Training Data

podcast@sequoiacap.com

For verified host and producer emails, sign up to view.

How Cursor Trained Composer on Fireworks: Distributed Infrastructure for High-Performance RL

May 26, 2026

Cursor's Federico Cassano and Fireworks' Dmytro Dzhulgakov explain how they collaborated to build Composer as a specialized foundation model. The core insight: models have finite capacity in their weights, and allocating all those bits to the singular task of software engineering in Cursor frees the model to be both better at the task and far more efficient at inference. Rather than start from pre-training and work up, they took an unconventional top-down approach — mid-training and RL on top...

Rebuilding IT From the Ground Up for the AI Age: Serval's Jake Stauch

May 19, 2026

Jake Stauch, founder and CEO of Serval, is building a ServiceNow for the AI era. His most contrarian bet is that the product should look like boring old enterprise software, but with unlimited intelligence. Serval's architecture splits work between two agents: an admin agent that uses code generation to spin up workflows from natural language, and a help desk agent that can only act through the tools admins explicitly approve. Jake explains why his team uses OpenAI models for end-user interac...

Suno's Mikey Shulman: Everyone Can Make Music Now

May 13, 2026

Most music platforms assume you're a listener. On Suno, 90% of daily users make something. Founder and CEO Mikey Shulman explains why that flips the  model: the act of creating IS the entertainment, with closer parallels to gaming and Claude Code than to Spotify. He breaks down the technical bets that got them here — modeling raw sound waves instead of encoding music theory, choosing autoregression over diffusion to prioritize full songs over crisp clips, and why music isn't a scale problem t...