Reasoning Search: When 'Instant' Isn't Enough
product
Two modes, one memory. Here's how SNAP's Thinking mode chains steps across your snaps for multi-hop questions Instant lookup can't answer.
Most AI search returns the first plausible passage. Reasoning Search plans, checks, and chains — across the sources you've saved.
Instant vs Thinking
Instant is a single-hop retrieval: ask, get the best-matching passage with citations. Ideal for lookups like "what did that PDF say about pricing tiers?"
Thinking decomposes the question, retrieves across multiple snaps, and reasons across them. Ideal for "compare how these three vendors handle onboarding based on the pages I saved last week."
What Thinking mode actually does
- Breaks a question into sub-queries
- Retrieves evidence for each from your snaps
- Cross-checks conflicting sources
- Assembles an answer with citations at every step
When to reach for it
- Comparative questions across several sources
- Timeline reconstruction from mixed notes and screenshots
- Literature review across a stack of PDFs
- Competitive synthesis from clipped pages
Still grounded
Thinking mode doesn't invent facts. If your snaps don't support a claim, the answer says so. See how grounded generation works in SNAP's source layer.