In-depth writing on software architecture, engineering practices, and the tools reshaping how we build — things I've learned, built, or figured out the hard way.
A second-hand BBC Micro, a dot matrix printer taught to print every line twice, and a Ninja Turtles game that never got past the intro screen. On why the how has a shelf life, the why doesn’t, and what that means for children growing up with the answers already on tap.
A three-year-old meeting his hero has no brakes at all. Twenty-five years in, the excitement is still there — it’s just buried under silt. On toil, the work that grows with everything you build, and why letting the machine take it is something to be cheerful about.
A brand-new fence lasted less than a day against three children under seven and a bag of pretend tools. On Hyrum’s Law, observable behaviour, and why the scarce skill is knowing which knots people are quietly leaning on before you knock one through.
An oven box that buries the one number you need under a temperature you already knew. On error messages, logs, dashboards, and why having information present isn’t the same as being able to find it when it matters.
A chip throttles before it melts. It reads its own temperature and acts before anything fails. On heatwaves, public transport disruption, proactive AI, and why the system that bends in full view of the people depending on it beats the one that snaps without warning.
One mathematician, seven bridges, and an idea that quietly runs most of your day. A tour of graph theory — vertices, edges, trees, the Travelling Salesman problem — and why it underpins GPS routing, dependency checks, and most of the software you used before lunch.
My five-year-old waved from the high rope because her confidence was earned on the low ropes first. A model’s confidence is earned nowhere. On AI, fluency, calibration research, and why the safety net matters more than the walk.
What a card trick and a broken departure board both know about failing well. On graceful degradation, circuit breakers, and why a fallback invisible to everyone isn’t resilience — it’s a slow leak with good manners.
ISO 27001 and ISO 9001, read as engineering documents rather than compliance paperwork. The security and quality controls are really the arguments senior engineers have spent years losing — written down in language the rest of the business has already agreed to take seriously.
Agentic BI is a real category, but the failure modes are architectural rather than technical. A field guide to the semantic layer, MCP, Skills, and the design choices that separate projects that ship from projects cancelled in 2027.
What ADRs are, why institutional memory matters in long-lived systems, and how to use them well. A practical guide to capturing architectural decisions, with worked examples and an honest account of where adoption fails.
AI is an amplifier. The research is unambiguous: the same tools produce opposite outcomes depending on the engineering system they land in. Here is the playbook for making sure yours bends the curve the right way.
On technical debt, normalised workarounds, and why the real problem isn’t shipping cold toast — it’s forgetting what hot toast was supposed to taste like.
The real job of a load balancer isn’t routing requests — it’s knowing when the service is cold. On health checks, managed infrastructure, and why the feedback loop is the architecture.
The scariest failure in an AI system is the one where everything looks fine. On polite failure, treating the model as a dependency, and why data gravity matters more than model choice.
A practical guide to the circuit breaker pattern in distributed systems — what it is, how to implement it with Resilience4j, the companion patterns that make it actually useful, and why fallback design is harder than it looks.
The first post was about the problem. This one is about what to do on Monday morning. Three frames, in a deliberate order: make the debt visible, make the trade-off honest, pick a rule about what comes in.
A Monday night, a rogue Lego brick, and a question every engineer has to answer: do you route around the thing under the sofa, or do you actually fix it?
Magic and software engineering are strangely similar disciplines. Both create the illusion of simplicity by hiding a lot of careful structure behind the scenes.
What my children accidentally taught me about prompt engineering. Ambiguous prompts, context, chain-of-thought reasoning, and why feedback shapes behaviour — whether the system is an AI model or a three-year-old.
Integrating NaPTAN, ATCO, GTFS, TransXChange, SIRI. Messy data, conflicting standards, domain-driven design in practice.
A practitioner's honest take on choosing architectural styles. When a monolith is the right call. When to introduce event-driven patterns. The real cost of distributed systems. Drawing on experience of using different styles for different products.