I work in software and spend a lot of time at the intersection of technical depth and practical usefulness. I'm less interested in building things for their own sake and more interested in building things that actually solve a real problem for a real person.
Right now that means: a legal compliance plugin for data protection professionals, research into AI reasoning and where LLMs genuinely break down, a deep dive into graph databases, and free learning programs for people who want to actually get good at something.
My most finished product is DPOkit: a WordPress plugin built specifically for Data Protection Officers. GDPR compliance is genuinely complex, and most tools in the space are either cookie-banner generators dressed up as compliance solutions, or enterprise software that costs as much as a small car.
DPOkit sits in the middle: serious tooling for people who actually carry the DPO role, inside the platform many of them are already running. If you're dealing with data protection professionally, go check it out.
I've been deep in LLM research, the harder questions. How do these models actually reason? When are they faking it? How do you build production systems on top of them that hold up over time? Where are the genuine capability limits right now?
Running parallel to that: serious work with Neo4j and property graphs. The core insight driving it is that when your problem is fundamentally about relationships and connectivity, a graph database doesn't just perform better, it changes how you think about the problem. Compliance networks, knowledge graphs, dependency mapping are all domains where the relational model is fighting the data, and graph is the right fit.
I'm particularly interested in combining the two: graph-grounded knowledge as a way to give LLMs something to reason against, rather than having them confabulate in a vacuum.
The learning programs started because I couldn't find what I wanted: a structured, achievable path to genuine competency in something, that fits into a real life, and doesn't require paying anyone. One concept a day, 15 minutes, four weeks, done at the end.
Finance, digital organisation, personal branding, and building an LLM from scratch are live now. A lot more in the pipeline. The dispatches also include tools like DPOkit. Everything is free where it can be, a decision I made early and don't intend to revisit.
I have 3 beautiful daughters with my, in case you haven't guessed yet, Japanese wife. We hyphenated our name with the idea of keeping it for the kids, but a word to the wise, not all forms accept it, so it can be a little troublesome.
Perth is technically the most isolated major city on Earth, which gives you a different view on a lot of things. Including what it means to work in tech without the usual proximity to where the industry concentrates. It turns out most of that proximity is overrated.
If you want to talk, about the plugin, graph databases, AI, the dispatches, or something else, email me at adrian@nakagawabennett.com. I read it.