Archive
Writing
Long-form writing on engineering practice. Architecture with the rationale on the table, not just the result.
2026
26 AugThe Last Published Date
Two client websites built alongside each other, with opposite architectures, decided by one piece of evidence that was free, public, and sitting under a date stamp before either proposal was written.
Building LapDelta: A Real-Time F1 Replay for a Pound a Month
A 60fps replay of real Formula 1 telemetry that runs for about a pound a month with no request-serving backend, built on a zero-copy data contract shared across three languages and a fair amount of geometry.
The Second Reader
University taught me every practice that defines good engineering, then removed the one condition under which any of them make sense. What's left over is a lesson about knowledge itself.
Let It Stabilise First
Don't commit to a structural contract before you know what you're agreeing to. On deferring irreversible decisions until real usage has revealed the domain.
Testing to the Risk, Not the Checklist
Deliberate exclusions are engineering decisions. Coverage isn't a number; it's a resource allocation problem, and the resource is finite.
The Sprint Review Isn't the Feedback Loop
Formal ceremonies create the conditions for feedback. The actual signal comes from elsewhere. Sprint reviews tell you what people decided to say; informal corrections tell you what they actually noticed.
The Pipeline Is the Product
Spending a sprint on deployment infrastructure before writing any application code looks like deferring value. The evidence pointed the other way.
Make the Easy Thing the Right Thing
When you start copying infrastructure code for the third time, you've already waited too long. A note on extracting shared patterns into a library, and being kind to future-you.
Building a Mathematics Interpreter in F#: From Parser to Symbolic Calculus
How we built a full interpreter with symbolic differentiation, six number types, and interactive graph plotting, and what the architectural decisions reveal about building extensible systems.
Building an Event-Driven Health Tracker with Three Lambda Functions
How we built a health tracking platform with event-driven notifications, scheduled jobs, and auto-completing goals, and why decoupling what happens from when it happens made everything simpler.
Building WeatherWise: A Weather Platform That Tells You What to Do
Building a weather platform that transforms raw API data into prioritised, actionable recommendations, and the full-stack architecture that supports it.
Alive to Guess Again
Karl Popper argued that a theory which can't be proven wrong isn't really saying anything. The same is true of engineering practices: if you aren't actively trying to break them, you don't know whether they're working.
The Pragmatist's Razor
Cargo cult engineering is adopting practices without understanding. But there's an equal and opposite failure: the engineer so principled they forget they're building software for people, not for architecture diagrams.
Knowing How Isn't Knowing Why
I started code reviewing for a junior developer a few months ago. It taught me more about my own understanding than I expected, and revealed a model of learning that changes how I think about software knowledge.
Cargo Cult Software Engineering
Richard Feynman warned about scientists who follow the form of science without the substance. The same problem is everywhere in software: teams adopting practices they don't understand, hoping the results will follow.