I have been pondering over an idea for quite sometime now. What if you could have a AI browser sidekick - let's call it Luna as a native extension to your browser. Since AI is already multimodal, what if this extension can see what you are looking at, read what the browser is showing.
You could interact with it based on text/audio. Just like how you would interact with a sidekick. You tell it her the job, how you do it. maybe show her how it's done once and she has the ability to take action on behalf of you. This would be real life Jarvis for the web
Some use cases I could think of:
fill basic ass forms for you
Sign into platforms on your behalf
scrape data from internet once you show her how to do it
It would learn from your screen and you can automate workflows.
The thing that separates her is that - she should feel like an extension of the browser, not some tool that is on top of it.
PS - I know some tools do some parts of it, but the friction is too high even to start using them. What do you think HN? Any other killer features you'd want in a browser AI sidekick?
