The Lobster Talks Podcast by Lobster Capital

What if you could simulate human reactions — and know exactly how your customers, investors, or audience will respond before you act? James He is building precisely that with Artificial Societies. In this episode, we dive deep with James, YC W25 founder of Artificial Societies — a wildly ambitious startup that simulates entire groups of humans using AI personas to predict how messages spread, markets react, and products succeed (or fail). From simulating 1,000 VCs to get into YC… to replacing $20,000 market research surveys, this conversation is a masterclass in founder execution, behavioral science, and AI-first GTM strategy. What you’ll learn: • How Artificial Societies works (and how founders are really using it today) • Why traditional A/B testing and market research are broken • Product-led growth lessons from 15,000+ activated users • The surprising YC tricks James used to get into the batch • Practical ways AI can upgrade your messaging, GTM, and PMF search • The future of simulating entire economies (and why we’re not there yet) 📌 Watch now and subscribe for more smart, fast, founder-first interviews. Chapters: 00:00 Simulating Investors: The Early Days 00:27 Introducing Artificial Societies 00:48 Journey to YC and Beyond 01:06 How Artificial Societies Works 03:02 The Science Behind Social Influence 05:27 Technical Insights and Applications 17:28 James' Personal Journey 22:29 Building and Growing the Startup 24:50 Using Internal Tools for Optimization 27:38 The Evolution of A/B Testing with AI 29:41 Challenges in Market Research 33:14 AI's Role in Market Research 34:40 Balancing Customer Feedback and Product Vision 42:55 Future Predictions and AI Capabilities 47:48 Closing Remarks and Final Thoughts 🎧 Listen on the go: Spotify: https://open.spotify.com/show/1u1JyRKH8JYFjhBkvTUOAP Apple: https://apple.co/4cZ8RMm My other YouTube Channel: https://www.youtube.com/@UCahCIZ9KfmeK4rLosGNYpdg My other YouTube Channel: https://www.youtube.com/@GJarrosson

YC just dropped its latest Requests for Startups — and they’re not just ideas, they’re roadmaps to the future. From AI-native enterprise software to 10-person $100B companies, these signals reveal where the smartest founders (and investors) should be looking. In this episode of Lobster Talks, Laurie and I break down Y Combinator’s newest RFS — from retraining workers for the AI economy to video generation as a computing primitive. We debate the real opportunities, the traps, and how these trends could reshape the startup ecosystem. What you’ll learn: - Why YC’s RFS is one of the best ways to predict upcoming demo day winners - The massive opportunity in retraining workers for the AI economy - Why video generation is shifting from an output to a primitive - How AI-native startups could disrupt giants like Salesforce - Whether 10-person $100B companies are actually possible - The risks, politics, and second-order effects of LLMs in government Chapters: 00:00 Introduction: The Value of Customer Money 00:26 Y Combinator's Request for Startups 01:14 AI and Worker Retraining 02:30 The Future of Semiconductors and Deep Tech 04:34 AR, VR, and AI in Training 09:02 Video Generation: The Next Frontier 16:12 The First 10-Person $100 Billion Company 22:31 The Future of Small Teams and Fundraising 24:03 Challenges and Opportunities in Multi-Agent Systems 28:12 AI Native Enterprise Software 34:39 Using AI to Revolutionize Government Consulting 41:24 Conclusion and Upcoming Events 📌 Watch now and subscribe for more smart, fast, founder-first interviews. 🎧 Listen on the go: Spotify: https://open.spotify.com/show/1u1JyRKH8JYFjhBkvTUOAP Apple: https://apple.co/4cZ8RMm My other YouTube Channel: https://www.youtube.com/@GJarrosson

