The Robots Already Working in American Factories | Tutor Intelligence...
a16z PodcastFull Title
The Robots Already Working in American Factories | Tutor Intelligence
Summary
This episode features Josh Gruenstein, CEO of Tutor Intelligence, discussing their approach to building general-purpose robots by deploying them in real-world scenarios, collecting data, and using human operators to train AI models.
The conversation highlights the challenges and opportunities in the robotics industry, emphasizing the need for a full-stack solution and how Tutor Intelligence aims to become a new labor substrate for the physical economy.
Key Points
- Tutor Intelligence's core thesis is to deploy robots into the physical economy to build necessary datasets, models, and supply chains for widespread robot adoption, unlike many robotics companies that focus solely on model development.
- Building successful robots requires a simultaneous solution across AI, hardware, manufacturing, and go-to-market strategies, with the belief that general-purpose robots could form a new labor substrate for the physical economy.
- The company's unique approach involves "tutoring" robots by connecting them to human operators who provide real-world work data to continuously improve robot capabilities.
- Robotics is viewed as a systems problem requiring trade-offs across multiple layers of the stack, including intelligence, hardware, software, supply chain, and customer integration, to achieve progress.
- Tutor Intelligence has developed two primary robot form factors, Cassie (for heavy work with suction cups) and Sunny (a wheeled industrial semi-humanoid), designed for specific industrial and logistics tasks.
- Customer adoption is challenging as most potential clients have never purchased robots before, requiring a value proposition like an affordable hourly rate to overcome initial skepticism.
- The company emphasizes building robots that are seen as functional equipment by management, while cultivating relationships between factory workers and robots through naming and integration to ease the transition.
- Tutor Intelligence is actively building out a mature robotics supply chain in the United States, including manufacturing its own hardware in Massachusetts, to support large-scale deployment.
- Their research program focuses on real-world reinforcement learning at scale, utilizing a 100-robot fleet as a data collection farm to train advanced models.
- The company believes that by making physical economy tasks cheaper and more efficient, robots can unlock new economic opportunities and transform industries like manufacturing and construction.
Conclusion
The future of robotics lies in building general-purpose robots that can adapt to diverse tasks, creating a new labor substrate for the physical economy.
Tutor Intelligence's strategy of focusing on a full-stack solution, from hardware to customer deployment, is crucial for scaling robotics adoption.
By addressing the entire robotics ecosystem and emphasizing economic viability, companies like Tutor can drive significant transformation in how goods are manufactured and delivered.
Discussion Topics
- How can the successful deployment of general-purpose robots reshape global supply chains and manufacturing strategies?
- What are the ethical and economic implications of robots becoming a new "labor substrate" for the physical economy?
- In what ways does the innovation cycle in robotics, as exemplified by Tutor Intelligence, differ from or mirror advancements in digital technologies like LLMs?
Key Terms
- Greenfield market
- A market in which a company is entering for the first time, with no established competitors or infrastructure.
- Full-stack platform
- A comprehensive technological solution that addresses all layers of a particular domain, from hardware to software and user interfaces.
- Labor substrate
- The underlying workforce or labor force that supports economic activity and production.
- Foundation models
- Large AI models trained on broad data that can be adapted to a wide range of downstream tasks.
- Reinforcement learning
- A type of machine learning where an agent learns to make decisions by taking actions in an environment to maximize a reward.
- Human-Robot Interaction (HRI)
- A field of study focused on the design, development, and evaluation of robotic systems that interact with humans.
- OEM (Original Equipment Manufacturer)
- A company that manufactures products that are then branded and sold by another company.
Timeline
Tutor Intelligence focuses on deploying robots to build data, models, and supply chains for widespread adoption.
The company's approach involves deploying robots, connecting them to human operators for training, and using real-world work data to enhance robot capabilities.
Josh Gruenstein founded Tutor Intelligence with the thesis that deploying robots in the physical economy would enable the creation of necessary data sets, models, and supply chains.
Robotics is fundamentally a systems problem requiring the integration of intelligence, hardware, software, supply chain, and customer considerations.
Many robot companies start horizontally and then move towards verticalization due to the immaturity of the market, a pattern seen in early computing.
The concept of a general-purpose humanoid robot mirrors the CPU in computing, serving as a foundational platform for various applications.
Tutor's name reflects its approach of leveraging networked human intelligence as "tutors" to train robots, analogous to how the internet's data trains LLMs.
The seamless integration of robots and their learning models, where customers focus on outcomes like cost reduction, indicates successful deployment.
The economic marvel of generative AI, where training data distribution matches economic viability, is a model Tutor hopes to replicate in robotics.
Advancements in LLMs, particularly their ability to perform 3D spatial reasoning through post-training on 3D spatial data, are accelerating robotics capabilities.
Tutor designs and manufactures its own hardware, moving beyond sourcing from OEMs to create robots optimized for frontier AI models.
Tutor produces two robots: Cassie, designed for heavy work with suction cups, and Sunny, an industrial semi-humanoid on wheels for faster, more reliable operation.
The majority of work in US manufacturing and logistics involves manual tasks such as transporting pallets and cases, which Tutor's robots are designed to address.
Customer adoption is challenging as many are first-time robot buyers, requiring a focus on economic benefits like a low hourly cost to gain traction.
The perception of robots shifts from beings to functional machines once they prove effective in operations and contribute to the bottom line.
Tutor serves a broad market, including many small and mid-market businesses, which forces them to build a general-purpose, adaptable product rather than custom solutions.
Increased robot adoption is expected to lead to lower costs, improved quality, and the expansion of physical economy businesses.
The increasing general-purpose nature of robots is expected to fundamentally change factory locations and organization, enabling more localized and flexible manufacturing.
The global supply chain remains highly interconnected, with both US and Chinese companies strategizing for long-term industrial capacity building.
Tutor is building its own robotic supply chain capacity in the US, verticalizing component manufacturing to reduce reliance on OEMs and drive innovation.
Tutor's research focuses on real-world reinforcement learning, using a large fleet of robots to gather data and train models at scale.
Hiring for Tutor involves a multidisciplinary team encompassing researchers, engineers, and technicians across all layers of the technology stack.
Tutor's customers are located across the United States, requiring a distributed team of market and technical professionals.
The episode concludes with standard podcast outro information and a disclaimer.
Episode Details
- Podcast
- a16z Podcast
- Episode
- The Robots Already Working in American Factories | Tutor Intelligence
- Official Link
- https://a16z.com/podcasts/a16z-podcast/
- Published
- November 2, 2026