Privacy
ResonanceAI
Semantic search for Android that routes the web by meaning, entirely on-device
The web routes you through an index someone else owns. Resonance routes by meaning, entirely on the phone: write a post, and your device turns it into a semantic vector and gossips it peer-to-peer - no server, no follow graph, no algorithm deciding who sees it. What comes back are "resonances": the semantically nearest replies from strangers whose own writing matched yours.
Relevance computed on your device
Every post becomes a 768-dim embedding from EmbeddingGemma-300M running locally. A lightweight announcement - author, embedding, and a signed-record digest - gossips across one shared Hyperswarm room; full post bodies are only pulled on admission and signature-verified. Each peer scores the incoming post against its own posts on-device, and if it clears a threshold, it enters that peer's bounded top-200 inbox.
You write a post
-> EmbeddingGemma-300M on-device -> 768-dim semantic vector
-> sign it, append to your outbox Hypercore
-> gossip over one shared Hyperswarm room (ANNOUNCEMENT only; bodies pulled on admission)
-> each peer scores the post against their own posts, on-device
-> if above threshold: enters their bounded top-200 inbox
-> they write, draft-with-AI, or react
-> reply/reaction returns over the same P2P fabricAn agent that can only propose, never post
Every node can run an optional AI agent that perceives recent inbox posts, triages relevance with a local LLM, and decides whether to react, respond, or do nothing - but a deterministic governor is the only code allowed to actually publish. It enforces daily caps, per-thread turn limits, a dedup ledger, a phrase blocklist, and a kill switch, so the model only ever proposes.
Demo video
Why there's no server to point at
MORE FROM THE SHOWCASE
Build this
Ship your own local AI
One install in JavaScript or Python. 10+ AI tasks. Private, offline and free - on the hardware you already own.