A second brain app that answers with sources: what to look for
Want a second brain app that answers from your own notes? Look for a citation on each claim, a way to open the source, and a set-up you'll actually use.
On this page 6 sections
If you want a second brain app that can answer questions from your own notes, look for three things: a citation on the claim, a way to open that source in context, and a set-up you will actually ask. A notes library you only write into is a filing cabinet. A second brain you can query, and then check, is more useful.
The short meaning of the term is in the glossary entry for a second brain. For the habit of capturing and organising work notes, start with how to build a second brain for work. This post is about the "answers with sources" part: what to require of the app, and how to set one up so you trust what comes back. Download Plexii free for Mac or Windows if you want to try the steps on your own files as you read.
Why second brains become write-only
Most people do not fail at collecting. They fail at asking, because asking has been disappointing.
- Search returns a pile, not an answer. You still have to open six notes and rebuild the point yourself.
- An answer without sources is a guess. You cannot tell whether it came from last week's decision or from an old draft you meant to delete.
- Old notes go stale. A pricing note from last year can look as confident as this year's, unless you can see which file the sentence rests on.
The fix is not another capture inbox. It is an app that will show its working, and a habit of opening the citation before you use the answer.
What "answers with sources" should mean
Require this much, at minimum:
- The answer is drawn from your material, not from a general model with your filenames sprinkled on top. You should be able to see which notes, files or pages were read.
- A citation is attached to a claim, not only listed at the end. Clicking it should open the passage, not a folder.
- You can compare what was read with what was used. A trace or source list that includes unused material is useful: it shows the tool looked, and what it skipped.
- Asking is something you do on purpose. An assistant that writes into your notes unasked is harder to check.
A list of "related notes" is not the same thing. Related notes suggest where you might look. A citation says: this sentence came from here.
When you do use an answer, how to check an AI answer before you rely on it is the short routine: open at least one citation, read it in context, and only then paste, send or decide.
How a few common set-ups compare
Different tools solve different parts of the problem. These descriptions come from each product's own documentation, not from a ranking.
| Set-up | Where your notes live | How answers point to sources |
|---|---|---|
| Obsidian (opens in a new tab) | Plain Markdown files in a folder on your computer, called a vault | Search and links are built in; AI answering comes from community plugins, so it varies |
| NotebookLM (opens in a new tab) | Sources you upload into a notebook in your Google account | Answers come from the notebook's sources, with citations. Account limits apply; check Google's current help for how many sources a notebook can hold |
| Plexii | Your workspace on your own computer first: desks, documents, tables, files, notes, and mail and calendar if connected | Numbered citations on each claim, and a trace listing the sources the search returned |
Obsidian is strong when you want ordinary files you can still open in twenty years. NotebookLM is strong when the material is a defined set of documents, such as a reading list for one project. Neither is wrong. The question is which shape your second brain has, and whether you can open the source behind a sentence.
How Plexii answers from your workspace
In Plexii, the second brain is a place called PlexiiBrain. It is not a separate notes app. It is what the whole workspace knows, and the Plexii assistant answers from it. Your desks, documents, notes and files are brought in as entries and kept in step with the originals. Delete the original and its entry goes at the next refresh. See PlexiiBrain: your workspace knowledge.
There are three ways in.
Search, then ask. Search everything is one box over desks, documents, tables, work widgets, files, knowledge entries, calendar, meetings and email. Press Enter to open a result. Press ⌘ Enter (Ctrl Enter on Windows) to ask for a written answer: the Answer card gives a plain-language answer and, under Sources, the documents and entries it drew on. The results list stays underneath so you can keep checking. Asking is never automatic. See Search everything and ask for an answer.
Ask the Plexii assistant. For longer questions, the assistant shows a trace of what it did. Each search line lists the sources it returned, numbered. In the answer, a statement that rests on a source carries a number, and clicking it opens that source in a window over what you are doing. Chips under the answer list only the sources the answer actually cites, so you can compare what was read with what was used. See See what Plexii read and check its sources.
Write knowledge down on purpose. In Ask PlexiiBrain you can add entries for things that are not in any file: a refund policy, a rule of thumb, how a process works. Tag them and pin the essentials. You can also add whole folders of files under Add files & folders, and they are indexed so Plexii can read them. See Build your knowledge base in Ask PlexiiBrain.
Set up a second brain you will actually ask
- Organise by project first. Retrieval is easier when material for one piece of work sits together. A simple system for organising project files, notes and links covers a two-level structure that works with any tool.
- Put one real question through the app this week. Use a question you already know the answer to, so you can tell whether the citations are pointing at the right place.
- Open at least one citation every time. It takes ten seconds, and it is how you find out whether to keep trusting the tool. How to check an AI answer walks through what to look at.
- Flag decisions that can go stale. A second brain should notice when the ground moves. In PlexiiBrain you can flag the work widget a decision rests on, such as the budget table, and Plexii warns you when it or its desk changes. The Decisions page lists what has been decided and what a change would put at risk.
- Know the rules your tool follows. Ours are written down in show your sources, ask before you act.
For more on AI you can check, see the AI you can check hub. The glossary entry for a second brain has the short definition if you need to brief someone else. Download Plexii free if you want to run the set-up on your own workspace.
Common questions
What should I look for in a second brain app that answers questions? Look for answers drawn from your own notes, a citation on each important claim that opens the source, and a way to see what was searched as well as what was cited. If you cannot check the sentence, do not rely on it.
Can AI answer questions from my own notes? Yes. The useful versions read the notes that match your question and write an answer from them. The difference that matters is whether the app shows which notes it used, so you can open and check them.
What is the difference between a citation and a list of related notes? A citation is attached to a specific claim and opens the source behind it. A related-notes list shows what might be relevant, without telling you which part of the answer came from where.
Why does an AI answer sometimes miss a note I know exists? Often because the search behind it matched exact words, not meaning. A note that uses a different word from your question may not be found. Try the other word, or use a tool that also matches by meaning.
Do I need to use the PARA method? No. PARA is one way to organise material by how actionable it is. Any structure that keeps each project's material together will make searching and asking easier. See organise by project.
Is a second brain the same as a knowledge base? They overlap. A knowledge base is usually shared and written on purpose, such as policies and how-tos. A second brain is usually personal and includes everything you collect. In Plexii, PlexiiBrain can hold both: entries you write and the workspace you already have.