The evidence is the billions of people who copy text to and from ChatGPT to other web pages.

But not just other web pages, other apps too. So again, why a browser. Why ask firefoxs pop out expansion panel instead of a right click context menu in word, or a chat interface in copilots phat app or a website chat interface or the precious space under a google search or any of the other 10000 places people are inserting this shit.

And a copy paste task isnt necessarily going to be aided by a pop out sidebar running a local LLM chewing up already precious RAM. There's no guarantee its going to integrate correctly with the users chosen LLM provider.

Like we are looking at having LLM's inserted into almost every customer facing application. At some point, they will want a subscription for each of them or they are all going to need local resources. They are all going to have to be interoperable and run off the same account. Or you are going to have to have something that just works with the whole stack.

It doesn't make a lick of sense to try and preempt that situation, with mainline features pushed to all customers.

Googles approach, having a separate AI enabled browser makes the most sense. If it takes off its because of affirmative user consent and they can merge it into chrome. If it doesn't work they can silently discontinue it like so many other things.

	▲	brokencode an hour ago \| parent [-]
		Google has a separate browser for AI features? Gemini has just shown up in Chrome for me. It’s of course a mess and a mad rush for market share right now. That’s just a product of a healthy, competitive market. I agree that I’d rather have one AI service I pay for that integrates with all my apps. AI integration in apps is about being able to feed in the context from the app into the model, like the web page or document. It’s much nicer to just ask the model about what you’re looking at directly rather than having to copy in context. I don’t have market research on this, but I do believe customers will expect it. Someday hopefully the OS will allow apps to expose context and actions for a systemwide AI assistant. This is what Apple is trying to do with their Apple Intelligence for instance. If this works well, that’d be great.