Why you might see "no results" when the data exists

Help

Why you might see "no results" when the data exists

Sometimes the assistant searches for the wrong thing. Here's how to spot it — and the ten-second fix.

Updated July 17, 2026

Why you might see "no results" when the data exists

Every so often you'll ask something you know has an answer — "what were my commits yesterday?" — and get back a confident "nothing found." Frustrating, and usually wrong in a very specific way: the data is there, but the search that ran didn't look where you meant.

This page shows you how to tell the difference in ten seconds, and how to get the right answer on the next try.

What actually happened

Before the assistant can search, it has to work out three things you didn't spell out: which account is yours, where to look, and what "yesterday" means. When a step goes wrong, it usually goes wrong silently — and the result is an empty answer delivered with confidence.

So "no results" means exactly one thing: the search that ran found nothing. It does not mean your data isn't there.

When the miss came from our side — the assistant guessed the wrong account, or used a stand-in name instead of a real one — that's our responsibility, and we treat it as a bug, not a fact of life. We catch most of these automatically now, and the rest is what this page is for.

Read the receipt

Every answer comes with a receipt: the reasoning trace, right there in the chat. Open it and check two things:

  • What it actually searched for. The trace shows the literal search. If you see an account name that isn't yours, or the wrong date range, you've found the miss.
  • How much work it did. One lookup that came back empty reads very differently from a chain of ten searches across your repositories. Both can end in "no results" — only one of them means the answer is trustworthy.

The fastest fix

Name things. Every word you make explicit is a step the assistant can't fumble:

  • Instead of "my commits yesterday" — try "commits by draftin1 in dialoguesai on July 15."
  • Instead of "did I get that email?" — try "emails from Sarah about the contract, this week."

More patterns like this: How to ask so the assistant finds it.

When there was more data than I could show

Long lists get shortened to keep answers readable. When that happens the answer says so — you'll see a note like "+10 more omitted." That's honest truncation, not missing data. Ask "show the rest" and the assistant will keep going.

What we do for you automatically

  • A search that reaches for a stand-in name (like "your-username") is stopped before it runs, and the assistant is told to find the real one first.
  • An empty search on a question about you makes the assistant confirm which account is yours before answering.
  • When we can tell an answer probably missed, a small note appears with the answer — one tap to retry, and a link back to this page.

Related