[evolving-silica]

I'm not sure if I ever used any LLM features from Firefox besides exploring, but so far it seems great they're trying local approach instead of sending it to some "partner".

In the comment section here I see lot of people complaining about the fact it's enabled by default as well as some concerns about resource usage. Could someone experienced in desktop app architecture explain if disabling them functionalities makes Firefox that much faster or using less resouces? I'd assume that those functionalities are kind of loaded on demand?

"AI" became such a keyword that seem to instantly give either positive or negative response, it's also an advertised feature of every second app with many of them just forcing AI into you just because of hype. This doesn't seem to be a case in Firefox - so I highly disagree with the title - the features are there but they don't go into your way if you don't want them, therefore it's easy to just use it, only when needed

[goku12]

They should just provide an easy way to opt-in and disable the AI. That should be in their settings page (GUI) too - not in the about:configs page or in the prefs.js. That should be enough for people to stop feeling like they're losing control.

To be more transparent about it and foster further innovation, they should simply move the AI parts out of the browser altogether (to an external engine like ollama) and provide sane defaults, easy ways to launch them and options to try out different models. Such models and engines can perhaps be shared with other browsers or applications without duplication. Why must they instead be separately integrated so tightly into every single traditional application?