Oh well . . .
Due to matters of ancient history (UI changes and plugin compatibility), it's literally impossible to run Firefox, since that would mean letting go of old grudges. No choice but to accept Google ramming its AI down our throat, I guess.
Mozilla has reiterated its opposition to Google's decision to build AI plumbing into its Chrome browser, though rather belatedly now that the technology, known as the Prompt API, is already being tested in Chrome and Microsoft Edge. Jake Archibald, Mozilla web developer relations lead, articulated the org’s concerns in a …
Wondering what "grudge" could possibly make Chrome more tolerable than Firefox. Work mandates, we understand, but what the hell about Firefox could possibly be worse than everything Google has done, still does, and won't stop doing to the web and its users.
Can you show me on the doll where the Mozilla touched you?
I remember... that was many years ago. Although, mine is PMTSD - Pale Moon. That is what I switched to and ran it for a few years with minimal issues.
When I finally did run into issues, and was called a fucking moron for bringing it up to the developers, I happily switched back to firefox-esr and was happy to find it's back to being awesome.
Are they proposing to shove a 22GB AI down our throats as part of a web browser now? Presumably I can just not do that and also not give away any keys to online AIs and this will be as good as an off switch for this "feature". I struggle to see the problem here. But then, I'm already running with NoScript so that's quite a lot of bullets dodged already. The tech bros probably don't even regard me as being properly online. (Suits me fine.)
Google has already decided Prompt API is coming. It will be whatever Google wants it to be unless another company can give Google a free idea which makes Prompt API better for Google.
This is the problem with dominant browsers. We went through it once with Microsoft and Internet Explorer, though what Google has done with Chrome/Chromium is arguably more toxic to the open web than anything from MSIE.
Chromium-based browsers are also a part of the problem. Anything on that engine gives Google leverage, even in Google-free variants.
Use Firefox. Get your friends and family using Firefox. Firefox needs to reverse its bad market share decline. We don't want a future where web "standards" are whatever Google dictates with only minor input from Microsoft and Apple.
Government, mainstream media, BigTech... in every corner of the world where there is power, it is being abused.
These things are not happening in isolation. Grifters, traitors, war crimes, rug-pulling cryptobros, the swarm of divisive bots, complicity, ex-spies working for BigTech, insider trading. The only thing on the menu is you and me.
There are rules, but people that supposed to enforce those rules just draw their wages that we pay and choose to look away.
They don't want to disrupt the wine and steak dinners and sometimes they themselves want some scraps from the table.
The notion that there is no rules is because civil service cares more about their gold plated pensions than the country.
…of the number of Web APIs out there.
Many are useful, some come in handy in very specific situations (e.g., serial API).
Others have that peculiar smell of someone's "toy project that'll get me noticed by management".
Web APIs should be, IMHO, in support of technology that is useful and long-lived which, again IMO, prompts will not be. They're just a stepping stone onto something else.
This is basically a concern that's reflected in multiple posts in that GitHub discussion.
It'd be nice to cite some examples. I've used Firefox or flavours thereof more or less exclusively since 2006 or so at home and at work. There were short dalliances with Opera about 15 years ago and more recently Vivaldi, but the web pages I've had issues with were in the single digits.
The most recent was one intranet page at work wouldn't show in Waterfox, it just borked out with a 404 error. However Firefox displayed it just fine.
As for search engines, like the commentard above I found Startpage rather good. Qwant is also decent if you want something lacking any U.S. tentacles, though.
Upvoted for recommending Qwant, they run their own index and are based in France where there seems to be a positive attitude to personal privacy. Has worked well for me for a couple of years now, just one sponsored ad at the top of each page. AI slop is available but you have to choose to use it. My default setup is Firefox with uBlock Origin and Privacy Badger, search set to Qwant. Oh and routed via ProtonVPN "just because".
Qwant doesn't understand fi and fl ligatures in pdfs. They return a recent paper of mine as by Julian Brad eld and Shanti Ulfsbjorninn.
Google did this about two decades ago, and my moaning about it is immortalized at https://harzing.com/popbook/ch13_2_2.htm , but they fixed it within a year or two.
Qwant doesn't have its own index, when Bing went down so did Qwant. Out of the well known alternatives, only Mojeek does have its own index.
Try using a user agent spoofing tool in Firefox.
