Building the World's First Visual Memory for Robots? | Shawn Shen

We sit down with Sean Shen (CEO & founder of Memories AI) to discuss how physical AI learns from video data. Learn the process of turning human observation into training data for robotics.Memories AI specializes in capturing human movement through video to teach robots how to perform real-world tasks. Sean explains the technical requirements for robots to successfully learn from watching us, moving beyond traditional programming to more intuitive training methods.This episode covers the current state of data collection for embodied systems and the specific challenges of scaling physical AI. We examine the workflows used to bridge the gap between video footage and operational robot behavior, providing a look at how Memories AI is building the foundation for more capable machines.Subscribe for weekly robotics breakdowns, and let us know your thoughts on the future of robot learning in the comments.Guest https://x.com/shawnshenjxLéo https://x.com/LeoKharonOskar https://x.com/hundreddaysxPodcast https://x.com/roboticsstack0:00 Podcast Overview 0:50 The Vision behind Memories AI 3:30 Scaling Law4:28 Leaving Meta & Leveraging AI 5:40 Company Structure 6:40 Team & Hiring 7:33 Data Quality 8:58 Good Data = Good Models ? 11:50 Simulation Data 13:40 Exploring Use Cases 15:18 Video Analysis & Reasoning 16:18 Sourcing Data & Processing 17:09 Project Luci & Upload 18:18 Camera Hardware 19:13 Model Architecture 24:03 Infrastructure 25:40 Livestreaming & Latency 26:44 Optimization 27:45 Partnerships & Use Cases 28:25 Data Vendors 30:02 Ray-Ban Meta Glasses 31:06 Personal AI 32:40 Capital Usage 33:40 Model Speed 35:00 Market Outlook 37:15 Outro #robotics #podcast #ai

We sit down with Xara, founder of Somnia, to discuss the engineering behind intimacy robots and AI companions. Learn how mechanical design and robotics integration shape the future of human interaction.This conversation covers the technical challenges of building physical ecosystems for AI. We examine anatomical diagrams, muscle mechanics, and facial expression systems that allow these machines to operate in daily life.Zara shares his vision for Somnia and the practical application of intimacy robots. We explore the intersection of hardware development and artificial intelligence, providing a clear look at the current state of this specialized engineering field.Subscribe for weekly robotics breakdowns, and let us know your thoughts on the future of AI companions in the comments.#ai #girlfriend #robot Guest https://x.com/xarathustraxLéo https://x.com/LeoKharonOskar https://x.com/hundreddaysxPodcast https://x.com/roboticsstack0:00 Podcast Overview0:52 Robot Companionship 1:48 Recent Breakthroughs 2:48 Emotionally Intelligent Robots 3:35 Xaras Journey 4:14 Doll Industry 5:21 Unitree Experiment6:32 Realistic Movement Sequences 7:28 Limitations of Dolls 8:04 Limitations of Unitree G1 8:55 Motion Capture 9:12 Reinforcement Learning 9:30 Voice Control & Navigation 10:15 Edge Device 10:55 Somnia Hardware 13:52 Teleoperation 14:18 Humanoid Robot Business 16:11 Realbotix 17:00 Full Body Control 17:44 3d Printed Muscles 18:02 Body Parts 18:52 Design 19:56 Customization 20:55 Manufacturing 21:14 Lessons from Human Anatomy 22:50 Skin Technology 25:50 Somnia Versions 27:25 Company 28:15 Building the Face 29:32 Hiring for Somnia Lab 31:10 Societal Implications32:08 Testing 32:50 Public Image & Mission 34:40 The Need for Companionship 37:00 Customer Experience & Maintenance 40:20 Technological Development Cycles 42:40 Expanding Use Cases 44:00 Market Adoption & Roadmap 47:30 Funding 48:20 Supply Chain 49:50 Research & Development 52:00 Safety 55:25 Vision 58:30 Outro

