undu

What’s not changing though is that most of their focus will be on integrating AI which most people don’t want.

I agree that AI chatbots are absolutely useless and have no place in a browser, but out of the three ML features in the screenshot, one is great for blind people, and another one is great for making the web more multilingual, so their usefulness is quite self-evident. Regarding ethics, at least for the last one it’s using a local model, and was trained using open-source datasets.[1]

What makes so-called “AI” bad is not the amount of users that can benefit from it, but how useful it is to the people that do use the feature, which usually means having experts tailor machine learning unto a single purpose.

I personally use the translation feature at least once a week when looking at news article that are not in English, and now I’m using a lot to translate Japanese webpages to plan a holiday there, so I’m very happy that Mozilla has invested time abd collaborated with universities to make this feature, I wish other people were less flippant about it just because it has “AI” in its name.

[1] https://hacks.mozilla.org/2022/06/training-efficient-neural-network-models-for-firefox-translations/

There’s a relatively new latex replacement, which is much easier to use, although its ecosystem is not as complete (obviously)

https://typst.app/

(The CLI tools are OpenSource and independent from the GUI thing they sell)

Did Synology just hire some brain dead Broadcom executive?

Well, Citrix’s CEO was Broadcom’ software boss

And also hasa place at the US treasury, he’s DOGE-affiliated as well: https://www.crn.com/news/cloud/2025/citrix-parent-ceo-krause-on-doge-role-we-re-applying-public-company-standards-to-the-federal-government

But the individual network packets are usually at most 1500 byes long, and applications encrypt the content. Hashing doesn’t prevent jack squat. It’s more likely to be DNS + IP blocks