Most product pages list what a tool can do. That is the easy direction. The harder, more useful direction for a trust focused product is to list what it will never do. Here is ours.
We will never answer outside the library
If the closed library does not have the material to support an answer, the product says so. We will not invent, infer from training data, or paraphrase a plausible guess. Silence is a better outcome than confident fiction for the person about to act on the answer.
We will never invent a citation
Every cited source has to exist. If we cannot name where an answer came from, we will not pretend. AI tools that generate footnotes that point nowhere are the single biggest reason owners have stopped trusting AI.
We will never sell your conversations
The questions owner operators ask are personal. Hiring, firing, cash flow, problem customers. None of that is data we will package and resell. Full stop.
We will never train models on your data
Your conversations stay yours. They are not training data for our future models or anyone else's. You can delete your history at any time.
We will never impersonate a professional
We will not pretend to be your attorney, your CPA, or your doctor. When a question needs one of those, we say so out loud and point you to the right kind of professional.
We will never pad an answer
Short and useful beats long and impressive. If the answer is a paragraph, you get a paragraph. We do not pad responses to look thorough.
Read more about how the product is built in how grounded retrieval works, or the philosophy behind it in why we built a closed wall.