The One Skill That Survives The AI Shift
About this episode
Ofer Mendelevitch has been building with LLMs since GPT-1 in 2019, and he wrote O'Reilly's "Hands-On RAG for Production" while everyone was declaring RAG dead, a claim he says resurfaces "every two months, mostly people trying to market something new." In this full conversation he takes Angelina's real hallucination-checking use case and rebuilds it live: why the magical one-PDF demo collapses at
Key moments
- 0:00 Is RAG actually dead?
- 4:08 Why write a RAG book when everyone says RAG is dead?
- 6:24 Can you build a hallucination checker with RAG?
- 10:26 Why does AI read 85% and 95% as the same number?
- 14:13 Why do agents beat one-shot RAG on hard questions?
- 20:06 How do you search tables, images, and video?
- 28:05 When is a knowledge graph actually worth it?
- 33:09 Which part of a RAG system costs the most?
- 39:21 How do you grade AI when there's no answer key?
- 45:59 Should you build your own RAG system or buy one?
- 49:41 Is developer relations worth it for a startup?
- 1:01:17 Will AI replace engineers and data scientists?
- 1:06:22 What should you study if AI can already code?
- 1:09:04 outro