RSSAmplifier

Blog

Where will you go next?

John Larkin's personal coding blog and thought journal. Exploring various ideas in software engineering, math, data, and more in my (limited) free time.

johnlarkin1.github.ioRSS feed ↗10 posts

Latest posts

Vanity Mirror

Context

Multi Armed Bandit

<!– This was meant to be for a take-home... I submitted some version of the first draft, but then couldn't stop and here we are. If parts trail off, it's because I shouldn't have even gone this deep into other more pressing matter

iMessage Data Foundry

🐙 Source GitHub 🐍 Python PyPI

larkin-mcp

You can either interact with mine, or clone the repo here , and get started with the second one. 🐍 Python PyPI 📘 TypeScript npm 🦀 Rust crates.io Check out any of the links above for the various published packages. Note, Claude did the css here. 📋 yourname-mcp template → I’m working on a much bigger project, but honestly, needed to take a break from that. It has been a grind. I have burned many…

Understanding Muon

So while I tried to mainly focus on optimizers, this post kinda splayed out some. It was my first time trying Pyodide and incorporating that logic into my blog. It was my first time using manim, which was exciting because I'm a big fan of the 3Blue1Brown channel. I also introduced quizzes (see AdamW section) for more interactivity. All of this is open source though, so if you have any questions,…

Disjoint Set Union

Recently, I did an interview. I got absolutely flamed and one of the reasons was I wasn’t familiar with a Disjoint Set Union (and I certainly couldn’t complete the C++ interview in time building this data structure naturally).

Teaching a Computer How to Write

✍️ Motivating Visualizations

Launching Scrollz

I’m happy to announce that as of September 16th, 2025, our application Scrollz App is officially in the iOS app store (and the Android store). With that, it’s out to 175 countries.

Tennis Scorigami

This post is going to be focused on discussing how we built our Tennis Scorigami project from a technical standpoint. I’ll discuss the current architecture, some of the design decisions I made, and where I want the project to go next.

Walk in the Parquet

At Mojo, we use Parquet files to store some of our simulation data. I - however - have been increasingly frustrated by the lack of support on MacOS to natively view them. They are (normally) compressed through a snappy algorithm, and Apple doesn’t have a native application to open them.