Can iDBQuery be wrong, and how does it avoid making things up?

An honest answer: no AI is infallible, and iDBQuery can occasionally misinterpret a vaguely worded question or a genuinely ambiguous schema. What it is designed to *not* do is hallucinate figures — and that design is the whole point.

Here's why iDBQuery avoids making things up:

• It computes, it doesn't guess. For a quantitative question, iDBQuery writes real SQL and runs it against your live data. A number like "total revenue" is calculated from your rows, not improvised from something that sounds plausible.
• It shows its work. You can see the query it ran, so the reasoning is inspectable rather than hidden.
• Every figure is cited. Each value traces back to the exact source rows it came from, so you can click through and confirm it instead of taking it on faith.
• It grounds itself in your data. It understands your schema through introspection and a semantic layer, so answers are anchored to what's actually there.
• It can say it doesn't know. When the data can't support an answer, the goal is to surface that rather than invent one.

So the right mental model isn't "a chatbot that might be confidently wrong." It's a tool that does the analytical work transparently and hands you the receipts. If an answer ever looks off — usually because the question was ambiguous — the citations let you spot exactly where the interpretation diverged and rephrase, which is far safer than a black-box number you can't check.