Notes from the field
Working notes on shipping AI and data systems in production at AU and NZ organisations. Mostly opinions, calibrated by what's broken. No listicles, no hot takes, just the things we'd tell a fellow engineer over coffee.
When pgvector is enough (and when it isn't)
A working engineer's case for Postgres + pgvector as the default vector store, plus the honest signals that mean you've outgrown it.
A pragmatic AI governance baseline for ANZ teams
What a small or mid-market AU/NZ organisation actually needs in place before shipping an LLM feature, without the box-ticking energy.
Your first agentic workflow should be boring
Most teams' first agent project fails because they pick something demo-friendly. The case for narrow, dull problems with measurable success.
Data mesh isn't a tool, it's an org chart
Most data mesh projects fail because the org doesn't actually want domain ownership. An honest read on the four principles and a smaller alternative.