Writing & Thoughts

Essays on AI, Engineering, and Software Architecture.

The Smart Pruning Paradox

Why the best-behaved compression method produces the worst-behaved model. An interactive walkthrough of weight pruning, where Wanda's score quietly removes the circuits responsible for safety.

How an LLM actually thinks

An interactive walkthrough of what happens inside a transformer when you send it a prompt, and three things the mechanics alone do not predict.

Building a byte-level parser on Tiny Tapeout: a design walkthrough

A JSON-grammar coprocessor that fits in two Tiny Tapeout tiles, plus three transferable design patterns and the parts of the TT submission flow that aren't well documented.

A Look at Google’s Agent-to-Agent (A2A) Protocol for Multi-Agent AI Systems

Deep dive into Google's A2A protocol specifications and potential.

Case Study: Building a Multi-Agent Research Assistant with A2A Protocol

Exploring the implementation of multi-agent systems using the A2A protocol.

Model Context Protocol (MCP) in Practice

A hands-on tutorial building a MacOS file assistant with MCP.

Fine-Tuning Mistral 7B on Apple Silicon

Guide to fine-tuning large language models on local Mac hardware.

Building and Running CodeLlama Locally on Your Mac

Step-by-step guide to running CodeLlama locally.

Rust Playground: A Fun Collaborative Way to Learn Rust

Exploring Rust through a collaborative playground.

Can ChatGPT Write Useable Machine Learning Models?

Evaluating the capability of ChatGPT in generating functional ML code.

Build a Neural Network from Scratch Using PyTorch

Fundamental tutorial on building neural networks.

PyTorch for Dummies

Beginner friendly introduction to PyTorch.

French Conjugations using Python

Automating french verb conjugations with Python.