A product manager is asked for the write-up he did on the subscription economy last year. He searches for the phrase and gets three results, all passing mentions in recent meeting notes. The note he wants is called "Membership retention ideas", and the phrase does not appear anywhere in it.
He has 1,800 markdown notes from five years of writing, spread across 12 folders, with a tagging system he stopped maintaining two years ago. This case covers what happens when those notes go into a Consilience workspace, and how he finds his own writing without remembering which words he used at the time.
1,800 notes over five years, 12 folders, tags abandoned
He wrote most weeks for five years and wrote meetings up the same day they happened, so the same subjects come up repeatedly. Subscription models and repurchase cycles appear in four separate notes across three years. None of the four link to each other, and each carries a different set of tags.
- 1,800 markdown notes. Five years' worth, all .md files on a local disk
- 12 folders. "00_inbox" alone holds 400 of them, and the other 11 each sort by a different logic
- Roughly 300 book summaries, 200 conference and talk notes, 700 meeting notes, and the rest loose ideas
- About 40 tags. Half of them sit on exactly one note, and #ideas is on 600 notes, which makes it useless as a filter
- No title convention. "Membership retention ideas," "0312 meeting," and "look at this again" all live in the same folder
Five search terms, no result
The routine is to type whatever word comes to mind. "Subscription economy." "Subscriptions." "Subscription model." "SaaS." "Retention." Each search returns only the notes containing that exact word. The note he wants uses his own vocabulary instead, words like "repurchase cycle", "membership" and "churn", so it matches none of the five.
The next step is opening folders. The 400 notes in "00_inbox" have titles that give nothing away, so the only way to check "0312 meeting" is to open it. After about ten he stops, opens a blank document and writes the note again from memory. An unknown number of the 1,800 notes were created this way.
Drag all 12 folders in at once
Setup happens once. Notes are not rewritten and tags are not cleaned up. The original markdown files stay where they are.
Getting 1,800 notes in
⌘⇧F → "Extract sources"
Type "extract" into the "Search everything" palette and an "Extract sources" row appears under "Actions" (its shortcut chip reads ⌘⇧I). Enter launches the extraction wizard. Drag all 12 note folders in at once.
Leave it running and go do something else
At this scale it isn't a one-shot job. The wizard shows its progress step by step, and you can keep working in the app while it runs.
Palette search keeps working while extraction runs
The "Documents" section of ⌘⇧F matches keywords on your machine, so it answers instantly whatever extraction is doing. It is still keyword search, which is not what he needs here, but it stays available throughout.
When extraction finishes, the "Concepts" section fills in
Type "retention" into ⌘⇧F and a row now appears under "Concepts," labeled "Concept · mentioned in {count} notes." Before extraction, that section was empty. Enter opens the entity page.
Scan the whole thing once with ⌘⇧L
The "Entity table" lists every concept pulled out of five years of notes, so it shows which subjects recur. Subscriptions and retention turn out to be among the most frequent.
Ask using words that aren't in the title
"what did I write up about the subscription economy last year?"
→ "Documents" section: 3 hits. All meeting notes that spell out the phrase.
From here, ⌘↵ → "Ask your workspace". A new chat opens with the question
pre-filled; nothing is sent until you press send.
(same question, in chat)
→ The answer arrives with a "Where this came from in your notes" card.
2 glowing nodes = "Directly matched" (those meeting notes).
Matte nodes = "Related". One of them is [[Membership retention ideas]] (2025-03),
which never says "subscription economy" in its title or body.
Each connecting line is labeled with the relation's name.
"retention"
→ A [[Retention]] row under "Concepts". Enter opens the entity page, where
five years of notes mentioning the concept sit together in one place.
"Open in graph" jumps to the full 3D view.The first question works because concepts such as [[Subscription economy]] and [[Retention]] were extracted from the sentences inside "Membership retention ideas", and those concepts are connected. Asking about the subscription economy therefore returns that note as Related, even though the note never uses the phrase. He does not have to remember which words he chose five years ago.
The card also shows why each note appeared. "Directly matched" holds the notes that contained the query. "Related" holds the notes reached through connections, and every connecting line is labelled with the relation. Clicking a node opens its entity page. Anything worth keeping goes into a case, which reopens later with ⌘⇧K or by typing the case name into the palette.
What changes
| Before | Now |
|---|---|
| Searching "subscription economy" returns the 3 notes that spell out the phrase | Hand the same question to chat with ⌘↵ and "Membership retention ideas", which never uses the phrase, comes along as "Related" |
| Skim 400 titles in "00_inbox" and open ten of them | Type "retention" into ⌘⇧F, get one "Concepts" row, and Enter lands on an entity page where those notes are gathered |
| No way to tell why a note matched, so you open all of them anyway | The "Where this came from in your notes" card draws "Directly matched" apart from "Related" and labels every connecting line |
| Every request from the manager means running the same search from scratch | Collect it into a case and reopen it with ⌘⇧K or by searching the case's name |
Try this workflow with your documents
Learn the steps behind this workflow, then try them with your own sources.
Open the lesson ↗