Y Combinator Startup Podcast

François Chollet, AI researcher behind Keras and founder of ARC Prize and Ndea, explores a radically different route to AGI. He gets into symbolic programs over neural nets. He talks about why coding agents are suddenly thriving, where deep learning may hit limits, and how ARC V3 tests learning, planning, and adaptation in new environments.

Gaurab Chakrabarti, Solugen CEO and former cancer researcher, joins Sean Hunt, Solugen CTO and chemical process inventor, for a look at turning biology plus chemistry into cleaner manufacturing. They talk about the cancer research spark, a scrappy PVC reactor, selling tiny batches to early buyers, scaling in Houston, and chasing a new future for American industry.

Mukund Jha, Emergent co-founder and former Dunzo founder, joins Madhav Jha, Emergent co-founder and ex deep learning leader at Amazon. They dive into turning AI testing into coding agents. They talk about building for non-technical creators, live app creation, lean hiring in India, personalized software replacing SaaS, and helping niche builders launch real products worldwide.

Max Hodak, Neuralink co-founder and founder of neurotech company Science, explores brain-computer interfaces that could restore sight. He talks about tiny retinal implants, how the brain adapts to new signals, what artificial vision feels like, and why neural engineering could transform medicine. He also gets into biohybrid implants, longevity, and brain-to-brain possibilities.

Raphael Schaad, founder of Cron (now Notion Calendar) and Y Combinator visiting partner, offers product and design expertise. He discusses how AI and no-code tools spread repetitive design trends. Short critiques cover distracting animations, poor hierarchy, scroll-jacking, and loss of brand originality. Practical focus: when to edit AI output and preserve clear product messaging.

Ian Fischer, co-founder and co-CEO of Poetiq and former DeepMind researcher, builds recursively self-improving AI reasoning harnesses. He explains how layering meta-systems on top of models can outperform fine-tuning. He describes dramatic benchmark gains, automating prompt engineering, and how small teams can achieve big improvements with code-based reasoning.

They explore the sudden rise of agent-driven tools like OpenClaw and Moldbook and the excitement around autonomous agents. They discuss agents acting without humans and a parallel agent economy. They cover how agents reshape developer tooling, docs optimized for agents, agent-native infrastructure like inboxes, and agent social networks and swarm behaviors.

Boris Cherny, engineer who built Claude Code at Anthropic, shares the origin story from a terminal prototype to a widely used AI coding tool. He talks about empowering workflows with CLI-first design, automating git/tests/DevOps, agent and subagent strategies for parallel debugging, and expanding form factors from terminal to GUI for nontechnical users.

A look at how tiny teams beat giants by baking automation into every workflow. Stories of AI teammates that let one person manage dozens of accounts. A unified internal source of truth that scales operations without hiring. Practical approaches for turning manual tasks into custom agents to keep companies lean and fast.

Peter Steinberger, software engineer and creator of OpenClaw, an open-source personal AI agent that runs locally and ties into messaging apps. He recounts the viral origin story and his key “aha” moment. Conversations cover bots delegating to humans, swarm intelligence versus centralized AI, local-first data ownership, and how personal agents might replace many traditional apps.