Advanced Vector Database Techniques That Power YouTube’s Home Screen

Date: 01/08/2026

Watch the Video

Liked This Week: Advanced Embeddings for Personalized Recommendations

Dive into vector embeddings beyond basic semantic search! In this Supabase tutorial, @dshukertjr demos three clever techniques—exponential moving averages, subtracting dislikes, and temporal analysis—to build dynamic, user-personalized content feeds like those on YouTube or Netflix. Perfect for no-code/low-code AI apps using pgvector.

Watch: Twitter thread/video
Full code: GitHub repo
Data source: The Movie DB API

Chapters:

• 00:00 Embeddings basics
• 01:25 Semantic search recap
• 02:53 Personalization logic
• 05:13 Exponential Moving Average
• 15:15 Subtracting dislikes
• 18:58 Temporal analysis

Next watches:

• Vector basics
• Supabase AI toolkit
• RAG with embeddings

Supabase: Open-source Firebase alt with Postgres + pgvector. Start free: supabase.com | Docs: supabase.com/docs | Subscribe: YouTube