Notes as vectors: semantic search over everything you write.
2026-08-05
Keyword search only finds the words you remember. imembr takes a different route: when you save a note or a bookmark, its text is turned into an embedding — a high-dimensional vector that captures what the text is about rather than which exact words it uses. Search then compares vectors, so a query and a note can match on meaning even when they share no words at all.
What "embedded as a vector" means
An embedding model reads a piece of text and outputs a list of numbers. Texts about similar ideas land near each other in that vector space; unrelated texts land far apart. Distance in the space is a proxy for relatedness in meaning. imembr stores one of these vectors for every item you keep.
Why it helps in practice
- Search for "how do I cancel a subscription" and find the note you titled "unsubscribe flow" — no shared keyword required.
- Recall a bookmark by describing the idea instead of the page title you forgot.
- Phrase a query the way you think about it today, even if you wrote the note in different words months ago.
Keywords still work
Semantic search does not replace exact matching — it augments it. When you do remember the precise term, typing it still finds the item. The vector layer is what rescues the searches where you only half-remember what you were looking for.