Video editing hasn’t changed in decades — until now. Mosaic is killing the timeline and replacing it with agent-powered automation that can take you from raw footage to polished video in record time. In this episode of Lobster Talks, I sit down with Adish Jain, co-founder of Mosaic, the YC startup rethinking how creators edit video. We talk about Mosaic’s agentic canvas approach, why timelines are becoming optional, and how AI will bring millions of new creators into the fold. You’ll learn: - Why timelines are a relic — and what replaces them - How agent-powered editing can cut 12-hour renders to seconds - The human-in-the-loop design challenge for AI video tools - How Mosaic balances speed, control, and creativity - Why fully autonomous editing is closer than you think - The contrarian bet that could reshape the editing industry 📌 Watch now and subscribe for more smart, fast, founder-first interviews. Chapters: 00:00 Reimagining Video Editing in the AI Era 00:37 Introduction to Mosaic and Its Vision 01:21 Challenges and Innovations in AI Video Editing 03:22 The Future of Autonomous Video Editing 08:18 The Inspiration Behind Mosaic 12:36 Building and Marketing Mosaic 17:43 Navigating User Expectations and Product Development 23:40 Debating the Perfection of AI 23:54 Utility of AI in Video Editing 24:48 Future of AI Agents in Video Editing 26:35 The Rise of Content Creators 27:43 Competing in a Saturated Market 30:04 Generative AI and Video Editing 33:14 Revolutionizing Video Editing with Mosaic 37:33 The Future of Video Editing 44:18 Exciting Developments at Mosaic 45:46 Conclusion and Farewell 🎧 Listen on the go: Spotify: https://open.spotify.com/show/1u1JyRKH8JYFjhBkvTUOAP Apple: https://apple.co/4cZ8RMm My other YouTube Channel: https://www.youtube.com/@GJarrosson

What if you built a tool so powerful, it became the #1 repo on GitHub—twice? That’s exactly what Daniel Farrell did with Onlook, an open-source AI-powered visual editor for code. From crashing Chrome extensions to crashing Hacker News, Daniel takes us through how he and his co-founder built momentum, community, and virality from nothing. In this fast, founder-first conversation, we cover: • How Onlook went from prototype to #1 GitHub repo • Why Hacker News beat Product Hunt for traction • The real benefits and tradeoffs of going open source • How open source GTM can win enterprise customers • What most people misunderstand about AI and design• Why taste, story, and community still matter more than code 📌 Watch now and subscribe for more smart, fast, founder-first interviews. 00:00 The Most Trending Repo in the World 00:16 Meet Daniel Farrell: Co-Founder of Onlook 00:59 The Vision Behind Onlook 01:57 Building the Prototype and Early Challenges 04:55 Finding the Perfect Co-Founder 11:57 The Open Source Journey 14:49 The Hacker News Effect 16:47 Global Impact and Community Building 19:40 The Importance of Open Source 24:31 Challenges of Maintaining a Community 25:00 Open Source Success Stories 25:21 The Super Base Playbook 26:33 Advantages of Open Source for Enterprises 33:42 Impact of AI on Design and Marketing Teams 42:25 Future Bets and Bold Predictions 48:00 Where to Find Onlook and Final Thoughts Listen on the go: Spotify: https://open.spotify.com/show/1u1JyRKH8JYFjhBkvTUOAP Apple: https://apple.co/4cZ8RMm My other YouTube Channel: https://www.youtube.com/@GJarrosson

What if launching a tech startup was as easy as filling out a form? Ben Collins, co-founder of Woz, joins Gabriel to unpack how his AI platform helps anyone — even non-technical founders — build full-stack software businesses. From betting against vertical SaaS to designing an AI cofounder you can trust, this conversation is a tactical deep dive into where startup creation is headed. 🔍 In this episode: Why Woz pivoted away from a holding company model What most AI-first code tools get totally wrong How mobile-first enables higher quality standards A unique approach to betting on future AI capability curves The real reason most AI startups are failing users 📌 Watch now and subscribe for more smart, fast, founder-first interviews. ⏱ Chapters: 00:00 Introduction to Scaling and Bottlenecks 00:18 Introducing Ben Collins and Woz 01:04 The Vision Behind Woz 03:43 The Founding Story 06:06 The Shift to a Platform Approach 08:53 Navigating the Competitive Landscape 12:39 Philosophical and Strategic Insights 17:02 The Mobile-First Strategy 25:34 The Journey to Product-Market Fit 27:15 Navigating Rapid Technological Changes 32:21 The Impact of Y Combinator 35:00 AI's Role in Scaling Service Businesses 36:42 Challenges for Non-Technical Founders 39:13 Bold Predictions and Future Bets 45:02 Exciting Announcements and Closing Thoughts 🎧 Listen on the go: Spotify: https://open.spotify.com/show/1u1JyRKH8JYFjhBkvTUOAP Apple: https://apple.co/4cZ8RMm My other YouTube Channel: https://www.youtube.com/@GJarrosson

