The real problem with AI, and the AI discussion is how muddy it is. Are we talking about llm's diffusion models, or classification systems? Do we mean to say that we're talking about transformers or the underlying architecture? Are we discussing huge data centers or device based AI? Nascent, active, or dormant compute? And the same is true for the ethics, legal, and data governance conversation.

Every single one of these things is a different discussion.

AI is not a monolith.

I'm sorry for your loss :( And that's always the risk of a low widthrawal rate, that you're leaving money (and enjoyment of it) on the table, of course, that's your choice, and depends on where you want the money you don't use to go, be it charities or family.

The numbers have been revised a little, closer to 4.5%, and that's still super conservative (that's the amount of money you'd need to spend to survive the worst case scenario for 30 years). With 1.5 Mil in savings, you could safely withdraw $65K/year and be safe. The average of safety is closer to 7%, which means that you only need $1M to retire (yeah, with a bit more risk). Here's a really good article that explains it: https://www.fa-mag.com/news/ch... And of course, there's also there's Bergen's book "A Richer Retirement: Supercharging the 4% Rule to Spend More and Enjoy More"

That is somewhat misleading. In this case you control (more or less) the client, so you can install a root certificate on your firewall and the client and let the firewall do its MitM on all your traffic. If Windows tries to evade that, the firewall will fail to decrypt the traffic and block it, which was the intended result. If Windows does not evade the MitM, the firewall can do full L7 filtering just like in the good old days.

It is wild that companies get to manage pension funds. That kind of thing is not legal around here...

I ran a dedupe on my drive and eliminated 8 redundant copies of the goatse guy.

#SaveWaterWithGoatse

You can't do that. The copilot is not a real key to the keyboard protocol. It sends something like Windows, Shift, F23. You cannot sensibly combine it with other keys or make it reliably control a modifier state. This is completely unlike the Windows key which is not only its own unique keycode but also typically gets non-conflicting lines on the keyboard matrix, so the hardware lets you combine it with any other key.

There are still unused keycodes available, AFAIK. It makes zero sense that the Copilot key was crippled. If it was only crippled in hardware, vendors could fix that, but the only way to fix the Copilot key is to reprogram the keyboard controller firmware, which then makes it incompatible with Windows.

Not the only one. My laptop and all my virtual machines (other than appliances) run Fedora, and the 6-month upgrade does not break anything.

Normal updates are automatically applied every morning.

"scraping" is maybe questionable, but there's no question that an AI is accessing your website.

Serious question.
Why?

Every time this happens, the people doing it pretend it's the first time this has happened in the last x number of years since the c64's release.
Although, this is the first time a project doing it has filled their entire site with unedited slop. Doesn't make me feel great about the process here.

Things I want from a project like this:
- Technical specifications and circuit board porn.
- Operating system details
- Wifi available, you say? Tell me more about the networking stack!

What exactly am I buying, other than a C64 case that's outfitted to look like an iMac from the early 2000s?

None of this is clear from the website.
It's an opaque project that provides almost no useful information on the product that they're selling.

Unlike, of course, the rest of us.

Until a year ago, the main thing I knew Overstreet for was running his mouth off about how braindead btrfs is, and how bad its design was. He may or may not have a point, i don't know enough to judge. But it seems at the moment that there are some horrific bugs in bcachefs, which suggests that Overstreet perhaps isn't the genius he thinks he is.

Phoronix's testing a few months later seemed to show that bcachefs is usually slower than btrfs. https://www.phoronix.com/revie...

It will be interesting to see if this continues in the next round of benchmarks, or if it was a temporary regression.

In every car except Teslas. In a Tesla, the lane assist will not allow deviations from its chosen path. If you try to correct it, it will fight you until you do it strongly enough, at which point it will turn off entirely.

There is no "encourage" in a Tesla.

Steeeve has been spouting nonsense, blaming the pilots without having any concrete evidence. Then his next video suddenly comes up with "oh, new evidence, the RAT deployed" even though there had been reports of this within hours of the crash.

He has zero credibility at this point.