We sit down with AK Schultz to discuss the realities of warehouse automation, assembly lines and the current state of robotics integration. SVT Robotics and their Softbots help bridge the gap between complex software systems and physical hardware in modern logistics centers & more. A.K. Schultz, co-founder and CEO of SVT, shares his perspective on the friction points between legacy infrastructure and new automation technology. We examine the specific challenges of scaling deployments in demanding environment and cover the strategic approach required for successful industrial application.#robots #technology #podcast Guest https://www.linkedin.com/in/a-k-schultz-9b5b899Léo https://x.com/LeoKharonOskar https://x.com/hundreddaysxPodcast https://x.com/roboticsstack0:00 Podcast Overview0:35 Getting into Robotics 1:42 Assembly Line Experience 2:45 Building for Industrial Automation 3:20 The Problem with Custom Code Interfaces3:58 Fixing Integration 6:48 Core Principles 7:17 Softbots 8:50 Advantages 10:10 Learnings from Communication Technology11:49 Software Robots 14:10 Robotics & Software in Production 17:05 Productizing Integration Code 20:22 Onboarding & Implementation 23:00 DHL Case Study 24:38 Expansion & Trends 27:32 AI 28:52 Local vs. Cloud31:35 Usage & Data 33:30 1 Billion Transactions a Month in Peak Season 34:12 Customer Size 35:22 Pricing & Connectivity 36:23 Vibecoding & Tokens 37:58 Humanoids 40:27 Automation Bottlenecks 43:45 Starting from Zero 47:05 The need for a business case 49:03 Outro

We sit down with Parsa, co-founder of Feather Robotics, to discuss their new modular hardware platform for real-world tasks. Feather Robotics recently emerged from stealth mode, bringing a fresh approach to how teams build and deploy systems. Parsa explains the philosophy behind their modular hardware platform and why they believe this specific architecture is the key to moving beyond prototypes into actual industrial application.This Podcast covers the practicalities of robotics deployment in real-world environments. We explore how their technology is already being used to solve complex automation challenges and what sets their robot hardware apart in a crowded market.#robots #technology #podcastGuest https://www.linkedin.com/in/parsa-bakhtiari-19a220113Léo https://x.com/LeoKharonOskar https://x.com/hundreddaysxPodcast https://x.com/roboticsstackWe sit down with Parsa, co-founder of Feather Robotics, to discuss their new modular hardware platform for real-world tasks. Feather Robotics recently emerged from stealth mode, bringing a fresh approach to how teams build and deploy systems. Parsa explains the philosophy behind their modular hardware platform and why they believe this specific architecture is the key to moving beyond prototypes into actual industrial application.This Podcast covers the practicalities of robotics deployment in real-world environments. We explore how their technology is already being used to solve complex automation challenges and what sets their robot hardware apart in a crowded market.#robots #technology #podcast Guest https://www.linkedin.com/in/parsa-bakhtiari-19a220113Léo https://x.com/LeoKharonOskar https://x.com/hundreddaysxPodcast https://x.com/roboticsstackSpotify: https://open.spotify.com/show/033M6IhFyhnY6dEWtoBZ1iApple Podcasts: https://podcasts.apple.com/us/podcast/the-robotics-stack/id67888853120:00 Podcast Overview 0:34 Building a hardware platform 1:48 Working at Tesla 2:36 Car Industry vs. Robotics Industry 4:00 Founding Feather Robotics 7:20 What was missing in Humanoids 9:40 Robot Development 10:17 Modularity 12:15 Achieving Competitive Pricing 13:12 Deployment & Use Cases 15:15 Robotics Software 16:21 Measuring Deployment Success 17:23 Trends & Form Factors 19:03 Grippers & Hands 19:33 Sourcing Parts & Manufacturing 21:22 Sales 21:48 Safety Measures 22:12 Funding & Revenue 23:41 Team 24:14 Safety Layer 25:48 Future Outlook 27:00 Humanoids 28:15 The Nvidia of Robotics 29:30 Data Capture Market 30:45 Foundation Models in Robotics 31:30 Adapting 32:28 Finding Talent33:22 Outro

