Articles

Why I write about both

At first sight this site covers two unrelated subjects, the way software projects are run and generative AI. I wrote both series because of one problem I keep running into.

I have worked in software since 1986, and I have seen good ideas lose their substance as they spread. Scrum is the case I know best. Many teams kept the daily meeting, the backlog and the sprint review, and dropped the project management that Scrum expected the team to take on. Planning stopped at the end of the next sprint. Domain-Driven Design later went through something similar, its vocabulary spreading much faster than the practice behind it.

Generative AI is going the same way, and faster. Plenty of companies have run a pilot. Far fewer have taken one into production after measuring what it changed.

The delivery articles describe what usually gets lost along the way. Someone has to plan beyond the sprint and own the architecture. The work has to be split between teams along sensible lines, and a regulated product has to build its evidence as it goes. The AI articles start by explaining what a language model does, in enough detail to judge a vendor's claims. They then look at what an AI project needs before a company can rely on it.

The two series meet in Claude + BMAD. It follows a product, part of which is a medical device, developed by one engineer working with an AI, under the process for the highest safety class of medical device software. The product is not finished. It has passed a critical gate, with the evidence a regulator would expect at that stage, and the article shows how Scrum, Domain-Driven Design and a quality system made that possible with an AI writing the code.

That is the thread running through all the articles. AI has changed how quickly software gets written. Whether anyone can rely on that software still depends on the disciplines the other articles describe, and that is the work I do as an interim CTO.

The articles are long because the subjects need room. You can read them in any order and stop whenever you like. Most chapters take a few minutes and make sense on their own.

Where to start Start from your question Questions clients and colleagues have asked me, each linked to the chapter that answers it.