AI answers have a strange quality. They sound equally sure whether they are right or wrong. A made-up detail arrives in the same calm, fluent sentence as a true one. For casual questions that is mostly harmless. For questions about your own work, where you are about to act on the answer or repeat it to a customer, it is a real problem.
So we built Stuart around one rule: every sentence in an answer has to point to where it came from.
What a source gives you
A source is not decoration. It changes what you can do with an answer.
- You can check it. The answer tells you which commit, note or call it came from, so you don’t have to take it on trust.
- You can judge it. A decision from last week’s call carries more weight than a note from two years ago. The source tells you which one you are looking at.
- You can say where it came from. “Per the call on Tuesday” is a much stronger thing to tell a colleague than “I think”.
The rules we follow
Every sentence points to a real source
Each line of an answer carries a small reference to the place it came from in your work: a commit, an agent session, a note, a call. If Stuart cannot find a real source for a sentence, that sentence is removed before you see it.
“Not sure” is a valid answer
Sometimes the honest answer is partial. If Stuart finds the commit that switched your billing database but nothing that explains why, it says so: “Not sure. I found the commit that switched the billing database. No note or call says why.” That is more useful than a confident guess, because now you know what to go and ask.
Sources stay one click away
Every answer lists its sources underneath, and the ones on your Mac open right where they live. Checking should take seconds, or nobody will do it.
A fluent wrong answer is worse than a slow right one.
What sources don’t solve
A source tells you where a statement came from. It does not guarantee the source itself is right or current. A note can be out of date. Someone on a call can be wrong. That is why a source shows when it was last updated, where Stuart knows it, and why the final judgement stays with you. The goal is not to make you trust answers more. It is to make checking them easy enough that you actually do.
Why it matters more for your own work
A general chatbot answers from the whole internet. Stuart answers from what is already on your Mac: your repos, your notes, your agent sessions, your calls. When the answer is about something you said, promised or decided, the source is the difference between “I think so” and “here it is”.
