See how the pros talk to AI.

Three chambers in one studio. The Library is a practice workbench — 212 UI problems, six engines render each prompt in parallel, you iterate toward the fewest-words-that-work. The Gallery is a community wall — every creation posted with the exact prompt and engine that made it. The Translator (coming soon) shows how Claude, GPT, Midjourney all read the same sentence differently. Learning, shipping, and sharing — the same craft from three angles.

Anything you made with an AI. Web pages, apps, images, videos, music, writing, code — eight categories, all open. The only rule: every submission must include the exact prompt you used and the engine that generated it. That's what makes a gallery post different from a regular social post — you're sharing the recipe, not just the dish.

Two signals, weighted together. Likeness — how close each engine’s rendering is to the target UI, judged pixel-level plus semantic. Brevity — how few tokens your prompt used. A ten-word prompt that hits 80% likeness often beats a fifty-word prompt that hits 95%, because craft is doing more with less. The scoring algorithm is open source — read it on GitHub, file an issue if you disagree with a call.

Yes. You retain full ownership of your creations and prompts. By posting, you grant Vibi the right to display your submission and let others copy or fork the prompt — that's the whole point of a community recipe book. You can delete your submission anytime and it disappears from the Gallery immediately. No wash-sale clause. Read the full terms.

Soon. The hard part is lighting up all six engines in parallel in a way that's honest (same prompt, same run conditions, real outputs — not mocked). We're targeting a preview for early members first. Join free and you'll be pinged the day it opens. In the meantime, the Library already does parallel six-engine rendering — just scoped to the 212 curated problems, not open prompts.

Because different engines hear different things. A prompt tuned for Claude might be quiet in GPT, loud in Kimi. The cross-engine view is what makes a prompt craft rather than a magic spell for one model. In practice, the prompts that score 90+ across all six are the ones that travel — you can reuse them in your own products, in your own code, and trust they’ll land.

By default, yes. Your practice prompts in the Library are stored only for your own logbook and stay private. Gallery submissions are public by design — that's the point. API keys never leave your browser — they’re passed directly to each engine’s endpoint, not routed through our server. You can read the network code, or just open devtools and watch.

Honest answer: possible. We’re a small team. But three safeguards — your logbook is exportable at any time as a JSON file with all your prompts, scores, and history. Your Gallery submissions are yours — exportable too. The problem set is open source — if we vanish, the community can fork it. And because you bring your own keys, you’ve never given us any billing hook or lock-in. Walk away whenever.

Absolutely. Use Vibi prompts in your own products, your client work, your startups. Everything you write is yours. We don't require attribution (though the community appreciates it). Tip jar / direct support for creators is on the roadmap — not in v1 but we're designing for it. For now: make good stuff, share the recipe, and grow the craft.