YC isn't just about AI anymore — it's quietly becoming America’s new industrial engine. In this episode, Gabriel and Laurie break down how Y Combinator is doubling down on hardtech, defense, and energy — and why this matters more than ever. From geopolitical risk to deep tech traction, we unpack how YC startups are shaping the future — and why top seed investors are scrambling to keep up. We also reveal how internal partner dynamics at YC shape which billion-dollar startups get funded. 🎙 This is your tactical, insider breakdown of what’s next for the world’s most important startup accelerator. What you’ll learn: * Why YC is going all-in on defense, energy, and hardware * How LOIs and demo day results work for hardtech startups * The “too late” moment for most seed investors trying to access YC deals * Why YC group partners matter — and how to track them * What the NextView “existential crisis” blog means for the seed market * Which partners are shaping the S25 batch behind the scenes 📌 Watch now and subscribe for more smart, fast, founder-first interviews. Chapters00:00 First Impressions and Initial Skepticism 00:36 Welcome to the Podcast 00:52 YC's New Request for Startups 01:57 YC's Hardware and Industrial Focus 07:08 The Importance of Energy and Defense 08:26 Political and Regulatory Considerations 14:27 Investing in Deep Tech and Hardware 20:15 Challenges in Seed Investing 21:25 The Impact of YC on Seed Investing 22:24 The YC Network and Its Influence 23:35 Challenges and Opportunities for Non-YC Investors 24:25 The Journey of Investing in YC 25:42 The Importance of Timing in YC Investments 29:39 Changes in YC Leadership 33:42 The Role of Group Partners in YC 37:35 Personal Experiences with YC Group Partners 39:30 Conclusion and Next Steps 🎧 Listen on the go:Spotify: https://open.spotify.com/show/1u1JyRKH8JYFjhBkvTUOAPApple: https://apple.co/4cZ8RMmMy other YouTube Channel: https://www.youtube.com/@GJarrosson

What happens when a college student builds an AI tool that goes viral overnight — and changes how thousands of students study? Summary: In this episode of Lobster Talks, Gabriel sits down with David Yu, co-founder of YouLearn — the YC-backed AI tutor that transforms PDFs, class recordings, and YouTube videos into quizzes, notes, and a personal tutor. From a dorm-room side project to a viral social growth engine, this episode is packed with tactics, conviction, and bold bets on the future of learning. You’ll learn: How YouLearn hit 10,000 users in 24 hours from a single post The exact tactics they used to blow up on Instagram and TikTok Why most AI tools in education fail — and how YouLearn avoids it The truth about retention, virality, and real product usage How David thinks about AI hallucinations and data safety What learning could look like in 2030 with Neuralink, glasses, and beyond 📌 Watch now and subscribe for more smart, fast, founder-first interviews. Chapters: 00:00 Viral Launch on Social Media 00:20 Introduction to YouLearn 00:47 Meet the Founder: David Yu 01:09 The Birth of YouLearn 02:37 Early Challenges and Successes 05:09 Going Viral Again 05:45 Handling Rapid Growth 08:14 Maintaining Motivation 10:53 The Importance of Co-Founders 12:00 Transition to Full-Time 15:30 Iterating on the Product 18:36 Joining Y Combinator 21:04 Investor Skepticism Towards YC 21:49 Misconceptions About Joining YC 22:33 Right and Wrong Reasons to Join YC 23:28 Maximizing YC Experience 25:08 Navigating YC with Existing Revenue 30:26 AI's Role in Education 31:48 Future of Learning with AI 34:05 Misconceptions About AI in Learning 35:10 The Evolution of Learning Tools 40:26 Final Thoughts and Future Prospects 🎧 Listen on the go: Spotify: https://spoti.fi/4m2K7ah Apple: https://apple.co/4cZ8RMm My other YouTube Channel: https://www.youtube.com/@GJarrosson

What if your inbox didn’t just organize itself — it acted on your behalf? Nizar Abi Zaher, founder of Zero, is building the AI-native email client that reads, replies, filters, and organizes without you lifting a finger. In this tactical, fast-paced conversation, we explore his path from Cal.com engineer to YC founder — and how he's going head-to-head with Superhuman. From open-source strategies to voice-powered inboxes, Nizar breaks down how Zero is redefining email — and what it takes to ship fast, build in public, and stay top of mind. In this episode, you’ll learn: • Why Zero was built around what people hate • How YC pushed them to launch before they were ready • How being open-source accelerates product velocity • What Superhuman got wrong — and what Zero does better • How social presence became their biggest moat • What email looks like in 5 years (hint: voice + AI agents) 📌 Watch now and subscribe for more smart, fast, founder-first interviews. Chapters: 00:00 Introduction to Zero: The AI-Powered Email Client 00:42 Meet the Founder: Nazar's Journey 01:24 The Problem with Email and Zero's Solution 02:56 Differentiating from Competitors 05:21 Rapid Development and Open Source Benefits 11:08 The Importance of Social Media Presence 14:57 Strategies for Effective Social Media Use 21:26 The Value of Authenticity on Social Media 24:07 The Importance of Onboarding 27:02 Launching and Iterating Quickly 29:31 Voice and Email Innovations 32:42 The Future of Email 40:57 The YC Experience and Final Thoughts 🎧 Listen on the go: Spotify: https://spoti.fi/4m2K7ah Apple: https://apple.co/4cZ8RMm My other YouTube Channel: https://www.youtube.com/@GJarrosson

