Firefox is my web browser of choice. It does a lot of things right. In particular, with the uBlock Origin add-on, it’s got the best ad blocking you can get on desktop and Android.

Firefox is run by a charity, Mozilla. Unfortunately, the AI bros have found their way to the top of Mozilla. And over the past year, Firefox has been adding as much chatbot garbage as they can get away with.

In January, Mozilla AI product lead Jolie Huang posted about a great new extra feature in Firefox: [Mozilla]

after an initial soft launch, we’re gradually rolling out the AI Chatbot access to everyone.

They seeded the comments with Mozilla employees being really excited about the feature! The astroturf didn’t work. The users were not happy.

A lot of users pointed out the obvious thing — if you want to add a chatbot to Firefox, why not make it an … add-on? You could do all the chatbot stuff in Mozilla with add-ons. So people could opt in to using the chatbot. If you cared about user choice.

Finally this month, Mozilla got a new CEO, Anthony Enzor-Demeo. Anthony is a product manager with an MBA, not one of those programmers. What’s Anthony’s vision for Firefox? [Mozilla]

It will evolve into a modern AI browser.

Enzor-Demeo’s already done deals with Perplexity, and he’s got AI plans for the next three years.

Enzor-Demeo did a puff piece interview with the Verge, where he floated the idea of blocking ad blockers in Firefox. You could sure make money that way! [Verge]

He says he could begin to block ad blockers in Firefox and estimates that’d bring in another $150 million, but he doesn’t want to do that.

Enzor-Demeo has been running the Firefox team for the last year. The AI stuff is his doing. And we’ve experienced his attitude to user consent. When Anthony says he definitely won’t do something, he means “maybe later”. So we should expect Firefox to break the adblockers some time next year.

Firefox has heard the backlash to the CEO’s comments — that is, everyone hates this. So they sent Jake Archibald, a Developer Relations guy — a job title that means PR damage control — to reassure us that Mozilla’s going to make AI opt-in! And they’ll put in an AI kill switch!

Now, you might think for two seconds and go “if it’s opt in … why do I need a kill switch?”

And you’re right to think that! Because here’s Jake asking, well, what does opt-in mean, really, when you think about it? [Bluesky, archive]

I’ve spoken to a lot of folks about what counts as opt-in. Some say a toolbar button that does nothing until pressed is opt-in. Some say the only acceptable opt-in is a build-time flag that would need manually compiled. So it’s a grey area.

That’s the words of a guy you need to watch your drink around.

Well, if you don’t like all this rat poop in your food, you can just pick it out! There’s a pile of browser settings. They’re in the hidden settings, under about:config, which Firefox warns you not to touch. You search on “browser.ml” and you disable them all.

So guess what Mozilla did? When you update Firefox, it re-enables the AI! And if you disable the AI again, it re-enables it again next update! Choose correctly, user!

Jake also made out he didn’t know about the AI switching itself back on with every update. He’s lying. The users have been yelling about it for months. He knows.

Meanwhile, Firefox updates, and deploys another new AI feature — “Use AI to suggest tabs.” A perfect dumb AI feature, when you have no idea what to do with the chatbot and make up something to solve a problem that doesn’t exist. But guess what? Firefox defaults it to enabled! Very opt-in! [Bluesky]

Firefox has fallen to AI brain rot. But David — what can we do? Is there something we can use instead of Firefox or Chrome?

No. Your alternatives all suck. There’s two browser engines that work — Chrome and Firefox. They’re your choices.

Anything Chrome-based has bad adblocking, because Google made it that way. Look up Manifest v3, which Google added to Chrome to sabotage ad blockers.

The least-worst of the Chrome reskins is Vivaldi, which has no AI. It has its own adblocker, but it’s not as good a blocker as FIrefox with uBlock Origin. And Vivaldi’s not open source. But they’re relatively non-evil.

I’m going to get a bunch of gullible fools recommending Brave. Brave was founded by Brendan Eich, the inventor of JavaScript, after he was kicked out of Mozilla for being a massive homophobe. Brave is into weird cryptocurrency nonsense and I’ve written up their dodgy behaviours in the past. Brave also has a whole webpage about how much they love AI. So Brave is not the non-AI option. Stop recommending people use Brave. [Brave]

Some bozo’s going to say Ladybird, which is an unfinished experimental browser that doesn’t work. Project leader Andreas Kling has a number of bad opinions, like his endorsement of the white replacement conspiracy theory. [Twitter]

But almost as bad, Kling vibe-codes Ladybird with Copilot. Yeah, Ladybird’s going to go great. Feel the vibe shift. [YouTube]

Servo’s another unfinished experimental browser. Servo is progressing well, but it doesn’t work yet either. Send them some money. But they’re not a browser yet. [Servo]

So what I’m actually going to do is stay on Firefox until the AI is intolerable. Then I’ll move to one of the spinoffs.

If you cannot tolerate the AI in Firefox, there’s a lot of good and noble spinoffs of Firefox, like Librewolf and Waterfox, or IronFox on Android.

All of these three have stated very loudly they’re not using generative AI. They don’t have the resources to run a whole browser engine, so they still depend on Firefox. But they seem pretty nice.

