Blog

Things I've figured out, written down

Mostly machine learning and software engineering, plus the occasional lesson I learned the annoying way.

10 articles9,148 total words

AI and ML

7 articles

The AI Crutch: What Happened When My Agent Mode Turned Off

The AI Crutch: What Happened When My Agent Mode Turned Off

A year into my first software developer job, I ran out of AI credits one afternoon and discovered I'd forgotten how to write code on my own. A shared story about a wake-up call.

Jul 21, 20263 min read
Attention Is All You Need: Understanding the Transformer Architecture

Attention Is All You Need: Understanding the Transformer Architecture

The Transformer model changed everything. From GPT to BERT to AlphaFold, virtually every state-of-the-art AI system today is built on one idea: self-attention. Here's how it works, from the math to the intuition.

Apr 10, 20268 min read
Convolutional Neural Networks: How Machines Learn to See

Convolutional Neural Networks: How Machines Learn to See

CNNs are the reason your phone can recognize faces, your car can read street signs, and medical software can detect cancer in X-rays. In this post, we break down the architecture that revolutionized computer vision — convolutions, pooling, and how it all fits together.

Mar 18, 20267 min read
Logistic Regression from Scratch: The Gateway to Classification

Logistic Regression from Scratch: The Gateway to Classification

After mastering Linear Regression, the next step is Logistic Regression — the workhorse of binary classification. We derive the math from first principles, implement it with NumPy, and understand why it's foundational to every neural network we've built.

Feb 12, 20266 min read
Mastering Linear Regression: From Scratch with Python and NumPy

Mastering Linear Regression: From Scratch with Python and NumPy

The maths behind Linear Regression, and how to implement it twice — once with Ordinary Least Squares, once with Gradient Descent. Formulas derived properly, models built in NumPy, nothing imported that does the thinking for you.

Jan 16, 20267 min read
Building a Neural Network from Scratch for MNIST Digit Classification

Building a Neural Network from Scratch for MNIST Digit Classification

Neural networks are one of the more beautiful things mathematics has produced, and the fastest way to understand one is to build it yourself. So I did: MNIST digit classification, from scratch, with nothing but Python and NumPy.

Apr 30, 20254 min read
Understanding the Core Concepts of Deep Learning with MIT 6.S191

Understanding the Core Concepts of Deep Learning with MIT 6.S191

Join me on a journey through the foundational principles of deep learning as taught in the MIT 6.S191 course. From perceptrons to regularization, we explore the mechanics that power neural networks today.

Apr 29, 20253 min read

Software Engineering

2 articles

General

1 articles

Humanity has come this far because of curiosity

© 2026 Breye Foka L. All rights reserved.