Searching
How search works
Search matches the words in your question. The same question always gives the same answer.
Word matching, not AI
AETHER-33 ranks memories by how well their words match your question, using a long-established search formula called BM25 (the same family of ranking that classic search engines use). Rare words count for more than common ones, and short exact matches rank above long rambling ones.
Why this choice
- Repeatable. The same memories and the same question always return the same results in the same order, on any computer, in any year.
- Explainable. Every result lists the words that matched and its score.
- No AI model, no network, no cost per search. Nothing changes when a model is updated.
Narrowing a search
- Space: each space is completely separate, like a folder. A search never crosses spaces.
- Tags: ask for tags and a memory must have all of them to match.
- Quarantined memories are never returned. You can ask to list them, but their text stays hidden unless you have the reviewer secret.
What it cannot do
It will not find a match that shares no words with your question: “car” does not find “automobile.” If you need meaning-based search, put a vector search in front and keep AETHER-33 as the trustworthy record of what was actually written.
AETHER-33 is a product over CAIN-42's governed-memory stack, not a new architecture moat.
Memory never carries authority: every response is annotated
grants_authority: false.
No third-party audit has been performed. Limits ·
FAQ · what is live now