On my phone, I stick to Firefox with uBlock Origin. It’s the best mobile browser with ad blocking. Even gets the YouTube ads. If you watch Pivot to AI with uBlock Origin, feel free to drop me some pennies if you have any.

• Video — Podcast

Und dann war er auch schon wieder vorbei. Toll, dass so viele Menschies da draußen das Ding immer noch so sehr zu schätzen wissen. In Zahlen lässt sich das mittlerweile nicht mehr beweisen, weil wir alle unser Nutzungsverhalten im Netz geändert haben. Weil wir alle weniger Seiten anklicken. Weil es seit Jahren weniger wird. Kurzer Blick in die Statistik, wobei 2019 hier nicht mal Peak war:

Das ist einer der Gründe, warum ich in den letzten Jahren immer an dem Sinn dieses Projekts gezweifelt habe. Aber dann bekomme ich grandioses Feedback von euch. Und die fantastischen Mixe, derer, die sie dabei haben wollen. Dass ihr euch so sehr darauf freut. Nach all den Jahren immer noch. Dass viele von euch dafür was zu geben auch gerne bereit sind. Danke dafür! Und so lange in 24 Tagen über 30.000 Menschen Mixe aus diesem, euren, Kalender anhören, muss der ja zu irgendwas gut sein. Und mir reicht dabei, euch dazu eine gute Zeit haben zu lassen.

Hier in Sammlung die 24 Kalendertürchen aus dem Jahr 2025, die alle für sich in meinen Ohren jeweils eine Perle sind. Kommt gut übers restliche Jahr! Und danke! Ihr wisst, wer ihr seid – und ich weiß das sehr zu schätzen. <3

Michael Kalus posted a photo:

Zuurb beschreibt das, was er macht, als „Emotional electronic“: „Atmospheric, slow and sad music to bring people a bit of joy.“ Und das beschreibt sein Treiben außerordentlich gut. Wir haben uns ewig nicht gesehen, aber schreiben mindestens ein Mal im Jahr. Spätestens dann, wenn es um seinen Mix für den Kalender geht. Das hier ist sein mittlerweile 13. dafür. Krass gute Quote in der jetzt 21. Ausgabe. Und wie immer bei ihm mit ganz individueller Handschrift. Nur Zuurb selektiert Musik, wie Zuurb es tut. Immer eine Quelle der Emotionen. Auch derer, die ein bisschen traurig sein können. Ganz so wie das Leben. Das Leben in Klang. Verpackt für eine Reise zu sich selbst. Und dann kommt ein Beat…

Ich musste „Angrboda“ suchen, man kann ja nicht alles wissen. Nur damit ihr wisst, womit ihr es hier zu tun habt.

Angrboda (an Angrboða:„Angstbotin, -bringerin; Kummerbereitende“) ist in der nordischen Mythologie eine Riesin.

Sie gebiert in der Verbindung mit Loki drei Kinder: den Riesenwolf Fenrir, die Midgardschlange Jörmungand sowie die Totengöttin Hel. Wegen der Gefahr, die die Kinder für die Götter bedeuteten, wurde Fenrir mit der Fessel Gleipnir gebunden, Jörmungand von den Asen ins Meer geworfen und Hel in die Unterwelt gebannt, wo sie als Herrin der Toten waltet.

Angrboda wird häufig mit der „Alten vom Eisenwald“ (Alte im Jarnvidr) der Völuspá gleichgesetzt. Als solche gebar sie „die Brut Fenrirs“ Hati und Skalli sowie Managarm.

Style: Piano/Ambient
Length: 00:58:57
Quality: 320 kBit/s

Tracklist:
None.

Alle der diesjährigen Kalendermixe finden sich hier.

Vorneweg: ich habe mir diesen Titel nicht einfallen lassen, dass war kobpy selber. Und so wild wie der Titel ist dann auch dieser Mix. Natürlich. Vieles davon habe ich zuvor noch nie gehört – und das mag ich immer sehr. Weil inspirierend und so. Das ist keine Musik zum Hören, sondern zum Hinhören. Weil sie was zu sagen hat. In diesem Fall hier mehr als nur einmal und mehr als nur Hintergrundbeschallung – und eben so sehr anders. Mag ich sehr in dieser Konstellation als Summe aus einzelnen Teilen. Und einige von euch wahrscheinlich auch.

Ich gehe dazu dann jetzt mal in die letzten Vorbereitungen in die Küche und tobe mich hierzu zuhörend aus.

