You are not any more secure with google authenticator for 2fa, are you?

These days you don’t get much extra benefit on a VPN over TLS which you get on 99% of websites.

It really isn’t superior. It’s just the hivemind that gets annoyed with Plex being stagnant, not open source etc. that claims it is. At best it has feature parity for some use-cases. Don’t get me wrong, it’s neat, but it’s not as polished as Plex.

Are you also a “guns don’t kill people, people kill people” kinda person?

Firefox also implements manifest v3, just without the user-hostile restrictions.

Thanks, but I’m specifically asking about Я̀ vs Я. I’ve never seen the variant with an accent before and the wiki page for it seems to be missing.

What is the accent on Я? (I’m a beginner Russian student and never cane across that so far)

The entire macromedia suite was so good. I had so much fun and learned so much as a teen.

Theoretically, what would the utility of AI summaries in Google Search if not getting exact information?

So what, you keep an ungoogled-chromium around and use it occasionally for compatibility, if you really need to. Doesn’t mean you are obligated to use it as your daily driver.

FWIW they are cannibalizing ads right now with AI summaries, since people will navigate less to websites (in the world where they are useful, which they don’t seem to be at the moment).

It works if you take it for its function: to signal the body that night is starting in 1-2 hours, and take it in a small dose (1-2mg). It’s not a sedative, like many seem to assume. It’s best to take it when you start your night routine.

Peer review, for all its flaws is a good minimum before a paper is worth taking seriously.

In your original comment you said tha model collapse can be easily avoided with this technique, which is notably different from it being mitigated. I’m not saying that these findings are not useful, just that you are overselling them a bit with this wording.

That paper is yet to be peer reviewed or released. I think you are jumping into conclusion with that statement. How much can you dilute the data until it breaks again?

The problem is that idiots drinking H5N1 milk might bring about human-to-human transmission of it which will be a dire situation. To date, around 50% of the 200-something human cases resulted in death. For comparison, the worst mortality rates of covid were around 20-25% in the very beginning.

Is your argument for bombing being the right decision the same (that it resulted in less bloodshed overall)? If so, how can you estimate the body count of the alternative (a prolonged conventional war, I assume)?

OpenAI clearly already scraped the pre-LLM (aka actually useful) content from SO, this entire deal is happening after the fact to avoid litigation.

I used to keep my steam games on a separate windows 10 partition and it worked exactly as you describe after a reinstall, it was all there. It’s still incredibly cool that this works on Linux and we get to use it as daily driver without being forced to dual boot for games. A windows installation still lingers on my desktop but it’s been years since I booted into it.

I went from a company that used github to one that uses gitlab. I thought it was going to be great and was excited for using a new thing. But it’s really clunky in comparison.

People need to learn what ragebait means