Minus One

Sergey Levine, cofounder of Physical Intelligence, joins South Park Commons Partner Aditya Agarwal to discuss the race to build general-purpose robots that can learn new skills and operate in environments they've never encountered before.

They talk about how Physical Intelligence is developing AI foundation models for robotics, what makes teaching robots so difficult, and how advances in reasoning and reinforcement learning are bringing us closer to robots that can perform a wide range of tasks.

Sergey also explains why solving robotics may require tackling many problems at once, how robots could eventually learn from their own experiences, and what it will take to make truly general-purpose robots a reality.

1. Sergey Levine: https://www.linkedin.com/in/sergey-levine-5a31a24/ 
2. Aditya Agarwal: https://www.linkedin.com/in/adityaagarwal3/ 
3. South Park Commons: https://www.linkedin.com/company/southparkcommons/ 

Apply to SPC: https://www.southparkcommons.com/apply

Chapters:

(00:00:28) - From Video Games to Robotics: Sergey Levine's Journey

(00:04:24) - Founding Physical Intelligence: Why the Timing and Team Mattered

(00:06:30) - Inside Pi Zero: Building a Foundation Model for Robots

(00:09:37) - From Pi 0.5 to Pi 0.7: Teaching Robots to Reason and Generalize

(00:16:45) - What Makes a Great Robotics Researcher

(00:21:10) - Solving Robotics' Data Problem and Building Learning Feedback Loops

(00:28:04) - The Future of Robotics: General-Purpose AI vs. Specialized Robots

(00:33:24) - World Models, Data Strategy, and Building a Robotics Research Culture

Dmitri Dolgov, co-CEO of Waymo, joins South Park Commons Partners Aditya Agarwal and Ruchi Sanghvi to discuss the 17-year journey from early self-driving experiments to putting fully autonomous cars on public roads.

They talk about how Waymo built its self-driving system, why simulation and evaluation became so critical, and what it takes to move from something that works in a demo to a product that can operate safely and reliably at scale.

Dmitri also explains why Waymo’s real advantage goes far beyond the model itself, and what it takes to keep improving a system where the margin for error is incredibly small.

1. Dmitri Dolgov: linkedin.com/in/dmitri-dolgov 
2. Aditya Agarwal: https://www.linkedin.com/in/adityaagarwal3/ 
3. Ruchi Sanghvi: https://www.linkedin.com/in/rsanghvi/ 
4. South Park Commons: https://www.linkedin.com/company/southparkcommons/ 

Apply to SPC: https://www.southparkcommons.com/apply

Chapters:

(00:00:32) - The First Time Waymo Felt Real

(00:03:39) - How a Fleet of Waymos Coordinates

(00:07:47) - How People Change the Way They Travel With Waymo

(00:12:51) - How Waymo Started: Understanding the Problem Before Building a Product

(00:14:58) - Inside Waymo’s AI System: Driver, Simulator, and Critic

(00:22:00) - How Waymo Adopts New AI Breakthroughs Without Breaking the Product

(00:27:02) - Why the Last 1% of Autonomous Driving Is So Hard

(00:25:16) - Where Startups Should Build in Physical AI

(00:44:30) - Can the Waymo Driver Become a General Physical AI Agent?

Helmy Eltoukhy, CEO of Guardant Health, joins South Park Commons Partner Mark Jacobstein to discuss his journey from electrical engineering to building one of the leading companies in cancer diagnostics.

They talk about Guardant’s path from liquid biopsy to early cancer detection, the difficult realities of building a diagnostics company, and why business model innovation can matter just as much as technical breakthroughs.

Helmy also explains how AI is helping Guardant make better use of its growing dataset and why he believes the same blood test infrastructure could eventually be used to detect hundreds or even thousands of diseases.

1. Helmy Eltoukhy: https://www.linkedin.com/in/helmy-eltoukhy-b011855/
2. Mark Jacobstein: https://www.linkedin.com/in/markjacobstein
3. South Park Commons: https://www.linkedin.com/company/southparkcommons/ 

Apply to SPC: https://www.southparkcommons.com/apply

Chapters:

(00:00:43) - From Electrical Engineering to Biology and Entrepreneurship

(00:08:10) - From Building Sequencing Technology to Solving Healthcare Problems

(00:12:51) - Guardant Health's Long-Term Bet on Early Cancer Detection

(00:15:31) - Early Mistakes, Pivots, and Difficult Business Decisions

(00:18:15) - Why Building a Diagnostics Company Is So Difficult

(00:21:15) - Why Business Model Innovation Matters as Much as Technology

(00:26:15) - How AI Is Unlocking Guardant's Massive Cancer Dataset

(00:30:05) - From Early Cancer Detection to the “Everything Test”

How close are we to AGI, and what happens when AI systems stop behaving the way we expect?

