← How it works
7.5 · DUAL RETRIEVAL
Fast when it should be fast, deep when it must be deep.
Inspired by human thinking: fast reflex for familiar tasks, slow reasoning for hard ones. Plus 5-layer context to save tokens while staying accurate.
SYSTEM 1
⚡ Fast reflex
Instant retrieval for familiar queries — near-zero tokens, answers right away.
SYSTEM 2
🧠 Deep thinking
Multi-step reasoning for hard work — only triggered when needed, no waste.
// 5-layer context
L1 · Current query
L2 · Session
L3 · Active persona
L4 · Knowledge graph
L5 · NOUS long-term memory
~0
tokens saved vs stuffing all context
Answers that are accurate, fast and cheap.
You get precise answers without burning tokens.