Postgres plans without the panic

Query plans intimidate newcomers because they look like noise. We slow the room down: identify the node that dominates cost, then ask whether the predicate matches an index, then check row estimates. Participants bring anonymised plans from work when they can; when they cannot, we use institute datasets sized to fail in instructive ways. The goal is not memorising every node type — it is knowing when to stop tuning and when to reshape the query. We also discuss when ORMs fight you, and how to document raw SQL exceptions responsibly. That balance keeps Prisma useful without hiding the database.