Recall: turn saved material into a library you can question and check
Recall keeps articles, videos and your own notes in one library, with summaries and AI chat. It suits people who collect material and want to find the original evidence when they write.
Form your own approach first, then explore the reasoning to check trade-offs and validation you may have missed. There is no single right answer.
01
DESIGN QUESTION
How can source choices make the answer easier to understand?
You are designing a teaching tool for questions about saved material. A user saved last year's training material, then asks for “the company's current expense policy,” forgetting that this year's version is absent. The tool offers “library only,” “web only” and “both.” This exercise does not claim Recall already has additional features.
- The tool must not publish private material to the web automatically. Choosing a scope does not make every source reliable. You may change prompts before submission, source explanations in answers, and feedback when material is missing.
- Walk through three tasks: summarizing the old training material, finding a current public standard, and finding a new internal policy absent from the library. Explain which tasks need the user to supply material rather than broaden the search.
- Deliver interface sketches for the three states, sample wording and one observation task. Check whether users can explain which scope an answer used and what remains unverified.
Explore the reasoning
- Separate the need for newer material from whether a source can be public. Public search may not fill a gap in an internal company policy.
- Compare the costs of a default scope, explicit selection and task-specific reminders. Do not silently change scope just because a question says “current.”
- Show the sources actually used and distinguish “not found” from “does not exist.” When an internal policy is missing, ask the user to supply an authorized file.
- Ask observers to explain the scope and the gaps in their own words, then record misunderstandings. Selecting a recommended option alone is not proof of a successful design.