S2-E13 · Fine-Tuning, the Tool You Reach For Last (LoRA and QLoRA)

Your assistant is great, but the way it talks keeps slipping. You write the perfect prompt, paste in examples, and it's right about 90% of the time, and that last 10% makes it feel like a toy instead of your product. So you reach for fine-tuning, the most misunderstood tool in AI, and for most of what you're building, the very last thing you should reach for. In this episode you'll learn what fine-tuning actually does (it changes the model's weights, not its desk), the line that saves you weeks ("RAG is for knowledge, fine-tuning is for behavior"), and why force-feeding facts through fine-tuning makes a model hallucinate more, not know more. You'll learn how LoRA and QLoRA make it cheap enough to run on a single free GPU by training a tiny swappable adapter, the decision ladder (prompt, then RAG, then fine-tune), and the honest catches: you need a real dataset, a real eval, and you risk catastrophic forgetting. Episode 13 of Season 2 of How AI Works, a free hands-on course on building with AI. 📺 Full course playlist:    • How AI Works · Season 2: Build With AI   New episode every week. Subscribe to @HowAIWorksHQ to learn how to build with AI.