Learning is anything but clean
Learning something new is messy. It's supposed to be. The journey to solving a problem is a series of plateaus, deviations from the original problem, and unexpected tangents — resulting in a messy solution and, ultimately, learning things you hadn't actually set out to learn.
So when you face a problem with LLMs, you get an immediate, articulated answer with no friction. That's not a good thing.
Craft a learning path
One answer is to not use LLMs at all. Coursera, Udemy — plenty of conventional online platforms. University being the default route. I also believe books (anything on printed paper) are the best way to learn, though that may be a biased belief!
The other is to prompt a journey that enables a dirty learning path. Use conventional methods to solve a problem, with LLMs hinting toward an answer — which ironically requires outside information or more prompts.
Using LLMs is a huge problem for learning
This technology is here to stay, so it's worth learning. Context engineering is a thing, as is using it agentically with skill markdown files, etc.
But people are literally atrophying their brains. It's like driving everywhere instead of walking — your legs weaken.
Cognitive debt is a thing. Without the struggle, you aren't using your brain properly.
...LLM users also struggled to accurately quote their own work. While LLMs offer immediate convenience, our findings highlight potential cognitive costs. Over four months, LLM users consistently underperformed at neural, linguistic, and behavioral levels. These results raise concerns about the long-term educational implications of LLM reliance and underscore the need for deeper inquiry into AI's role in learning.
You can read the full article here.
How I prompt for learning
Don't give me the answer, I am learning how to do this
I'm not even joking.
My preference is to use LLMs for things I already know — boilerplate code or maintenance on this very website, for example.
Printed content is still superior for me when it comes to learning. Unlike LLMs, you can't get a direct answer. You might have to look for a specific page, read another book, or look for different sources of information. Again, that exploration makes learning messy, therefore more productive. It also supports this "Shadow Study Technique" which boosts retention and active recall with whatever it is you learn.
What I think this means long-term
With most things, I try not to slip into a neurotic take with LLMs. The technology is amazing, but incredibly dangerous.
Using it to learn is not a good idea. It's like eating takeaway every day when you know home-cooked food with vegetables is better long term for your health.
It's very clear to me: the messy path to finding an answer to your problem is best. You need to struggle.
Moving forward, I'm creating hard boundaries when it comes to learning:
- Use books first, if possible/relevant
- Look online - Google, Youtube
- Courses
- Prompts that don't give me a direct solution
- If using LLMs, breakdown the answer
- Write about the journey: Internalise to memory the learning process