Srinivas Narayanan, former CTO of B2B Engineering at OpenAI and former VP of Engineering at Facebook, joins South Park Commons Partner Prateek Mehta for a wide-ranging conversation on his Minus One to Zero journey, enterprise AI adoption, and AI safety.

They discuss the career choices that took him from two startups to Facebook and OpenAI, why a beginner’s mindset has mattered along the way, what is slowing AI adoption inside large companies, and how builders should think about reward hacking, alignment, and increasingly autonomous AI systems.

1. Srinivas Narayanan: https://www.linkedin.com/in/srinivasnarayanan/ 
2. Prateek Mehta: https://www.linkedin.com/in/prateek-mehta-a571972/ 
3. South Park Commons: https://www.linkedin.com/company/southparkcommons/

Apply to SPC: https://www.southparkcommons.com/apply

Chapters:

(00:01:26) - How Close Are We to AGI? 

(00:02:26) - From IIT Madras to AI and OpenAI 

(00:07:32) - How to Find Problems Worth Spending a Decade On 

(00:12:38) - Building OpenAI Without a Playbook 

(00:15:24) - How OpenAI Built a High-Talent, High-Agency Team 

(00:18:12) - What’s Holding Back Enterprise AI Adoption 

(00:29:12) - The Future of AI Beyond the Digital World 

(00:36:42) - The Real Risks of AI and How to Think About Safety 

What does it take to build something that lasts? Snowflake CEO and former Google SVP of Ads Sridhar Ramaswamy joins South Park Commons General Partner Aditya Agarwal to share his answer.

Ramaswamy traces his journey from Google to founding Neeva, an ad-free search startup that shut down before the generative AI boom, to leading Snowflake through the AI era. He shares what defines great founders, what doomed Neeva, and why most software will lose its moat.

Ramaswamy and Agarwal also explore the future of AI, what remains valuable in an age of AI-generated software, and the search belief Ramaswamy got wrong.

1. Sridhar Ramaswamy: https://www.linkedin.com/in/sridhar-ramaswamy/ 
2. Aditya Agarwal: https://www.linkedin.com/in/adityaagarwal3/ 
3. South Park Commons: https://www.linkedin.com/company/southparkcommons/

Apply to SPC: https://www.southparkcommons.com/apply

(00:00:52) - Choosing When to Take a Career Risk

(00:05:01) - Lessons From Google’s Founders 

(00:09:52) - What Went Wrong at Neeva 

(00:13:39) - The Future of AI Models 

(00:22:59) - Snowflake’s Enduring Value in the AI Era 

(00:29:24) - How to Keep Up With AI 

(00:32:24) - Rapid Fire: Books, Search & Google

Why did OpenAI’s model hack into Hugging Face?

Tom McGrath, co-founder and Chief Scientist at Goodfire, joins South Park Commons Partner Jonathan Brebner to explore one of AI’s biggest challenges: interpretability.

Using OpenAI’s recent reward hacking incident as a starting point, they discuss why today’s most advanced models are still difficult to understand and why gaining visibility into how they work could be essential for building safer, more reliable AI systems.

Tom shares how interpretability could change the way researchers develop models and how Goodfire is creating tools to help understand and steer frontier AI.

1. Tom McGrath: https://www.linkedin.com/in/tom-mcgrath-7337bb151/ 
2. Jonathan Brebner: https://www.linkedin.com/in/jonathan-brebner/ 
3. South Park Commons: https://www.linkedin.com/company/southparkcommons/

Apply to SPC: https://www.southparkcommons.com/apply

(00:00:30) - Why AI Interpretability Matters Now

(00:02:15) - When AI Reward Hacking Becomes Real

(00:05:07) - The Core Problem of Interpretability

(00:07:23) - Intentional Design: Steering What Models Learn

(00:12:43) - Neural Geometry: The Shapes Inside AI Models

(00:22:29) - Turning Interpretability Research Into a Product

(00:25:28) - How Research Changes Inside a Startup

(00:32:48) - Science Fiction, AI Risk, and the Future

Thariq Shihipar, a member of Anthropic's Claude Code team (and SPC alum), joins South Park Commons Partner Evan Tana to discuss what it takes to build at the frontier of AI, where the models, tools, and best practices are constantly evolving.

Thariq traces his journey from a gaming startup to Anthropic after a chance text about Claude Code, shares why his team stopped using Plan Mode, explains concepts like capability overhang, harness engineering, and the underrated art of writing evals, and discusses why RAG is becoming an anti-pattern. He also breaks down how founders should think about building when the capabilities of AI systems are changing faster than anyone can predict.

