One-Liner

Project intro

One-Liner is a one-page web app that turns long, over-the-top posts into a single line saying what they actually mean. It was built for the author's friend, who was starting to use social media for job searches and found that most professional-feed posts took a few hundred words to say one thing — 'I got a new job' or 'please comment so I can message you'. Paste the post, and you get the gist in one line, then read the full post only when it is worth the time. The whole app is a single index.html file with no backend, no server and no API key: open-weight models run entirely inside the browser with WebLLM over WebGPU, defaulting to Qwen 2.5 1.5B Instruct with Llama 3.2 1B and 3B in a dropdown for comparison. A system prompt asks for exactly one sentence under 20 words, paired with four worked examples, and a small cleanup function keeps only the first line and first sentence, stripping labels or quote marks the model sometimes adds. Resilience is part of the design: the AI library loads from three CDNs in turn so one blocked source does not break the page, and a Check connection button tests each download step after the author's first version failed silently inside a preview pane. It was entered in the Hacktoberfest Weekend Challenge: Build for a Friend. Honest limits, as the author states them: a small on-device model is less accurate than a big hosted one, it can miss sarcasm, the first load is a large model download, and it needs a WebGPU-capable browser; and the author credits 'Claude' (not specifically 'Claude Code') as a coding assistant for scaffolding and debugging.