They didn’t sell to insurance brokers — they replaced them. Dakotah Rice didn’t build AI tools for the industry. He built an AI-first insurance company — and now Harper is scaling faster than they can handle. In this episode, Dakotah shares how Harper went from kitchen-table experiments to a high-growth, fully autonomous insurance brokerage. He breaks down the real bottlenecks, the AI architecture behind their workflows, and the bold bets that could reshape an entire industry. What you’ll learn: Why Harper ditched SaaS and chose vertical integration How to embed engineers directly in operational workflows The underrated bottlenecks to AI-first company scaling Why relationships are overrated in insurtech How to build trust with real-world customers — without mentioning AI The real limits (and future) of fully autonomous firms 📌 Watch now and subscribe for more smart, fast, founder-first interviews. Chapters 00:00 Introduction to Harper and Dakota Rice 00:38 Dakota's Background and Harper's Mission 02:13 Challenges and Pivots in Building Harper 04:29 Entering Y Combinator and Scaling Up 06:19 The Importance of Workflow Understanding 13:17 Lessons from Past Failures 18:34 Vision for AI in the Future 21:51 Building Complex AI Systems 22:49 Human Element in AI Services 23:53 Customer Perception of AI 27:37 Scaling Challenges and Solutions 35:14 AI's Role in Future Business Models 37:58 Future Predictions and Industry Insights 42:46 Conclusion and Contact Information 🎧 Listen on the go: Spotify: https://spoti.fi/4m2K7ah Apple: https://apple.co/4cZ8RMm My other YouTube Channel: https://www.youtube.com/@GJarrosson

What if a single AI voice agent could handle 100,000 calls a day and free up your entire support team? Episode Summary: In this episode of Lobster Talks, Gabriel sits down with Kevin Wu, founder of Leaping AI, to unpack how voice-powered AI is automating 50–70% of repetitive support calls, self-optimizing over time, and achieving 90% customer satisfaction. We dive into the origin story, technical stack, ROI metrics, human-in-the-loop safeguards, and bold predictions for 2030. You’ll learn: How Leaping AI began: from BCG ideation to the first major enterprise customer The differentiator: combining voice AI with evaluation and self-optimization modules Managing edge cases and human fallback to maintain 90% satisfaction Tech stack overview: orchestration layer, Twilio, multiple TTS/ASR providers, LLMs Calculating ROI and setting realistic metrics with weekly check-ins Future bets: AI vs. human empathy, integration challenges, market structure by 2030 Practical advice for CX leaders overwhelmed by AI hype Chapters: 00:00 Introduction to Leaping AI with Kevin Wu 00:59 Kevin Wu's Journey and Inspiration 02:10 Winning the First Customer 02:41 Leaping AI's Unique Selling Points 03:59 Self-Improving AI Agents 05:17 Handling Edge Cases and Human Involvement 06:32 Early Days and First AI Call 07:33 Challenges and Continuous Improvement 09:00 Market Position and Competitor Insights 13:47 Customer Metrics and ROI 17:18 Technical Stack and Partnerships 19:09 Sponsor Message and Text-to-Speech Choices 19:32 The Best AI Voice Provider: Does It Exist? 21:10 Challenges in Voice AI for Sales 23:31 Implementing AI in Customer Support 25:37 Customer Reactions to AI 29:32 Future Predictions for AI in Customer Service 37:45 Getting Started with AI in Customer Service 📌 Watch now and subscribe for more smart, fast, founder-first interviews. 🎧 Listen on the go: Spotify: https://spoti.fi/4m2K7ah Apple: https://apple.co/4cZ8RMm My other YouTube Channel: https://www.youtube.com/@GJarrosson