I Built YouTube's Algorithm - here's What I Got Wrong

Описание: Jing Conan Wang spent years at Google Brain optimizing YouTube's algorithm for engagement, and discovered it was making people miserable. We dig into why, and why the same problem is repeating with today's AI systems.

In this conversation, we explore:

The moment he realized optimization metrics ≠ user happiness
Why modern LLMs actually lost personalization compared to older recommendation systems
How RAG systems are missing the personalization layer entirely
The "context modeling" framework that could fix it
How high token costs change the entire go-to-market strategy for AI companies

This is essential viewing if you're building AI products, thinking about business models in the era of expensive compute, or wondering why your RAG implementation doesn't feel as personalized as YouTube recommendations did 10 years ago.

Jing is now building DeepVista, an AI assistant designed to replace executive assistants for founders. Sign up for beta: deepvista.ai

Perfect for: AI engineers, founders, product builders, data scientists wrestling with production AI challenges.

Resources: https://jingconan.com/2025/07/09/cont...
Jing's linkedin:   / jingconan