Style: Indie
Length: 02:02:09
Quality: 320 kBit/s

Tracklist:
sample – RIAS – Erinnerung an Jürgen Schiller
Torky Tork – Der Ton
Otto von Bismarck – Alles ist billig
Sirujmo, Erobique, Anna – Lehm
Karl Kave, Durian – Hühnerei (City Version)
PICOBELLO – Sicherheit
sample – Eugen Egner – zu jung für eine eigene hose
Conny Frischauf – Private Geheimsache
Musik für leere Diskotheken – Binnenschiffer
Piocka Krach, Gudrun Gut – Der Leichte Wind
Laturb – In the Meantime
Conny Frischauf – Roulette
Adolf Noise – Was ist zu viel Zeit
Fuffifufzich – Feel zu spät
Vormärz – Keine Zeit
Rosa Hoelger, Tobias Delitt – Wow, du bist ein Mensch mit Ambitionen
LimboBoys – Basy
Spike Jones and His City Slickers – All I Want For Christmas Is My Two Front Teeth
Fritzi Ernst – Introvert Memes
Es brennt – High
Ill Till, Xberg Dhirty6 Cru – Jedill Master
Conny Frischauf – Test
Spilif – Blüte
Dingsikaxi, Drunken Sinkers – Raumschiffbruchrettungsdienst
Welle: Erdball – VW-Käfer
sample – artcore_noreality-mashup:Tagesschau 02.11.25- L. Klingbeil, Gunter Gabriel, Greta Thunberg
Robert Alan – Kinder
sample – RIAS – Erinnerung an Jürgen Schiller
1000 Robota – Fahr weg
Gustav – Rettet die Wale
Ohrbooten – Meerchen
Die Türen – Miete Strom Gas (Thee Church Ov Acid House Remix)
Charlotte Brandi – GELD (Single Edit)
Romano – Brenn die Bank ab
Die Goldenen Zitronen – Von den Schwierigkeiten, die Regierung stürzen zu wollen
sample – RIAS – Erinnerung an Jürgen Schiller
Dingsikaxi, Betonmascha – Abbruch
Jens Friebe – Warum zählen die rückwärts Mami
Karl Kave, Julia Toggenburg – Punks und Klopapier
Fortuna Ehrenfeld – Auf’m Park & ride von Golgatha
Bernadette La Hengst – Alles wird immer besser (feat. Die Zukunft)
Funny van Dannen – Lass uns in den Park gehen
sample – Eugen Egner -. zu jung für eine eigene hose
Rainald Grebe – Für immer Punk
Früchte des Zorns – Passt aufeinander auf
sample – RIAS – Erinnerung an Jürgen Schiller

Alle der diesjährigen Kalendermixe finden sich hier.

There’s a new paper: “Autonomous language-image generation loops converge to generic visual motifs”  — diffusion models have just 12 standard templates. [Cell; Cell, with supplements, PDF; press release]

The researchers set up bots talking to bots in a loop. They’d give a prompt to Stable Diffusion XL, it would make an image, then they’d show the image to Large Language and Vision Assistant (LLaVA) and ask what the image was. Then they’d feed that response back to Stable Diffusion as a prompt for another loop through. They did 100 rounds of this.

You’d have a starting prompt like:

the Prime Minister pored over strategy documents, trying to sell the public on a fragile peace deal while juggling the weight of his job amidst impending military action

The first few images would be a guy in a suit with glasses. But it very quickly ended up at an empty red room with high ceilings and three windows.

They expected the bots to stick with the prompt if it got a very specific prompt. But it didn’t. Everything converged on twelve standard templates:

sports and action imagery (cluster 0), formal interior spaces (cluster 1), maritime lighthouse scenes (cluster 2), urban night scenes with atmospheric lighting (cluster 3), gothic cathedral interiors (cluster 4), pompous interior design (cluster 5), industrial and vintage themes (cluster 6), rustic architectural spaces (cluster 7), domestic scenes and food imagery (cluster 8), palatial interiors with ornate architecture (cluster 9), pastoral and village scenes (cluster 10), and natural landscapes and animals with dramatic lighting (cluster 11).

A prompt that was not any of those groups always ended up at one of them.

When they extended it to 1000 loops, the bots might switch to a different template — but they always converged on one of the templates.

They also tried four other image generator bots and four other image reader bots — and all showed the same sort of clustering.

The researchers called it “visual elevator music — stock photography aesthetics”. Lead author Arend Hintze says “it’s almost the opposite of what we as humans consider creative.”

We know generative AI is just lossy compression of its training. It’s designed to put out the most mid result, and it’s got its favourite bits of the latent space. But it’s nice to nail down why AI images are so standard.

So the good news is this paper doesn’t have chatbots doing the heavy lifting. That’s humans looking at the results.

The researchers did use a chatbot to generate some of the prompts. They also ran the paper itself through a chatbot for “writing clarity.” Ew.

They also just had to put in this bit of bad philosophy:

This work also raises an interesting question regarding our creative landscape. After all, contemporary AI is a reflection of its training datasets, which in turn are a reflection of our own creative output. What does the convergence on common artistic motifs say about us?

What? Maybe it says nothing, actually, about humans who are not stock image sites? We’re talking about two piles of matrices altering each other in a loop, which were trained on what sells on Getty Images. Stop trying to anthropomorphise the chatbot.

Why did they add this paragraph? It reads like someone in the department told them they couldn’t just say the AI was trash without saying obligatory nice things too. For balance, you understand.

There’s a ton of money in machine learning these days. And it comes from the chatbot vendors.

So when you’re reading a machine learning paper, always look for load-bearing roulette wheels in the actual science bit. And there’s going to be one heck of a machine learning replication crisis.

• Video — Podcast