In this conversation, we sit down with Matt Trossen, founder of Trossen Robotics. We speak about the history of his company and what it takes to scale Physical AI with Teleoperation and AI Models. He shares his unique perspective on the robot market and presents the new product lineup which includes reworked Robotic Arms but also UMI Grippers and software like Trossen Cloud. His journey is one of a kind: Coming from USB controlled hardware for sensing and automation to working close with researchers on the ALOHA Project. The recent advancements in Artificial Intelligence opened a completely new world to robots for him and his company. Enjoy!#robotics #ai #teleoperation0:00 Podcast Overview0:40 Getting the word out1:40 Starting a robotics company2:50 ALOHA Project3:16 Early Stages6:35 Moving towards Robotics6:57 ALOHA Kit7:31 Adapting Products8:35 Listening to the Market9:50 New Product Lineup10:25 WidowX Pro10:47 Core11:01 Heavy11:30 Workbench12:10 Glide14:55 Rivet15:30 IP & Design19:00 Form factor convergence19:50 Production & Manufacturing20:48 SDK & Trossen Cloud22:00 The Next Wave of Automation23:10 Going Full Stack24:50 Investment & Scaling25:58 CEO Anish Shah29:25 Partnerships & Clients32:20 Team & Product Development34:25 Data Collection38:31 Humanoids42:51 The Art of Reverse Engineering Nature45:00 UMI Data46:10 Data Mix46:46 Deployment Success48:14 Robot Foundation Models50:00 Market Outlook51:10 Opportunities & Challenges54:45 Revenue & Raising Money58:00 The Value of a Platform Based CompanyGuest : https://x.com/trossenroboticsLéo https://x.com/LeoKharonOskar https://x.com/hundreddaysxPodcast https://x.com/roboticsstack

In this conversation, we sit down with Varun Nair to dig into the high-stakes world of data for robotics and physical AI. We talk about what actually matters for training robot intelligence. From egocentric video to UMI over teleop, simulation and the gaps between them. You will also learn why the field is still figuring out which data scales and the limits of human-operated data, evaluation challenges, and what it will take to move from demos to real deployment. If you’re building, investing in, or just obsessed with robotics and embodied AI, this one’s for you.#robot #data #ai0:00 Podcast Overview0:30 Data Market1:30 Varun's Background3:01 Robotics & Crypto6:27 Physical AI Definition8:32 Simulation for Deployments10:45 World Models11:45 What is good data ?12:42 GPT 6 Astra Evaluation13:50 Data Life Cycle17:01 Modalities & Pseudo Labeling19:40 Depth for LLMs20:19 How labs use data21:38 Deployment Data = Post Training Data22:17 Teleoperation24:00 Patient Capital vs Impatient Capital24:47 Benchmarks26:35 UMI Data for Mid Training27:35 Has Openai been training on robot data?30:59 Humanoid vs other Embodiments31:36 Hard vs Easy Tasks for Robots32:51 Simulation33:59 Niantic Spatial34:42 Is robotics data broken?36:17 Model Companies37:38 Robot Everest38:11 Avea38:47 What Varun looks for40:17 OutroGuest https://x.com/_varunnairLéo https://x.com/LeoKharonOskar https://x.com/hundreddaysxPodcast https://x.com/roboticsstack

Asimov is an Open Source Humanoid Robot and today we have the CPO of the company behind it on the podcast. His name is Matthew Moore and he shares unique insights from Menlo Research. You learn what it really takes to build a robot from scratch, what you can do with Asimov and what the future holds for open robotics. #humanoid #robot #asimov 0:00 Podcast Overview 0:38 Why we need Open Robotics 1:28 Matthew's Journey 5:16 Cooking Robots 6:35 Arriving at Menlo Research 7:34 Asimov Development 10:13 Team Building & Supply Chain 11:40 Actuators 12:50 Asimov 1 14:40 Actuator Simulation 16:35 Component Sourcing 18:58 International Operation 21:22 Compute & Model Training 23:49 Software & Team Ressources 25:10 Design 25:56 Business & Data Center 28:06 Customers & Applications 31:06 Data Collection & Market Reality 33:22 Hands & Grippers 35:28 Product Ambitions 36:12 Asimov 2 37:32 Pre-Assembled Version 38:18 Future of Robotics 40:24 Industry Observations 42:02 Hiring & Partnerships 42:52 Scaling Production 43:35 Outro Guest https://www.linkedin.com/in/matthewpaulmoore Léo https://x.com/LeoKharonOskar https://x.com/hundreddaysxPodcast https://x.com/roboticsstack

