what is internals-series fr?

This repository is the code companion to a blog post series called Internals, published at suryal.dev. The series explains how AI systems actually work under the hood, and this repo holds the runnable scripts that back up the claims made in each post. The author states that every claim in the writing is verified against a real model, so if a script reports that its numbers match, those numbers were produced by actually running a model at the time the post was written. The README is short and describes the project at a high level rather than going into deep detail. It lists the posts in the series alongside links to their matching code folders. So far there is one entry: a post called Inside the Forward Pass, with its code kept in the 01-forward-pass folder. The scripts run models locally using MLX, a machine learning framework built for Apple Silicon chips. This means the project is meant to be used on a Mac with an Apple Silicon processor, such as an M1, M2, M3, or M4 chip, rather than on other hardware. Setup is a single pip install command that pulls in mlx and mlx-lm. By default the scripts use a small model called Qwen3-0.6B in a 4-bit quantized form, which is about 350 megabytes to download, so trying the code does not require a large amount of disk space or a long wait. A command line option lets you swap in a larger model from the mlx-community collection on Hugging Face if you want to experiment further. Overall this is a small, focused reference project for readers following along with the Internals blog series who want to run the same experiments described in the posts themselves, rather than a general purpose tool or library.