You're not being too picky. The summary pretends there's only one temperature in play, whereas the original source mentions at least two temperatures (one defined after a phonon-phonon equilibration time; and a more conventional thermodynamical temperature which is only well-defined after a longer time). The sensationalism comes from thinking the longer-time-scale one applies at very short time-scales.
It's shit like that which makes headlines and then makes people distrust research.

'Exonate'?

:) Of course you're right, which means extrapolation is gonna be a /real/ shit-show.

Careful. The unpacking of all those parentheses may cause us to run out of memory on the acronym stack.

It's far faster than exponential (assuming that's 1% of a constant-size market, then 2% of the same market, etc.). The first doubling time was 8 years; an exponential growth process would keep the same doubling time, but it took another 2.9 years to double again.
So your "not even close" is correct, in that exponential growth is far slower than whatever this is.

Thanks. It doesn't appear in my Apps list, so I probably took it out via the ADB interaction. However, I don't like seeing it in the list of possible "Digital Assistants" if it's been uninstalled.

Oblig. Dr. Ian Malcolm: "Yeah, but your scientists were so preoccupied with whether or not they could, they didn't stop to think if they should."

That's good to hear!

Now, how do I rip out Gemini, or ensure it's never installed?

EDR is sometimes all you have to know something happened. Waiting for DLP to note a loss can be too late if there is behavior which isn't currently being flagged as suspect. I've seen cases where employees attempted to establish a new baseline of behavior which EDR caught before they got around to leaking things and were told by management how they should be doing backups of their work machine and to stop the ways they were trying. If it happens again, then you have stronger reason to think they are up to no good and need stronger re-training or axing.

Macs have indeed changed, it was certainly doable on Intel units, though some options could be turned off from afar to make it harder. With Apple Silicon + FileVault + disabled external boot, it's pretty much impossible unless you've an insider who knows the needed keys, which should be safeguarded well away from easy single person access.

Eventually your device will go back online and whatever EDR or DLP your company has installed will send the offline logs to a server, same logs which are generated when online. How big do you think the ring buffer is for those logs?

You might have gotten away with it in past, it suggests your employer was utterly incompetent if they didn't notice activity like this.

It's a very good question. It looks like it was mainly failures to generate a result within a predetermined time. Some of the failures were due to cryostat hardware failures (a fridge went out during a NIST campus closure); some due to fiber + interferometer polarization drifts; and so on. It also appears that [perhaps?] a few of the misses are due to latencies in the timetaggers to record a common timebase. I can't quite tell from the arXived version of the paper: https://arxiv.org/abs/2411.052...

All in all, it's a marvelously good overview of the impressive experiment!

It's also happening tomorrow, according to the summary. Hmmmmmmm...

Hm. Perhaps this is a use of 'grifter' I wasn't aware of.

How is Geerling a "grifter"?