions? Because historically there wasn’t a platform to build on

When software shapes the car, what’s left for design?

The human. See how we’re building the next generation of intelligent cars with @Stellantis.

appliedintuition.com/blog/stellanti…

Simulation is still our bread and butter and it’s only becoming clearer that it’ll unlock robotics at scale

Reinforcement Learning (RL) in simulation is a scalable strategy to learn dexterous skills, but naive exploration from scratch often fails to discover manipulation strategies for diverse objects.

At @AppliedInt, we propose X-Reset, a straightforward method to resolve the exploration problem for generalist RL policies by randomly sampling resets from human hand-object states.
- We train a single dexterous policy to manipulate a variety of objects to any goal pose, ranging from small markers to heavy powerdrills with zero teleoperation data
- The learned policy zero-shot generalizes well beyond the training set (DexYCB), and transfers from sim-to-real using object pose-tracking models

We are continuously expanding robotics research at @AppliedInt beyond X-Reset, with the bounds of simulation and synthetic data for dexterous manipulation pushed. Stay tuned and join us!

🌐 Project Page: xreset-applied.github.io
📄 Paper: arxiv.org/abs/2609.35715

Dexterous manipulation in simulation has an exploration problem.

Here is how our team is fixing this problem.

“I don’t believe you can be very successful and not read a lot.”

4 books @qasar recommends, the ones that helped him build @AppliedInt, a $15B physical AI company:

📍I Am That, Sri Nisargadatta Maharaj
📍Sam Walton: Made in America, Sam Walton (with John Huey)
📍The Story of My Mahatma Gandhi
📍Long Walk to Freedom, Nelson Mandela

You join. You ship. You find what’s broken and fix it.
That’s the bar.

review.firstround.com/firsthand-appl…

Come build alongside some of the best engineers in the industry, solving the hardest problems in physical AI.

appliedintuition.com/careers?utm_so…