The AI revolution isn’t being held back by data or algorithms. It’s talent. Of the two million people working in machine learning, only a fraction can compete at the highest level. Just 600 are Kaggle Grandmasters — the global benchmark for elite AI engineering.

Neo claims to be the first autonomous AI engineer that can operate at that level. Built by Saurabh and Gaurav Vij, Neo recently beat OpenAI on their own benchmark, outperforming them across 75 Kaggle competitions. This isn’t a future promise. It’s a measurable, functioning system trained to solve real-world machine learning problems better and faster than the best human teams.

In this episode, Mark and Jeremy Think on Paper with the founders of Neo about:

Why top-tier ML engineering is scarce, expensive, and often wasted on grunt work

What it means to automate judgment, not just workflows

How lessons from CERN and GPUs inspired Neo’s architecture

What happens when software can outperform your best talent, and learn on its own

If you care about AI infrastructure, automation, and what comes after the era of human-led engineering, this episode is essential.

Please enjoy the show. And share with a curious friend.

—
Links and resources

NEO: https://heyneo.so/
—

Chapters

00:00 Why There Are So Few Top AI Engineers
01:54 Meet the Founders Behind Neo
04:08 What CERN Taught Them About Building at Scale
06:51 How to Explain AI to People Who Don’t Work in AI
10:06 Misinformation, Education, and the Stakes of Understanding AI
12:16 What Makes Neo an Autonomous Machine Learning Engineer
16:17 Can AI Beat the Best in the World at Their Own Game?
23:48 The OpenAI Benchmark and Why It Matters
30:22 Hollywood, Dogecoin, and the Future of AI Culture
40:41 Could AI Learn to Want Something?

—

More AI: https://www.youtube.com/watch?v=vTQoVHVjd8Y&t=59s&ab_channel=WhereEmergingTechnologiesMeetHumanity

source