ArsTechnica has more on this: https://arstechnica.com/information-technology/2024/04/aws-s3-storage-bucket-with-unlucky-name-nearly-cost-developer-1300/
Amazon waived the bill, and agrees that it shouldn't be this way.

In response to Pocwierz's story, Jeff Barr, chief evangelist for AWS at Amazon, tweeted that "We agree that customers should not have to pay for unauthorized requests that they did not initiate." Barr added that Amazon would have more to share on how the company could prevent them "shortly." AWS has a brief explainer and contact page on unexpected AWS charges.

Decwriter II printing terminal, mark-sense card reader, an acoustic coupler modem, and a Western Electric rotary dial telephone talking to a DECsystem-10 run by the county-wide educational consortium was my first experience with computers. My first program would calculate how old you would be in the year 2000. More syntax errors than lines in that one, but I wrote it just guessing, with no instruction in BASIC. Later on it was the pair of Apple ][s in school. In College it was a DECsystem-20, and my dad bought an Apple ][ for home. I bought a //e when they were brand new. 80 columns! Upper and Lower Case!

I just got a beefed-up 15" MacBook Air from Apple. I chose "12 months same as cash", rather than the 3% rebate on my Apple card. But yeah, at 9% financing, it would have to be something that I couldn't live without (like a new furnace), rather than a toy.

It's going to use a RTG.

One of the high technology parts of the French TGV trains is that they are capable of running on old track at non-TGV speeds. BrightLine really should do the same, the train slows down a lot, but does continue on to Union Station after stopping at Rancho Cucamunga. There should not be a need to change trains.

Because the regulations haven't kept up with the markets. It used to be that there was idle capacity, sometimes even excess baseload (electricity generated from sources like nuclear, that are very slow to ramp up or ramp down), at night, so users were encouraged to shift their loads to the off hours.
Now, in the spring and fall (when the sun is out but it's not too hot), solar panels can go full tilt in the daytime, while demand is still low.
The fact that electricity is now cheap, or even free, during spring and fall days, and still very dear later on in the evenings, is a price signal that tells people to invest in electricity storage that can soak up cheap electricity during the solar peaks in the day, and release it after the sun is down when it's really needed.

The machine I had was a HeathKit H19. This had it's own OS called HDOS. Not sure what the quality of that was or how it compared to CP/M. However the hardware also had ROM mapped to the first 2K or so (to run the program controlling the front panel display) which made it incompatible with CP/M. I somewhat remember it was already clear that all the good software was only for CP/M and I had the wrong machine and HeathKit screwed up. Anybody else remember these, have any comments on them? It does sound like creating HDOS was not a trivial amout of work, was anything interesting lost with it?

Thanks, that makes sense. I was wondering if this was industry-planted “research” so that they could tell regulators “no, don’t regulate PFASs; nature produces far more than we do on its own.”

No, ignoring the XY position of windows is a specific design decision by Wayland. They did it on purpose because they think it is a security problem. The idea that the desktop could just look at the requested positions and only ignore bad ones apparently is foreign to them. Instead they made it impossible for an application to store window positions.
They also purposely designed it so it is impossible to work with overlapping windows, by requiring that clicking in a window always raises it,a design that was removed from X10 to make x11. Their arrogance shows no bounds.

Have another AI that examines everything that the main AI attempts to write. All it has to do is identify that an output is objectionable. If true then it is never printed. The main AI can generate a new piece of text, repeating until it generates something non-objectionable. Or it can print "congatulations you got me to say something objectionable".

This is all irrelevant to whether censoring is good/bad, just that technologically it seems possible to fix any and all such bugs.

WTF? The left was hard against invading Iraq long before it happened. I know because I disagreed with them, Saddam seemed very dangerous and the left naive. They may well have been wrong, but they were against it.

There is no text size slider under accessibility on my machine (4k monitor, M1 Studio Ultra.)

Just verified that the option is there on both an M1 Max Studio (with 43" 4K monitor) and an M3 Air (built-in screen): Settings/Accessibility/Vision/Display/Text Size. MacOS Sonoma 14.4.1. You can set a default size, and then override that size for any applications you want. 9 point to 42 point, with the default (on my Air) at 13.

"The computer we will release in 6-12-18 months will be much faster than what our competitor introduced 3-6-9-12 months ago" is just empty marketing. Apple's upcoming M4 MacBook Airs will be much faster than the M3s, for some marketing-specified definition of "much faster".

When 2025 rolls around, will Apple be hyping up the M5?
I hope no one installs it in a starship...

Training is how you make an AI. The AI by itself is as capable as a newborn. Can't speak, can't listen. Training is what makes it useful. Without training data, those large neural networks won't do anything useful.

"Teaching" a computer so that it "oh, goes and finds it" is called programming, and that is done by programmers, whom one must compensate if you want the program to do what one specifies, rather than what the programmers choose. So it's expensive. And, actually a fairly complicated problem, when you get down to the nitty gritty of "what did the user mean by that", and push to increase the precision (i.e. how many of the documents returned were what the user was looking for) and recall (i.e. how many of the documents of the ones that the user was looking for, did the algorithm return). You can easily get 100% on either score, by returning all of the documents in your corpus (so that none were missed, 100% recall), or none of them (so that there are no irrelevant documents returned, so 100% precision). Getting a high score on both is hard.