There are two ways to use Anchor: following the queue (Study) and pulling up a specific card (Browse). Buried in settings, the second would never get used, so it is a tab of its own.
Search
It searches ids, titles, deck names and body text, and orders hits by strength (id → title → deck → body).
- Search is entirely local. GitHub’s search API is never called: the synced vault is already on the device, so there is no reason to go out, and it works offline
- Dash characters (
−–—) are normalised before matching, so characters that merely look alike do not hide a card
Reading a card
The card screen shows everything at once: front, back, hint and note. References in the back are already expanded, and you can jump to the referenced card.
On a card with a rubric, the keywords are listed and nothing is matched.
A single .md file can also be read as a whole (in v1 a file holds several cards). </> switches to the raw source.
Nothing here is graded
If merely looking something up moved the review log, the scheduler’s input would be polluted. The log is the only record of your history, so reading stays reading.
Handing cards to an AI agent
- Copy on a card: you get the ID only (why: having your cards written)
- “Copy weak cards for AI”: up to ten cards, most-forgotten first, with text, source paths and ids (iOS / Mac)
History
A summary of everything you have graded — counts and breakdown — is available on iOS and Mac. On Android, Browse covers search and reading.