What are the key takeaways from this The TWIML AI Podcast episode?

Key insights include: **Sim-to-Real Gap:** Closing the simulation-to-reality gap requires deep understanding of both worlds, mapping every software layer from high-level commands down to motor currents. Real-to-sim processes identify critical parameters like torque limits and delays by hanging robots and collecting motor data, then abstracting these into fast simulators.; **Locomotion Progress:** Blind locomotion (reactive only) is robust for quadrupeds, but perception-based locomotion remains challenging. Adding cameras increases the sim-to-real gap significantly because sensor noise and visual rendering must be accurately simulated. Switching robots to new hardware takes only days, but adapting to new tasks requires substantial engineering effort.; **Robot Deployment Reality:** Most robotics demos use either live teleoperation or policies trained on thousands of hours of collected teleoperation data. No humanoid robot currently generates economic value in warehouses or factories because they perform approximate tasks requiring more human handlers than the original workforce, making value negative.

How long is this episode of The TWIML AI Podcast?

This episode is 66 minutes long. SignalCast provides an AI-generated summary so you can get the key insights in about 3 minutes.

The TWIML AI Podcast

Intelligent Robots in 2026: Are We There Yet? with Nikita Rudin - #760

January 8, 2026

66 min episode · 2 min read

Episode

66 min

Read time

2 min

Topics

Productivity, Leadership, Software Development

AI-Generated Summary

Published Jan 9, 2026

Key Takeaways

✓Sim-to-Real Gap: Closing the simulation-to-reality gap requires deep understanding of both worlds, mapping every software layer from high-level commands down to motor currents. Real-to-sim processes identify critical parameters like torque limits and delays by hanging robots and collecting motor data, then abstracting these into fast simulators.
✓Locomotion Progress: Blind locomotion (reactive only) is robust for quadrupeds, but perception-based locomotion remains challenging. Adding cameras increases the sim-to-real gap significantly because sensor noise and visual rendering must be accurately simulated. Switching robots to new hardware takes only days, but adapting to new tasks requires substantial engineering effort.
✓Robot Deployment Reality: Most robotics demos use either live teleoperation or policies trained on thousands of hours of collected teleoperation data. No humanoid robot currently generates economic value in warehouses or factories because they perform approximate tasks requiring more human handlers than the original workforce, making value negative.
✓Model Architecture: Production systems use three-tier hierarchies: large VLMs for abstract task planning (running off-board or in cloud), vision-language-action models for motion planning at ten hertz (onboard, compute-limited), and small whole-body trackers at fifty hertz on CPU for motor control. Diffusion models in VLAs create the biggest onboard compute bottleneck.
✓Training Approach: Combining reinforcement learning with imitation learning works by using human demonstrations to guide RL exploration, not as pure pretraining. Training separate primitive skills (walking, picking, placing) then distilling into one VLA shows early generalization signs. Industrial deployment focuses on repetitive tasks where abstract reasoning can be precomputed, unlike home environments.

What It Covers

Nikita Rudin, CEO of Flexion Robotics, explains the gap between robotics demos and real-world deployment, covering simulation-to-reality challenges, reinforcement learning techniques, and why no humanoid robot generates actual economic value today in 2025.

Key Questions Answered

•Sim-to-Real Gap: Closing the simulation-to-reality gap requires deep understanding of both worlds, mapping every software layer from high-level commands down to motor currents. Real-to-sim processes identify critical parameters like torque limits and delays by hanging robots and collecting motor data, then abstracting these into fast simulators.
•Locomotion Progress: Blind locomotion (reactive only) is robust for quadrupeds, but perception-based locomotion remains challenging. Adding cameras increases the sim-to-real gap significantly because sensor noise and visual rendering must be accurately simulated. Switching robots to new hardware takes only days, but adapting to new tasks requires substantial engineering effort.
•Robot Deployment Reality: Most robotics demos use either live teleoperation or policies trained on thousands of hours of collected teleoperation data. No humanoid robot currently generates economic value in warehouses or factories because they perform approximate tasks requiring more human handlers than the original workforce, making value negative.
•Model Architecture: Production systems use three-tier hierarchies: large VLMs for abstract task planning (running off-board or in cloud), vision-language-action models for motion planning at ten hertz (onboard, compute-limited), and small whole-body trackers at fifty hertz on CPU for motor control. Diffusion models in VLAs create the biggest onboard compute bottleneck.
•Training Approach: Combining reinforcement learning with imitation learning works by using human demonstrations to guide RL exploration, not as pure pretraining. Training separate primitive skills (walking, picking, placing) then distilling into one VLA shows early generalization signs. Industrial deployment focuses on repetitive tasks where abstract reasoning can be precomputed, unlike home environments.

Notable Moment

Rudin demonstrates live on-stage training where a quadruped robot learns to walk from scratch in three minutes on a laptop, with policies updating every fifteen seconds. The robot progresses from falling over to taking steps to walking around the stage, visually showing the reinforcement learning process.

Know someone who'd find this useful?

Episode Transcript

I'd like to thank our friends at Capital One for sponsoring today's episode. Capital One's tech team isn't just talking about multi agentic AI. They already deployed one. It's called Chat Concierge, and it's simplifying car shopping. Using self reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love. It helps schedule a test drive, get preapproved for financing, and estimate trade in value. Advanced, intuitive, and deployed. That's how they stack. That's technology at Capital One. My hot take on that, and I'll be happy to be proven wrong, but I think there is not a single humanoid robot today that actually generates value. Meaning, there might be a robot that does something fairly close to what it's supposed to do in a factory or in a warehouse, but it's not the exact task. In the end, it's not generating value because it's not doing the actual thing, I suppose. Alright, everyone. Welcome to another episode of the TwiML AI podcast. I am your host, Sam Cherrington. Today, I'm joined by Nikita Rudin. Nikita is cofounder and CEO of Flexion Robotics. Before we get going, be sure to take a moment to hit that subscribe button wherever you're listening to today's show. Nikkita, welcome to the podcast. Thank you. We're excited to be here. I'm excited to have you on the show, and I'm looking forward to digging into our topic for the conversation, which is, really digging into the gap between, you know, where we are today with robotics and where we need to be to fulfill the vision of the technology. You've been working in this space for quite a while. You did your PhD at ETH Zurich and spent some time at NVIDIA. Why don't you share a little bit about your PhD and the focus of your research? So when I started, we were trying to use simulation with reinforcement learning to teach a legged robot very simple things, like just walking on a on on flat ground. And when the robot could take a few steps, that was already a big success. And the core focus was to reduce the training time needed to to achieve that. And when you say legged, like a quadruped? Like like a quadruped, a big legged, robot dog. We're not using bosonamic spots. We're using antibiotics animal. Antibiotics is a is a Swiss startup that was a spin off from from our lab. Very very similar to to a spot, but it's red and made made in Switzerland. Yeah. We're really trying to reduce the training time needed to to achieve that. So be before I started, there were some results of reinforcement learning for for such quadrupeds, but it would take weeks of computation to it to achieve anything. And using GPUs and massively parallel simulators, we managed to reduce that to just a few minutes. It's actually we had a demo on stage at some point at at …

Get the full transcript (11,344 words) + summary by email — free

One-time email with the complete transcript and AI summary of this episode. No account needed.

One email, no spam. We’ll also show you what SignalCast does.

Browse all The TWIML AI Podcast transcripts →