YC Bench - Predicting Top Y Combinator Startups

​Startup investing is still largely driven by human intuition and personal relationships, but what if we could replace VCs with AI agents ?

​ This 30-minute webinar introduces YC Bench, a live benchmark designed to test and rank machine learning models on their ability to predict short-term success within Y Combinator batches.

​Y Combinator offers a uniquely structured environment: hundreds of startups per batch, extreme selectivity (<1% acceptance rate), and a clear evaluation milestone at Demo Day.

​YC Bench turns this into a high-frequency, real-world testbed for applied ML—where researchers can iterate every 3 months instead of waiting years for startup outcomes like IPOs or acquisitions.

​ What you’ll learn:

• ​How YC Bench is redefining startup prediction as a measurable ML task

• ​The dataset, evaluation framework, and prediction targets

• ​Why short-term outperformance is a tractable and meaningful signal

• ​How to participate and submit your own models

​Why attend?

​If you’re a data scientist, ML researcher, or engineer interested in real-world, high-stakes prediction problems, YC Bench offers a rare opportunity to:

• ​Benchmark your models on a dynamic, economically meaningful task

• ​Collaborate with top researchers

• ​Get involved in the future of AI-driven investing

​Top contributors will be invited to present their work and explore further opportunities.

​👉 Learn more: https://ycbench.com

​👉 Paper: https://arxiv.org/abs/2604.02378

​Join us to explore how machine learning can move startup investing from intuition to infrastructure.