Learning through LLMs

Deepak Kumar — Sun, 15 Mar 2026 13:23:18 +0000

My intention this past week was to build a solid foundation in programming so that I am able to pick up learning to code and also understand when I read the code on what it actually does. I wanted to get familiar with terms used by software developers and be able to understand what they mean by that. I know very little python and I was unsure on how to start. I looked at some videos where they explained their approach, setup etc. Frankly it quickly became overwhelming enough for me to quit this pursuit.

This is where I decided I will leverage LLMs for my learning.

I have used LLMs extensively for quick learning of certain topics, getting up to speed on something. One of the advantages i felt it had over traditional learning is that you will be able to tailor is extensively according to your own learning style. The problem with learning with courses, tutorials or any other structured learning is that first, the learning path is not customized for your current knowledge and skill level. Either it is oversimplified or too complicated to catchup. All the knowledge was thrown to me all at once wherever I went.

One of the approaches which has always worked for me is if I am able to dissect the information on the go as I am learning.

I never got the opportunity to that in my previous learnings. Conversations with LLMs gives me the edge and I think this way of learning is way more efficient. The only skill here needed is to ask the question properly. If you know that, then you are able to extract the right knowledge in right context to be able to learn smoothly. It's about filtering, ordering and breaking down of information in a manner which adds positively to your knowledge curve of that subject.

I will explain the setup and then also show another benefit which i believe would be very important. I had LLMs and google collab on a split screen. I used LLM as my guiding light and something that I would like to refer if i dont understand something. i had LLM chart out my path, daily goal and my time which i should be spending on the problem solving. I tried to be not guided on the code creation, debugging as much as possible unless i really had no other way to move ahead. In that also i tried to do an active recall using another scenario so that it stays with me. What i realized is if you approach programming like creating a sentence understanding the connecting words and their meaning it's much easier to understand and recreate it.

Of course, there are bad days where you do not get it even with hints and that's okay. If you have not been able to hit a wall when solving a problem where you have to wonder your capability or took a walk to come back again unable to solve it, I don't think you have pushed yourself enough

I don't think I have discovered something new but in terms of individual development it is a win for me. One thing I would be looking to improve is to also start taking care of planning the task and do more things myself. You need to be able to do things which your area delegating to LLM tools otherwise you develop a skillset gap which can be a problem when servers are down, and you have work by yourself!!

I would be interested to know your experience as something starting off or even experienced people leveraging LLMs for the learning to code.

PS: I have used ChatGPT until now and I am using Claude for a change. It is already blowing my mind on the way it is different from ChatGPT and I will share my learnings on this soon

Week 1 with python

Deepak Kumar — Sat, 07 Mar 2026 18:39:42 +0000

I spent a week with python and here is my experience

I am learning python specifically for data analytics and science perspective, so I found that identifying your goal with learning this general-purpose language really helps you being on track.
I am using LLMs to assist me in learning this, so developing a roadmap with goals, guiding me through initial setup and choosing right tools was incredibly useful. I think this is where LLMs shine.
It also helped me in improving my prompting skills. I did not want it to completely feed me everything and allow me to challenge myself to think critically when retaining information, debugging a problem or even constructing instruction.
I am also leveling up in SQL so some relation between how queries are constructed has helped in developing some core understanding.

I am not sure if I have been slow, but I am not worried. I think I have good retention, and I have been consistent in getting more comfortable and more importantly I am able to think in those terms.

I am now comfortable in exploring data, identifying inconsistencies and develop basic metrics out of those dataframes and I am excited to see what is next.

Forem: Deepak Kumar

Learning through LLMs

Week 1 with python