So Spotify is a loop (it automatically plays the next recommended song)

True. I feel like "primary attention" should be a part of it too.

And privilege escalation.

>I think the people who need those abbreviations spelled out are not the target audience of the article.

I agree this is true for the economist readers, but I clicked the link thinking this was about how Dropbox (in a company) is an obvious target for privilege escalation (for red teams and hackers), which also sounds plausible and interesting.

But I don't mind, after a few seconds of loading my mistake was obvious.

Compared to the naive implementation presented in the blog post yes, but the PR uses AVX, which can process 256 bits (32 bytes) at a time, so I'm not so sure

>and when you want to do anything that can't be accomplished by plain JSON

Like comments? ;)

YAML then, or something like it. I guess I just like dumb formats for this stuff.

>METR measures the longest software engineering task a frontier model can complete successfully half the time

Longest? Yes, they actually use actual wall clock time. I don't think this metric makes any sense. I don't know about the others, I stopped reading at this point.

The others are an IQ test (TrackingAI), a test of Ph.D. questions across multiple domains (Humanity's Last Exam), and graphical pattern matching (ARC-AGI-2).

What's interesting is that while they are rather different in nature (yes it is odd that METR measures clock time as opposed to iterations etc.) but the behavior of the resulting improvement curves are extremely close.

That's the punchline: 4 independent measures point to the same conclusion. That increases the chance that the conclusion is correct.

I did that when I was younger and internet forums we're still a thing. The difference is that I had to do actual research, which often took me half an hour or more, to solve the problem for the asker. I got knowledge and "the feeling of being a notable figure" as you have said. The other side for the answer. I believe my answers were good.

The difference is that nowadays people can just skip the research part and copy-paste Llama. No value added.

>The difference is that nowadays people can just skip the research part and copy-paste Llama. No value added.

and why would the person doing that do that? if the person asking the question wanted to know what an llm thought, wouldn't they have just asked the llm for an instant response instead of having to wait for someone else to respond? i'm at a total loss of being able to understand the point. it's obviously different if the response is a bot posting just in case that needs to be explained to a bot.

I assume it's unambiguous in Japanese.

That's what hackers do in popular stories, right? In reality, most cybercriminsls just don't care that much (they usually use Tor browser, though). I mean they mostly use discord and telegram (both unencrypted).

Serious "hackers" are usually state sponsored now, or members of mature groups.

They don't care, I would say they don't know.

I saw an article where they analyzed the leaked IP addresses from a breach forum, and some of the top ten were Surfshark and iCloud Private relay.

>and some of the top ten were Surfshark and iCloud Private relay.

what's wrong with icloud private relay? It uses a 2-hop architecture so it's probably more private than any single-hop VPN.

It's good, but it's funny to think about cybercriminals running Safari and an iCloud+ Apple account.

I don't get it. Why did you buy a third spatula, if you consider it a significant expense (i.e you can't replace ones you have if they break), and you already have two.

Perhaps they didn't have the other two already. From the phrasing, it seems they bought one when they moved out of their parents house and received/inherited the others