Everyone Is Using RAG Wrong

Added: 2026-05-25

[00:00:00]Rag is the most overbuilt and under-thought technology right now. Let me ask you series of questions and help you re-evaluate on your decision of rag.

[00:00:08]First question, is your data small and it fits under the model's context window or let's say your data is under 100k tokens, then you don't need rag. Just take your data and insert it into model's context window. It's free, [music] it's fast and it doesn't have any infrastructure cost. Second question, does your data change too frequently? Let's say your data changes every day, every hour or every 12 hours, then you need rag. Third question, does your product needs to cite resources?

[00:00:34]Let's say you want to show your user where the data came from, then you need rag. I have made a detailed YouTube video on how rag works. Comment rag and I'll send you the direct video link.