Blog
The Start
Why this blog exists — the real origin of Atmos Football as an AI stress test on private football data, plus the honest reasons behind building the blog itself.
The Data
How the Atmos Football dataset went from chaotic OneNote scribbles (2015) to a clean, historic master table — and why the early AI tests kept failing until Claude finally cracked it in 2026.
The Elo Problem: When Fair Teams Don't Feel Fair
The team generator says the teams are balanced. The players say they're not. Then the game happens and the numbers were right all along. How Elo ratings work in casual 5-a-side football, and why trusting them is harder than understanding them.
Building with AI: How Claude Helped Write This App
An honest account of using Claude as a development partner for a real side project — what works, what doesn't, and when it gets in the way.
From 60k Reads a Day to Under 1k: Cutting Firestore Costs on the Free Tier
How a casual football app was burning through Firestore's free tier quota and the techniques used to reduce reads by over 95%.
The Two-Environment Workflow: How We Actually Build Features with AI
Planning in one environment, implementing in another. After 40 versions and several months of building an app with AI, here is the system that made it work — and why separating thinking from doing turned out to be the most important decision.
What 40 Versions Taught Me About AI-Assisted Development
After several months and roughly 40 releases built with AI assistance, here is what actually worked, what did not, and what I would tell someone starting the same journey today.
Push Notifications for a PWA — FCM on Web and Android with Capacitor
How we added weekly match notifications to a React PWA and Android app using Firebase Cloud Messaging — covering web push, Capacitor native, and the permission flow.