aditya

i like sim & rl. @uwaterloo, prev @nvidia

[Palatial @PalatialSim]

Dexterity is the last frontier for robotics. Egocentric, dex-umi, and tele-op have so far failed to deliver the quality and scope of data required to solve manipulation. Simulations will.

For the past few months, we’ve been working with @origami_robo on the first Dexterity Factory, a real2sim2real pipeline that we believe will solve dexterous manipulation.

Objects go in, dexterous policies come out. By combining Palatial’s new real2sim capture system and Mad Max agent with Origami’s S1 Skill-capture gloves and robot hand, we’ve been able to close the sim2real gap on hundreds of complex tasks.

We want to contribute this research to the robotics community. Next week, we’ll open source the first 100 hand object interaction dataset, and we’re just getting started.

[aditya @_AdityaAjay]

@PalatialSim @origami_robo really cool!

[aditya @_AdityaAjay]

I strongly believe two important considerations for robotics startups in 2026 are:

1. does your output improve as frontier models become more capable?

2. do better frontier models improve your business case or put it at risk?

[aditya @_AdityaAjay]

@AnkaHeChen Excited for it!

[Dwarkesh Podcast @dwarkeshpodcast]

You can read a million books and listen to a million podcasts, but are you actually learning anything?

Do you integrate the knowledge and turn it into something useful? Or does it just fade away?

@michael_nielsen

[aditya @_AdityaAjay]

really cool demo, especially the coffee mixing with the water

[NVIDIA Asia Pacific @NVIDIAAP]

What does it take to simulate more than rigid bodies? 🤖

Explore open GPU physics, differentiable simulation, and multi-solver workflows for physical AI and robotics with NVIDIA researchers at #SIGGRAPHAsia2026.

Join this full-day workshop to see how these approaches connect simulation with real-world robot learning. 👉 https://t.co/ASCW6cnFy0

📅 Thursday, December 3 | 9:00–5:00 p.m. MYT

[aditya @_AdityaAjay]

@kevin_zakka @kevintracy48 simple and elegant!

[aditya @_AdityaAjay]

@qiayuanliao Incredible work!

[aditya @_AdityaAjay]

@BronarsToni really cool approach - super intuitive!

[aditya @_AdityaAjay]

@erwincoumans @voxagonlabs Really interesting! Using voxels is an inexpensive and reliable approximation. I wonder if use MPM can be used for more realistic dynamics for debris

[aditya @_AdityaAjay]

@CamronSabahi This is facts

[aditya @_AdityaAjay]

I find deformables very interesting!

[aditya @_AdityaAjay]

@Luckyballa for sure! let's do it

[aditya @_AdityaAjay]

@Luckyballa really cool!

[aditya @_AdityaAjay]

@BaoYanda I think the economic ceiling for data in a niche physical domain (like plumbing) will be much higher. purely because it is much harder to define criteria for success in those domains

[Lucky Iyinbor @Luckyballa]

Simulation will be a bitter lesson for human data folks 😁

[aditya @_AdityaAjay]

@octi_zhang @GoogleDeepMind really really impressive, Octi! makes me excited for how far RL can scale

[aditya @_AdityaAjay]

I’m a big fan of using RL for anything

[aditya @_AdityaAjay]

you can just scale ordinary PPO to make some really impressive policies

[aditya @_AdityaAjay]

@octi_zhang @patrickhyin @iggydagnino @mateoguaman @rosario_scaling really cool work, octi !!