> Define best I suppose?
I don't define best, because it only *seems* to be the best option. The question here is not if it is a good option, but why they consider it to be a good option. One could try to discuss if it indeed can be a good option (e.g. when criticizing how universities currently work) but that was not what I wanted to do here.

The problem now is, that people short-term try to pick the fight to play detective and catch cheaters, but the long-term question is how to shape learning. First there is a perfect cheating tool, and second this tool could also be used for learning. How does the educator get the student to learn (using whatever method the student prefers)? And in another aspect, can the new tools help students to learn better?

The main goal of educators is not to fail lazy students, but to get students to learn so they can pass the exam. Some educators may not be good at it, but in the end students and educators have the same goal. So how should the process be shaped that they still can cooperate to reach that goal in the future?
I don't have the answer, but I don't think universities should be a battle ground between teachers and students if cheating or cheat detection works better, but serve the long-term goal of the student not only passing the exam, but also graduating having the knowledge they wanted to have when they started.

A URL to a paginated website does the same for good websites. There are a few mistakes you can make though, such as starting to count pages from the most recent page instead of the oldest one.

The thing why the EU is against it is, that endless scroll (among a few other patterns) is designed to keep users engaged more than they intended to use the app. That's why they want you to scroll. If you want to go deeper they indeed only want you to click ads, but to show you apps they need you to use the app. They also need user generated content (including interactions) to keep the site alive for others.

Synthetic data contains too many model names to prevent that from happening. Even without direct distilling you can bet that they incorporate every dataset they can find in the different repositories. And many of that are distilled from a wealth of different models. Anthropic would be stupid if they wouldn't do that themselves as well.

And Anthropic created content that is not copyrighted and cries foul when someone uses it like it were not copyrighted. It is even unclear if they could really ban you for that, but of course they can ban using their "... or without a reason" clauses.

It's still a design choice in the app. There could easily be pages to, for example, jump to a specific day, or you could set bookmarks. The technical side of a good UX may not be all that simple, but currently nobody is working on it anyway. They want you to doomscroll. And the EU wants to help you not to doomscroll that much unless you really want to.

Apple mostly buys AI from others. They trained a few models (unremarkable but not bad for their time) but most of them aren't in use anymore and they talk about licensing Gemini.

^Walignment to protect us from rogue AIs?!

There are two points for grading homework that work in favor for students (even though not all may understand this).

First, it motivates you. If it is optional, then you may skip over it or tell yourself you do it when you have more time. It needs to be graded with deadline to make your really work on it on a regular basis.
Second, it filters you, if you're unprepared. If you fail the graded homework, you (probably) fail the exam. By failing you without negative consequences (except for having the course again), the professor does you a favor.

Copying (with or without AI) helps you to avoid these factors and get into the exam to fail there. Previously copying (often) was easy to detect so people could be failed for the homework before the exam. Now they have "unique" solutions and tutors are required to play detective, which is not their job. They are also not hired as police officers that control you're not cheating, but as someone who should help you learning.

If everyone works against the system, the system has a problem. And the problem is not, that the students are cheating, but that cheating seems to be the best option for them, either because it really makes things easier (should teaching be adapted not to require homework then?) or because it seems to make them and then they fail the exam (should be adapted how to motivate them to learn?). Making rules and enforcement stricter only treats symptoms.

I never really understood TikTok, possibly for that exact reason. They also already got their share of criticism and they may be next for the EU to look into if they are addictive to children for these reasons.

Books literally use pagination ...

Who says these companies are not targeted by the law? This is the article about Meta, when they continue with TikTok Slashdot will probably also report about it.

Large companies are adopting similar patterns leaving the user without a choice to help themselves. Do you expect to chat with your friends in Slashdot comments? If you want a social network where your contacts are, you can choose if Meta, Google, or Musk provides you endless scroll. If you want a social network without endless scroll, you can choose one of the few and be pretty lonely there.

But are they convenient features?

Autoplay: Is there anyone who wants a random video to start catching their attention while reading their feed? That's totally an advertiser feature.
Endless scroll: Just consider the alternative. With pagination you can bookmark where you stopped scrolling, so you can continue later.

Endless scroll forces you to read until the gap between the top and where you left off last time is closed, otherwise you never will find the position of where you stopped reading again. Chronological ordering would help a bit, but you still have to jump around between positions you can only find by scrolling and not by page numbers.

They are selling your attention and so they are optimizing how much content gets your attention. They are not optimizing for you to use the app effectively, they optimize for the ads to catch your attention.

The list doesn't look too useful. Obviously inspired by the Google list in that style it mostly lists deprecated models being phased out. It's not like they killing for example GPT-5.1 left a gap. GPT-5.2 was a drop-in replacement and they didn't promise GPT-5.1 to exist for longer.