Fine-tuning vs. prompting
When do you change the weights vs. change the prompt?
The idea inside
Same task via a better prompt vs. a few adapter weights (LoRA intuition).
After this lesson
You can explain when to prompt vs fine-tune (LoRA) and the cost trade-off.
Where it leads
Inside this lesson
That's the real lesson stage, paused. Claim your pass to operate it.
See how AI actually works, end to end.
This lesson is one stop on the full arc. Unlock all of it, and keep it for life.
What you get
- The 34-lesson main path, a finishable route from a word to agents
- Goal tracks for using AI at work and building AI features
- Boss labs that make you apply a whole act, not just recognize it
- Spaced recall that brings each idea back before you forget
- Course memory: every term defined, with links to where it first appears
- A shareable capability card when you finish the main path
- Lifetime access on every device, every future lesson included
Not videos to watch. You predict, operate the machine, then prove it. That is why it stays.
99 interactive lessons and challenges. No videos, no code.
Free launch pass: lifetime access, no card needed
New here? The first lessons are free to try. Start with lesson 0.1
What this lesson shows
Same task via a better prompt vs. a few adapter weights (LoRA intuition).
The question it opens with
When do you change the weights vs. change the prompt?
The walkthrough, in the lesson's own words
- Model too formal? Add a prompt, or fine-tune the weights. Open a new chat to see which one sticks.
- Reach for the prompt first; fine-tune only when you must.
- The prompt peeled off; the edited weights stayed. Only fine-tuning sticks.
- A prompt lives only in this chat's context window (from 4.2), a new chat starts blank, so it's gone. Fine-tuning changes the weights themselves, and the weights ARE the model, so they carry over.
- Prompting. A one-off is cheap and instant, just say what you want, no training, no cost.
- Fine-tuning. Baking the style into the weights means you never have to repeat the instruction again. In practice that means collecting hundreds of example conversations in the style you want and running extra training on them, which permanently nudges the weights.
- Prompt first; fine-tune only when the behavior has to stick.
- A prompt only enters the context window, so it's gone next chat. Fine-tuning recolors the weights themselves, and the weights ARE the model, so the change lasts.
- Your support bot keeps answering in the wrong format, even though you spelled the format out in the system prompt. A teammate says to fine-tune it. Right call?
- Usually not yet. A prompt lives in the context window, so first push it harder: add examples of the exact format, and if you need a hard guarantee, constrain the output to a schema. Reach for fine-tuning only when prompting can't reliably hold the behavior and you need it baked into the weights for every call.
Key takeaway
Compare cost and effect.
What you can do after this lesson
You can explain when to prompt vs fine-tune (LoRA) and the cost trade-off.
Prefer reading first? This lesson has a companion explainer: Fine-tuning vs prompting
This is the written summary. The lesson itself is interactive: you predict, drag and operate the mechanism above, and the reveal answers you.