Recap and next steps

What you built

Over this course you built a set of habits, not only a set of queries:

• Read a query before trusting its output.
• Name what one row represents before writing or accepting SQL.
• Check every join for fan-out.
• Verify a headline number with a second, independent method.
• Keep a short list of anchor numbers and reconcile against it.
• Write a correction into the repository, not only into the next prompt.

Keep the workflow

The loop you practised is durable: you ask, and then you read what comes back, interrogate it, verify a headline number, and accept or reject it. Those four steps - read, interrogate, verify, accept or reject - are how experienced people work with SQL they did not write, whether the author is an agent or a colleague. The dataset here was small so you could check every total by hand. On a real warehouse the totals are too large to eyeball, which is exactly when the loop earns its keep.

The catch that matters most is the second, independent method: a wrong number that a single query returns confidently only shows itself when a different route disagrees. And a correction only holds for the next conversation if it lives in the repository - an AGENTS.md note or a quality check - rather than in a prompt you will not repeat.

If you want to keep practising at a bigger scale, the project README's "Working at a bigger scale" section lists three paths, from a 6-million-row generated benchmark to connecting your own warehouse.

The next course, Design the Model, turns the question around: instead of auditing a query someone handed you, you design the query and the model behind it, and you make the repository something an agent can reason about well enough to get more right on the first try.

Check your understanding

• Which habit catches a wrong number that a single query would miss?
• Where should a correction live so it holds for the next conversation?
• Name the four steps of the loop.

Do it with your agent

This is a concept lesson - nothing to build. Say next lesson and your agent walks you through it and checks you followed. Then say next lesson again to finish the course.

Next steps

• Design the Model - the intermediate course, shown on the programme overview.
• Bruin documentation - the reference for bruin query and every other command.
• Data quality checks - turn a caught mistake into a check that catches it next time.
• Join the Bruin community on Slack - ask questions about the material or your own data.