1. Thariq Shihipar: https://www.linkedin.com/in/thariqshihipar/ 
2. Evan Tana: https://www.linkedin.com/in/evantana/ 
3. South Park Commons: https://www.linkedin.com/company/southparkcommons/

Apply to SPC: https://www.southparkcommons.com/apply

Chapters:

(00:00:00) - From Startup Founder to Anthropic 

(00:03:33) - Building Better AI With Claude Code 

(00:07:21) - The Capability Overhang 

(00:12:17) - How Anthropic Builds AI Products

(00:17:51) - Prompting Claude More Effectively 

(00:30:58) - How AI Is Changing Product Teams 

(00:41:11) - Advice for AI Founders

What does it actually take to become an astronaut? Shubhanshu Shukla, the second Indian to travel to space after more than four decades, joins South Park Commons general partner Aditya Agarwal to share the answer. 

During Axiom Mission 4, Shukla became the first Indian aboard the International Space Station (ISS). An astronaut with ISRO and a Group Captain in the Indian Air Force, he shares what it takes to get there, from the intense selection process and psychological testing to a seven-day isolation challenge and other unexpected requirements candidates must meet.

He also reflects on life in orbit, from speaking Hindi aboard the ISS for the first time to the disorientation, the overview effect, and the physical changes the human body experiences. Along the way, Shukla and Agarwal explore India’s space ambitions and what it takes to chase an outsized dream.

1. Shubhanshu Shukla: https://www.linkedin.com/in/gagan-shux/ 
2. Aditya Agarwal: https://www.linkedin.com/in/adityaagarwal3/ 
3. South Park Commons: https://www.linkedin.com/company/southparkcommons/

Apply to SPC: https://www.southparkcommons.com/apply

(00:00:00) - Introduction: Becoming Comfortable With the Uncomfortable

(00:01:54) - Why Saying "Yes" Changed the Course of His Life

(00:07:15) - What It Really Takes to Become an Astronaut

(00:11:40) - The Physical, Medical, and Psychological Tests

(00:15:21) - What It Feels Like to See Earth From Space

(00:17:20) - Microgravity: How Space Changes Your Body and Mind

(00:25:35) - The Pressure and Responsibility of Representing India in Space

(00:28:29) - The New Space Economy and Why the Next Decade Will Be Different

(00:35:03) - Advice for Entrepreneurs Building the Future of Space

(00:44:19) - The Future of Human Space Exploration: Moon, Mars, and Beyond

Nato Saichek, cofounder and CTO of Longshot Space, joins South Park Commons Partner Jonathan Brebner to explain why rockets won the space race for the wrong reasons—and why a 15-kilometer space gun may be the most rational path to cheap access to orbit. 

Nato traces Longshot's origins from garage experiments with his cofounder Mike Grace, through building the world's largest operational cannon, to a first-principles reimagining of launch economics. He explains why $4 trillion in rocket investment locked civilization into a local optimum, how aerospace-grade hardware can be replicated for $5 in parts, and what it actually feels like to shoulder the weight of a deeply unconventional bet.

1. Nato Saichek: https://www.linkedin.com/in/natosaichek/ 
2. Jonathan Brebner: https://www.linkedin.com/in/jonathan-brebner/ 
3. South Park Commons: https://www.linkedin.com/company/southparkcommons/

Apply to SPC: https://www.southparkcommons.com/apply

Chapters:

(00:00:00) - The 15-Kilometer Space Gun

(00:06:27) - The History of Space Guns and Why the Idea Was Abandoned

(00:09:33) - Why Rockets Aren’t the Best Solution

(00:18:39) - The Story of Longshot Space

(00:22:09) - Nato On Leaving His Dream Job

(00:28:25) - Fundraising for an Unconventional Idea

(00:39:36) - The Perks of Having a Cofounder

Amit Jain, cofounder and CEO of Luma AI, joins South Park Commons Partner Finn Meeks to explain why the most defensible AI companies aren't just building tools—they're developing systems that can understand and simulate the world.

Amit traces Luma's path from a free 3D capture app designed to gather training data, through an early bet that video generation was finally viable, to the unified model architecture powering Luma Agents today. He explains why most AI products still break at execution, why foundation labs blur the line between product and research, and why they are the blueprint for the companies of the future.

1. Amit Jain: https://www.linkedin.com/in/gravicle/ 
2. Finn Meeks: https://www.linkedin.com/in/finn-meeks/ 
3. South Park Commons: https://www.linkedin.com/company/southparkcommons/
   

Apply to SPC: https://www.southparkcommons.com/apply

Chapters:

(00:00:00) - Apple's LiDAR work seeded Luma's founding vision

(00:05:05) - A free app was secretly a data collection operation

(00:06:49) - H100s made video generation stop feeling impossible

(00:10:00) - Most definitions of "world model" are simply wrong

(00:14:37) - Why unified models beat pure video scaling

(00:20:03) - Luma Agents and the end-to-end creative production loop

(00:23:00) - The knowledge gap no model can close alone

(00:29:34) - Blowing James Cameron's mind in five minutes

(00:31:16) - Why consumer video generation failed—and who it's actually for

(00:35:23) - Product and research aren't two things at a foundation lab