"Helsinki has a yellow tram" matches all three tokens of "Hello".
MidBrain's lexical index stores keyed HMAC trigrams, not letters. The server returns candidates; the client decrypts and checks.
Different coordinates, same similarity.
MidBrain stores embeddings after an orthogonal transform derived from each agent's search key. Dot products survive it, so ranking doesn't change.
Every MidBrain agent gets four keys.
The one that decrypts conversations and documents never reaches our servers in client-managed mode. Search and consolidation get only the key they need.
Most failures in our DolphinBench run came after the facts reached the model.
89.1%: right tool, facts in context, response omitted a requirement.
7.9%: retrieval misses.
3.0%: expected tool never called.
83.2% on DolphinBench, across 600 tests.
Same agent and model as the leaderboard: Hermes with GPT 5.6 Luna. $64 in harness cost, plus 52k memory credits.
Production MidBrain, with two experimental prefetch features on.