Latest Interviews
Showing 1–2 of 2 transcripts.
Clear all filters- Y Combinator38 min
2024: The Year the GPT Wrapper Myth Proved Wrong
Jared Harge, Diana, Harj, Lily Yang, Cheng Cheng, Suk Peng, Yiu, Francesc Campoy Flores, Priyanka Vergadia, Anan, Mark Mandelbaum, Melanie Warrick, Mark Mirchandani, Gary Miles, Leslie Kendrick Magnuson, Harjit
The 2024 startup landscape shifted toward capital-efficient growth, where companies like Opus Clip and Perplexity achieved tens of millions in revenue with under $5M in funding by leveraging open-source models and vertical-specific applications. Enterprise adoption accelerated as AI agents attained enterprise-scale reliability, driving a record-breaking aggregate weekly growth rate of 10% for YC batches while converting pilots to revenue at unprecedented speeds. This ecosystem revival was fueled by regulatory relief, a resurgence of in-person Silicon Valley activity, and a strategic pivot from model monopoly to multi-model orchestration that prioritized product execution over raw compute ownership.
- Jane Street1h 6m
The Uncertain Art of Accelerating ML Models with Sylvain Gugger
Sylvain Gugger, Ron Minsky, Jeremy Howard, Mark Mandelmann, Mark Mirchandani, Francesc Campoy, Gabriel Sanchez
Former fast.ai co-author Jeremy Howard discusses his transition from mathematics education to optimizing machine learning infrastructure at Jane Street, highlighting breakthroughs in learning rate schedules and image resizing that previously secured top benchmark placements. He details the development of the Hugging Face Accelerate library, a lightweight tool designed to abstract complex hardware parallelism and eliminate boilerplate code for training across diverse GPUs and TPUs. The discussion further explores the architectural constraints of financial data, the dominance of PyTorch's iterative execution model, and Jane Street's rigorous approach to reproducibility and custom model development for high-frequency trading.