I answer first, then ask AI to check my thinking

While working through a Google Cloud quiz, I gave my AI assistant a specific job. I would read each question and choose my answer. Once I clicked, it could look at my selection, tell me whether it agreed, and explain why. We did that for ten questions in the same browser tab.

What an order can teach you about databases

One question asked me to distinguish Cloud SQL from BigQuery. I chose the answer about transactions versus large-scale analytics. The assistant agreed and gave me an example: recording a customer’s order fits Cloud SQL. Comparing sales across millions of orders fits BigQuery. The same orders appear in both examples. What changes is the job I need the database to do.

That example gives me a way to check my own understanding. If a store needs to change the shipping address on one open order, which system would I reach for? What if the store wants to know which regions accounted for last quarter’s growth? I can work through both questions even if I forget the textbook wording.

What agreement tells me

The assistant agreed with all ten selections. It could see what I clicked, but it couldn’t tell whether I had a solid reason, recognized a familiar phrase, or made a lucky guess. Next time I’d add one sentence before asking for feedback: “I chose this because…” That gives the reviewer something specific to challenge, even when my answer is right.

A prompt I’d use again

Wait until I make my choice and give my reason. Then tell me whether the choice and the reasoning hold up. Give me one concrete example. Change one detail and let me try the new case before you explain it.

I see room for this in troubleshooting, too. Before looking at an error log together, I could write down my leading explanation and what I expect the log to show. Then I could ask the assistant which observation would count against my explanation. I’d have a record of my judgment before seeing the evidence, and a specific reason to revise it afterward.