Some sites are very lazy and do user agent sniffing and if it the browser text it parses is not on their allow-list you get rejected.
My FF user agent extension default I have set spoofs it as Chrome.
This tends to resolve "does not work in FF" issues.
On the subject of balkanised internet, Cloudflare does not always play nicely with spoofed user agent (when you set your agent to something very silly - not sure which particular parts it disliked* but Cloudflare "protected" sites often gave me grief when I was using one of my very silly user agent test strings - cannot remember the full details, but a very tongue in cheek, I remember OS was Acorn Atom!)
* who cares, I was just having fun
> "My FF user agent extension default I have set spoofs it as Chrome."
Unfortunately that shows up in site stats and bolsters the value of Google's useragent dominance.
> "On the subject of balkanised internet, Cloudflare does not always play nicely with spoofed user agent (when you set your agent to something very silly - not sure which particular parts it disliked* but Cloudflare "protected" sites often gave me grief when I was using one of my very silly user agent test strings - cannot remember the full details, but a very tongue in cheek, I remember OS was Acorn Atom!)"
They can tell it's spoofed by other browser ID checks. Sites and CDNs are increasingly cracking down on useragent spoofing, as these days, it's a much more reliable sign of a bot trying to cloak itself as human than a privacy-focused human with a browser extension.
Firefox at least listens to users, and in recent versions the AI cruft can be turned off fairly easily. I think their point is mostly about the vendor lock-in, and new prompt-injection attack surface.
Google's update at the bottom of the article is pretty much, "thanks for your complaint, we're doing it anyway."
Wait one moment! Let me try to understand this: the Prompt API lets the *remote site* provide prompts to some *local* LLM running within the realm of the *local* browser.
Firstly: how is that in any way useful? Is the whole point of a remote site, accessed by the browser, not purely that it provides some kind of information, data, or 'content' to the user of the browser itself? Even if that 'content' to be shoveled in direction user must be LLM-generated slop, I cannot imagine any possible scenario where the superior technical option wouldn't be to generate it, and store it, remotely and to serve the slop instead. At the very least, that would allow for the same content to be served upon every request or allow the service, itself, to know and audit what their user was served.
But, when all is said and done, everything above and indeed everything I could have challenged would boil down to a rational argument about a totally irrational population of A.I.-pilled idiots and the outcome of such argumentation will never serve a purpose and, so, it is only left to instead wonder: …
The local LLM will operate within the realm of the *local* browser and, either deliberately or indirectly, it will have some kind of local data[^1], history, profile or fingerprint[^2] that somehow falls within its context window: WHAT COULD POSSIBLY GO WRONG?
[2]: I'd *bet* the answer to why this is supposedly useful goes something along the lines of giving the local LLM some user data or letting it track history or past interactions. This is possibly the only capability this has over remote 'content' emission – and I'm not rating that 'capability' as a good thing, though, personally.
[1]: Even the graphics APIs like WebGL and WebGPU – nay, even the high-precision timer! – play or have played a role in browser fingerprinting. There's absolutely ZERO chance that anything running on specialised hardware like an NPU and consuming any kind of random entropy, as the inference stage of these models does, is going to somehow be different.
> "Firstly: how is that in any way useful?"
In theory, this would give users a choice of LLMs, letting them potentially use a local model and not send their data to a tech company which retains it or commits advertising against the users.
That's not why Google's doing it, nor is that likely to be the dominant mode going forward. But that's the theory. The broken web is littered with ideas which sounded good in theory but made things worse.
I think I saw an advert, and I think it was for Google Android phones, where one of your friends texts you to ask where/when your restaurant booking is tonight, and the phone texts back that information without intervention by you - because the phone knows. This seemed to me like something that you don't want a phone to do on its own.
From one of Vivaldi's developers: quoting this article
At @Vivaldi, we agree with Mozilla's Jank Architect. The Prompt API isn't a Web API, and we drown it in a bucket at compile time.
IMHO this seems to be quiet the correct approach.
> Firefox maker torches Google for building Prompt API into browser
How else are our corporate overloads going to control our access to information.
Q: What did Orwell's 1984 say about Winston Smith being under constant surveillance?
ChatGPT: ChatGPT isn't designed to provide this type of content.
--
Posted on 01/05/2026: 21:41 BST