Open Models Change The Economics of AI
Y Combinator · 商业与创投
In this episode of the Lightcone, Ollama Co-founder Jeffrey Morgan sits down with the hosts to talk about the massive shift toward open models in enterprise, the 150X growth in tokens since the start of the year driven by coding agents, and how Chinese models now dominate cloud token consumption. Apply to Y Combinator: https://www.ycombinator.com/apply Work at a startup: https://www.ycombinator.com/jobs 00:00 — Intro 01:16 — Who's Actually Using Open Models? 02:13 — Is It All About Cost? 03:31 — The Token Usage Explosion 05:31 — Fine Tuning: Hype Cycle or Here to Stay? 07:27 — Where Open Models Beat Claude 08:26 — Launching Models at Scale 11:31 — Olama as the OS Layer 13:58 — Hidden Layers Between Model and App 17:28 — Open vs. Closed: The Steady State 20:41 — Local vs. Cloud Models 23:12 — Why Chinese Models Dominate Cloud 24:00 — NVIDIA's Open Source Play 26:03 — The Return to Local 27:15 — Getting GPUs Is Hard 29:16 — The Flash Model Revolution 31:35 — God Model vs. Orchestration 33:37 — The Geopolitics Question 36:14 — From Docker to Ollama 37:12 — Applied to YC With the Wrong Idea 40:36 — Lost in the Wilderness for Two Years 42:15 — The Pivot That Changed Everything 44:49 — 100K GitHub Stars, No Revenue 47:40 — How Do You Monetize Open Source? 49:43 — Why Do YC as a Second-Time Founder? 51:49 — What Seeing "Good" Actually Does for You 53:38 — Old DevOps Rules That Don't Apply Anymore