I’m Ben Lvovsky. I build software, and I write here about the parts of the job that don’t fit in a pull request: architecture decisions that quietly become permanent, metrics that reward the wrong behaviour, and incidents that turn out to have been misdiagnosed for months.

Alongside that, I’m a PhD candidate in software engineering at RMIT University in Melbourne. My research is on multi-agent LLM systems for privacy and security in software delivery — roughly, whether a set of specialised language-model agents arguing with each other can find the sensitive data that leaks through tickets, logs, and repositories. That work is separate from these essays, but it shapes what I find interesting about them.

Everything here is written in the first person because all of it comes from somewhere specific: a team I worked on, a system I broke, a queue I mis-tuned. Where a story involves a client or a colleague, the details are changed enough that they aren’t identifiable.

Getting in touch

Every article has comments open at the bottom, backed by GitHub Discussions. Corrections are especially welcome — several of these pieces exist because somebody told me I was wrong about something.