I let six months of Gmail job alerts pile up, got curious, and turned them into a 5,598-job dataset with a zero-cost Python pipeline. Here's the archiving craft and what the corpus said about Australian ML hiring.