addandsubtract

That's not how captchas work...

This is just an ongoing trend to address coding agents first in documentation, and people / manual methods second.

Thanks for the break down. I somehow missed the 2nd point. Crazy that they just used an LLM like that.

> Notice that the spider is carefully moving out of the way, just a second before the ant is launched.

It's not moving before the ant is launched. It's moving as soon as the tension in the web is gone, ie. there's movement in their web. Most spiders react to movement in their web.

Or Palantir films your hands while trying to enter a venue, border, or voting booth.

They said hands, not feet.

As soon as I get more than 2 captchas, I'm out. At that point, I'm assuming they're just testing my patience to train their models. Switching VPN servers is much more effective.

Welp, might as well accept the status quo and go back to consuming our soma.

You've mentioned the Iran airstrike twice now. Do you have a source for them using an LLM for their "intelligence" / decision making?

Meanwhile, I'm trying to figure out how to give up, because I can't find any more words. I found 8 (2 bonus ones), and am stuck finding any more. There should be a "reveal" button, so you can see what else you were supposed to find.

Edit: solved it! Being stuck is annoying, though.

The AfD is a far-right extremist party in Germany with currently 28% projected support[0].

[0] https://dawum.de/Bundestag/

But assuming a random draw, you're more likely to select people with higher luck.

Label AI as porn and the payment processors will cut their ties automatically.

I mean, there are already places where you can buy tokens at 10% of their original cost.

The same way a company wouldn't torrent books?

Why would the model know what "Claude-like" is?

The AI can cheat, because it can move a blocking piece out of the way to then attack you in the same turn. Or maybe it's just something to keep in mind, but it caught me off guard.

Does the trend only show absolute numbers? Because I think it should be divided by the number of posts during the time frame (day?).

Didn't they originally introduce those tokens to make the models smarter by second guessing their "thoughts"?

That requires actual activism, getting involved in local politics, and mobilizing. All things that Americans, ironically, can't afford. So here we are, creating trillionaires instead.

Unpopular opinion: a government funded and run lab, available to the public. The EU could make it happen.

Shouldn't we still have people in the loop for selecting/proposing the best implementation (plan)? Vibe coding an entire solution from a prompt still doesn't feel like the optimal way to write software.

Apple just rolled out all of the liquid glass components and style guides to update apps with. No way they're getting rid of it any time soon.

Now that you mention it, why don't we encrypt injectable data that comes from users and only decrypt it on the client?

Depending on how much access the AI agent has, there are worse things to inject it with than a link.

I thought this article was missing a (1999) in the title.

There is/was also 2-minute papers[0], but the videos lost their plot a couple of years ago.

[0] https://www.youtube.com/@TwoMinutePapers

> Sadly, that website is no longer maintained after its acquisition by Meta.

Wait, I thought it was aquired by Huggingface, because that's where the domain points to: https://huggingface.co/papers/trending

Anyway, as a huge fan of PWC, I'm glad to see it revived! One of the main annoyances of the old PWC was the searchability / discoverability of papers. I hope that now, you can create embeddings of the paper (summaries) to improve the search and make finding related papers easier.

Bayesian statistics to the rescue! You hear about the things that (can) go wrong a lot more, than the babies that are born without an issue. Basically, the opposite of survivorship bias? Anywho, millions of babies are born each day. Only 3-4% are born with a major birth defect or severe health issue (in the US) [0]. Now, a pregnant woman will undergo a routine screening, such as non-invasive prenatal test (NIPT) or a detailed anatomy ultrasound. With a positive test result, the probability of the baby being born healthy goes up to 99%+. Even with a negative test, the probability of the baby being healthy only drops to 80-50%, and then further testing will dial those probabilities in even more.

All this to say, all will be good – don't let the rare anecdotes get to you.

[0] https://www.cdc.gov/birth-defects/data-research/facts-stats/...

Bro is in this thread defending Google like he owns stocks. Oh wait...