Reverse Engineered AI

Reverse Engineered AI

A personal note on learning AI backwards, from LLM curiosity into the ML foundations behind modern systems.

I started as a software engineer six years ago. One of the important things about software engineering, especially for people trying to learn on their own, is that you get almost infinite attempts to make something work locally at no cost. That is part of what makes it beautiful. It stays open to anyone curious enough to start down the path.

I was always interested in AI, but from a distance. It was a minor curiosity: enough to stay interested, but not enough to make me seriously start learning. Then ChatGPT arrived in late 2022, and that curiosity multiplied overnight.

I began learning anything I could find about large language models. I watched Andrej Karpathy's videos and followed whatever helped me understand what was happening. But once I started doing actual project work, I realized I needed more than LLM intuition. I also needed the basic AI and machine learning concepts underneath these systems.

So in a way, I learned AI and ML in reverse order.

The series

This post is the first in a series called Reverse Engineered AI. In it, I plan to work through AI and ML topics from the foundations upward, eventually making my way back to LLMs.

This time I want to go in the proper order, but with the memory of having learned it the other way first. That may lead to a few useful observations along the way. I am excited to share those here.

Back to insights

Get in touch

The contact page is there if useful.

If a note here connects with something you are thinking through, the contact page is open for a short message with context.

A few lines are fine.

Go to contact