Ziv needs your fingers, not just your eyeballs — feedback requested
Ziv is our hackathon submission: a haptic message channel. An open NVIDIA model writes the line; your Android phone plays it as vibro-braille you feel — no audio, no spoken reply, no scrolling. Every turn is framed by an attention vocabulary: an arrival cue, working ticks, Grade-1 braille cells, and a descending close so "message over" is felt, not guessed from silence. If the queue is busy, the next message waits politely instead of barging in; a full queue answers the sender with a loud 429. All of it is proven in tests and a live repro — JUDGE_REPRO.md maps every claim to its proof.
Try it first (no install, no key):
- The feel-tool — plays the timing vocabulary, name marks and any word you spell. Android Chrome vibrates; desktop/iOS show the rhythm visually: https://site-5505fea77ec84b4f8e1a5d70c82d5877.freebuff.page
- The 76-second demo — the whole loop on the phone: scheduled intent → model-written line → gate → queue → close: https://youtu.be/t9ovgEjm8i8
Two design questions we're actively deciding on — tell us:
- The close (200→140→90→50 ms buzzes): did it land as "this message is over," or just as quiet?
- Letter pacing (530–770 ms per character): bearable, dragging, or rushed?
Where to reply: questions and ideas → Discussions. A proper "Tried it" report → the structured issue form at https://github.com/razukc/ziv/issues/new/choose.
One asking rule: describe what you actually felt or saw — not what you predict "users" would feel. Nothing here is validated with deaf-blind readers yet; first-hand experience beats speculation, and it's the input we can actually act on. The repo is the submission artifact, not an active open-source project — so questions, not drive-by PRs. Thank you
Log in or sign up for Devpost to join the conversation.