This is a great application of machine learning to improve in-context search! Rather than typing for a mere exact string, why can't you ask a question about a webpage, and have the search point you to the answer in the page?

?makers this is awesome. a few questions: * does it scan web page text semantically? * when does it actually start scanning the web page's text? * how much cpu load does this add? * will it run a separate copy on every webpage by default, or does it use a service worker?

Looks interesting!

This is a good idea, congrats with your launch! I will test it out.

Really interesting idea ,I like this vision. Altough shift-ctrl-F didn't work ,and always model loading, maybe cause by bad proxy.

Hey Product Hunt 👋 Shift-Ctrl-F is my attempt to "supercharge" ctrl-F for Chrome. Search the information available on a webpage using natural language (e.g. "what is the difference between test mode and live mode") instead of an exact string match (e.g. "difference", "test mode", and "live mode"). String match is a proxy heuristic for the true content you're looking to uncover -- in most cases it works very well, but in some cases it can be a bad proxy. Now that we have sophisticated question-answering models that can be run in Javascript (thanks to Google / TensorFlow), it's a great opportunity to experiment with new ways to search a webpage. Shift-Ctrl-F is open-source at https://github.com/model-zoo/shi.... The model is embedded in the extension and runs on your device, so it's safe to search webpages with private content.