Model Specialization & Fine-Tuning: a16z Podcast — How OpenAI Builds for 800M Weekly Users
Update: 2025-11-28
Description
When an app reaches 10% of the planet, engineering becomes a social and strategic problem — and OpenAI’s approach to models reveals why. Condensed from 53 to 4 minutes, this summary with host Reid Hoffman and guest Sherman Wu explains how OpenAI balances ChatGPT’s first-party scale with a broad API-driven developer platform. Learn why the industry moved from “one model to rule them all” to model specialization, how reinforcement fine-tuning and context engineering (not just prompt hacks) unlock product-market fit, and when deterministic agent builders make more sense than free-form autonomy. Sherman breaks down pricing trade-offs (usage-based vs. outcome-based), operational separation across modalities, the role of RAG and retrieval, and why open-source models haven’t materially cannibalized incumbents. Ideal for product leaders, ML engineers, and founders exploring fine-tuning, RFT, agents, RAG, pricing, and scaling inference. Listen now to get the key ideas in minutes.
Comments
In Channel
Download from Google Play
Download from App Store
United States00:00
00:00
1.0x