GPT 6 Astra is OpenAI's most powerful model. So can it be used for robotics? In this episode we look at various demos that show people implementing GPT 6 Astra to control robotic arms and more. Some work in simulation, some work in real environments. And the results are actually surprising. The LLM even starts showing up on various benchmarks and sometimes outperforms SOTA Physical AI models. Of course, there are caveats to this but the performance is seriously impressive. #chatgpt #astra #gpt #podcast 0:00 Podcast Overview 0:32 GPT 6 Astra 3:35 Demo 1 - Painting Robot Arm 5:43 Demo 2 - In Context Learning 8:30 Demo 3 - Robocurve Office Tasks 11:29 Demo 4 - Hand Simulation 15:16 Demo 5 - Drone Flying 17:38 How to use Astra with Robots 18:07 Policy-like 18:51 Agent 19:32 Teacher-like 19:58 Is Astra a VLA ? 21:20 Was it trained on Robot Data? 22:38 Task Annotation Experiment 24:09 Demo 6 - Safety Testing 26:44 Limitations 29:40 Comparison & Outlook 33:52 Outro Léo https://x.com/LeoKharonOskar https://x.com/hundreddaysxPodcast https://x.com/roboticsstack

In this episode of the robotics stack podcast, we explore tactile data capture hardware with Matt Wulff from 6thSense. He is one of the co-founders and joins us to discuss the specialized robotics hardware required to give robots a realistic sense of touch. We examine how current systems struggle with physical dexterity and why new hardware approaches are essential for advancement. We break down the entire data capture pipeline from deployments to engineering the gloves & camera involved and how to turn this into useful robotics training data. We also speak about what the future holds for hands, humanoids & more!#robot #data #podcast 0:00 Podcast Overview 0:39 The Future of Robotics & 6thSense1:44 Background 2:20 Egocentric Video & Tactile Data 3:15 Stereo vs. Monocular Video 4:43 6thSense Products & Services 5:28 Hardware5:33 Cameras 6:06 Gloves 6:50 Product Development 8:18 Team 9:19 Funding 9:50 Deployments I 10:14 Data Capture Methods 11:06 Deployments II12:38 Competitive Edge 16:50 Tactile Sensors 19:26 UMI Data 21:40 Data Management & Processing 23:10 Data Quality 24:25 Strategy & Partnerships 26:38 More Cameras 28:20 Hand to Gripper Data 29:07 UMI / Teleoperation 31:54 Robot Model Training 33:45 Capture Rig Implementation 35:28 Robotic Hands 36:30 Who will win Humanoids? 38:19 GPT Astra 40:00 Future 41:35 AI 42:55 Stock Picks 45:32 Robot Skin 47:30 Is Video Enough? 49:14 Outro Guest: https://x.com/makeitmattoLéo https://x.com/LeoKharonOskar https://x.com/hundreddaysxPodcast https://x.com/roboticsstack

In this episode of the robotics stack podcast we have our second interview. The guest is Randall Briggs who is working on Steel Bot, an open-source robot platform that is made in America. We are discussing the product, the history and alternatives like the Unitree G1 or Asimov. Randall also shares some interesting historical insights and explains why he has such strong conviction for the project. #opensource #robot #podcast 0:00 Podcast Overview 0:38 Background 2:22 Returning to MIT 6:47 Backdrivability 9:19 Steel Bot 9:40 Inspiration 11:19 The Need for an Alternative14:01 Steel Bot Progress 14:48 FCC Ban 16:24 Permanent Magnets 18:00 Steel Bot Specs 20:49 Modularity 22:08 Hands 23:14 Funding & Team 24:19 Unitree G1 Limitations 28:02 Steel Bot Software 31:30 Open Source Importance 33:39 Apple History & Applications for Robots 39:05 Use Cases & Job Market 43:42 Mass Production & Demand 47:07 Market Dynamics 48:43 Steel Bot Customers 49:38 Contact 50:30 Conviction & Vision 51:34 Business Opportunities 53:35 Selling Actuators 54:32 Outro Guest: https://x.com/randallmbriggsLéo https://x.com/LeoKharonOskar https://x.com/hundreddaysxPodcast https://x.com/roboticsstack