SQL for the age of AI

Atterna SQL: write it, read it, check it.

SELECT SUM(o.amount)
FROM orders o
JOIN items i ON i.order_id = o.id;

An AI assistant writes a query in seconds. It also writes queries that double-count after a join, drop people in a LEFT JOIN and average away a skew — and they all run. Atterna SQL trains the part it can’t do for you: knowing what a query really counts, and what the number means.

Ways to practise SQL at Atterna

• Start · SQL from zeroTen short days on a practice marketplace: from reading a table to JOIN and window functions. Days 1–2 need no account; all 10 days are free with one.
• Cases · SQL on real requestsA colleague asks something vague; the obvious query gives a convincing, wrong number. Case 1 is free, about 17 minutes.
• SQL interview questionsEight problems analysts get in interviews, answered right on the page with the reason behind every option.
• JOIN practiceTwo tables, ten rows each: how many after the join — and why the sales total doubled.
• AI-generated SQL: 6 silent mistakesQueries from ChatGPT, Claude or Copilot that run, look right and answer the wrong question — try each one.
• Check an AI’s SQLPaste an answer or a query from ChatGPT, Claude or Copilot and see the traps it leaves open. Runs in your browser.
• SQL level testSix questions, three minutes: where your SQL starts and which day to begin with.

What makes it different

• Real mess, not toy tables: cancelled orders, refunds and baskets split between several sellers.
• Every query is also run on a second dataset with different numbers, so a query that fits one by accident is caught.
• Each exercise asks what the number means for a decision, the part an AI cannot answer for you.

SQL terms, with the trap in each

• Primary key
• NULL
• GROUP BY
• WHERE vs HAVING
• INNER JOIN
• LEFT JOIN
• Join fan-out
• COUNT(DISTINCT …)
• Window function
• CTE
• Grain

FAQ