There's more to ssh than just interactive shell access, and a few passes with sftp to upload a new key that's bound to executing the shell is why you don't rely on a flawed interpretation. Don't guess when it comes to things in /etc/passwd and /etc/shadow. Go and read the docs carefully.

Except "!" has an already defined different meaning to the library that reads the passwd database, so you're just creating a new problem.

Would anyone happen to know the OUIs used by this company? Ten million plus devices means it might be worth our time to just go ahead and blacklist those things enterprise-wide.

I'm just thinking there's no way all that code was actually reviewed in earnest by humans in eleven days, so... it's a cute idea but that's a "hell no" for actually running it on anything.

When the focus is on being "joyful" instead of things like "useful", it's no longer educational, it's simply entertainment.

Had Reddit not forgotten it's core utility was people helping people figure things out when they began prioritizing a focus on ad revenue, they'd be happy about this because it would mean fewer site visits where the reader isn't likely to contribute anything useful in exchange for the limited bandwidth and storage their access requires.

I would not particularly *mind* going to Reddit to participate in those threads, but they've become a bit "user-hostile" since they decided to start working up to an IPO, and now technology is turning them into the place where answers are decided upon and born, but not really any easier to get them than the SEO-obsessive sites. The bulk of straightforward questions about common things are now something an LLM can reliably answer without feeling like making things up.

For the things many people are searching for that normally sites like Reddit would have threads discussing it to death, crappy ad-laden websites have made it their mission to climb ahead of Reddit and other discussion forums in search results so they can give the exact same information, but only after wading through 40 megabytes of ads, two video panels, and a captcha. Now that LLMs are able to make sense of silly questions and provide straightforward answers, the LLM's answer is able to make that particular information workflow almost obsolete.

For example, with a question about a video game or piece of complex electronic equipment, I'd gotten used to skipping over the first few spam-laden search results to reach discussion forums where actual people had already worked out the answer because the problem was relatively straightforward (I just didn't have the answer, yet), and those answers were simply being distilled with very little genuine effort by the webshites into scrolltext dross that was 85% ads by volume. Now an LLM usually just tells me the answer as if I had asked it a direct question about the subject so I don't have to scroll or even consider the existtance of some wikia monstrosity or listicle written by someone who only learned about cell phones a year ago.

Clearly the execs at Amazon watched The Congress and just completely misunderstood the message.

It's weird how when someone finds a fly in their soup the answer never seems to be to just add more soup.

Because once again, hiring a "cloud" vendor and then storing your data in just one remote datacenter is not a smart thing to do.

The SMS spam is largely being done by an outfit called Winred, and you can guess what their political affiliation is and why they consider themselves exempt from the legislation. It's going to be a couple of years before that changes.

The Do Not Call database actually works, but the spam callers are almost exclusively criminal enterprises calling for the purposes of fraud, so... you can guess how much they care about the law.

The last time I checked it was the FCC's job to stop them. I may have to lower their grade from a 'D' to an 'F' if they continue to only ban the companies once or twice a year for taking these calls for money.

There are some very obvious errors in their methodology if they've claimed to find "significant signs of AI authorship" in webpages that were written before ChatGPT.

That actually is how it is for everyone--except Texas because they were gullible enough to allow themselves to be talked into doing it the dumbest way.