Show HN: Vibe Linking (vb.lk)

Vibe Linking is a new URL shortener that encodes a natural-language query into a short link and uses a lightweight model (Gemini 1.5 Flash) to infer the most relevant web destination at redirect time. Instead of storing a mapping to a specific long URL, vb.lk links contain the user's query and the service runs a quick model inference — often falling back to “I’m feeling lucky” results from Google or DuckDuckGo — to forward the visitor to a best-guess target (examples include searches, Wikipedia pages, YouTube videos, maps, or product searches). Users can report incorrect redirects to feedback@vb.lk to help improve evaluation and model behavior. For the AI/ML community this is an intriguing real-world use of small LLMs for low-latency routing decisions: it showcases semantic linking (natural-language anchors that resolve dynamically), hybrid retrieval via search engine heuristics, and an online feedback loop for iteratively refining redirect accuracy. Key technical implications include trade-offs between model size/latency (Gemini 1.5 Flash enables fast inference), reliability and security risks (misdirects or phishing concerns), and content drift: as targets change, the model’s best-guess behavior may diverge from intent. Vibe Linking is a lightweight experiment in replacing static URL mappings with semantic, model-driven routing — useful for sharing intentful links but requiring careful evaluation and safety controls.