Radar is cheap now, thanks to semiconductors getting smaller and faster the analog front-end which used to be a long expensive chain of components is now much smaller, the ADC is now faster, more accurate, and cheaper, the processor is now fast enough to keep up with a higher bandwidth signal. You could probably drive a very rough radar system directly off a Pico's GPIO and ADC, maybe toss in a decently fast op-amp for a receive amplifier.

Where you will run into issues is processing radar signals into usable data. If you're happy with the results that radar was giving 30 years ago then it's fine and dandy, but the magic of modern radar is in the software, not the hardware.

Someone I know has developed radar hardware, specifically around signal processing, and while they couldn’t say much, my impression from them was that due the very large amount of data involved using off the shelf CPUs would be very limiting, which is why his job for a period of years was developing a design to be put on an ASIC which is significantly more effective at the task.

Can confirm, radar signal processing is very bandwidth heavy and not suited to a typical CPU.

Can also confirm that it is generally regarded as impolite to talk too much about it!

"impolite" is probably an understatement in the context of discussing details about the development of defense systems outside of work.

Wow! You managed to break almost every financial law and regulation in one go. That's amazing!

Reminded me of Arrested Development's George Bluth's "Light Treason".

I feel like it would be unfair to make them include the SVG decoder, the artical uses TypeScript and I believe that is a feature most runtimes (browsers) provide. However they do have to include the SVG files themselves. The Canadian flag from Wikimedia comes in at 699 bytes and France at 262 bytes. So I'm fairly confident that even with compression they'll be over 5kB

(I fix airplanes, have been in aviation for 15 years)

Yes, it does. I can't speak universally but I would generally expect it to be a written rule, if not then at least it's the expectation.

We should probably assume that the authors are aware of ceramics, Curie points, and the earth magnetic field. While the article isn't very clear on what effect causes the change in temperature needed to erase the VRM. I suspect is has something to do with how long the object has spent stationary with respect to the earth's field. At least that's my takeaway from having to periodically rotate magnetometers to prevent them from acquiring a bias.

I'm reminded of thermoluminescence for dating objects. Radiation creates defects in the material over time. When heated the defects release light.

Imprecise but will fink on ceramics that were fired 30 years ago vs 3000 years ago. Difficulty you need to heat a fair sized sample.

"Reads article"

Impression, author didn't understand any of what he was told.

But yes similar, over time the object picks up a faint magnetic field superimposed on the one locked in when the ceramic was made.

StimDeck, for autism on the go.

Neuromancer had stims and cyberdecks... I guess you've brought it full circle.

AI inference is basically 1 operation iterated over every single model weight. That's what's makes computer in memory attractive and feasible.

General purpose workloads would require much more silicon real estate and are not (typically) nearly as bandwidth hungry. Just like not all CPU workloads benefit from being adapted to a GPU, even fewer will benefit from being adapted to processing in RAM.

AI is not intrinsically matmul. Current models are a framing of the problem as matmul to enable the bulk operations necessary. There are many calculations that can (and have been) reframed into matrix form to achieve similar results. More than once I have seen a matmul version of something that I thought couldn't have been representable that way. I'm not sure if the greater benefit is the performance or the ability to conceptually manage the problem in you head.

Not all algorithms can be converted of course, but a lot of that is because there are some inherently unparallelizable problems.

Many algorithms can be reasonably converted into matmul. Not many algorithms can be reasonably converted into very low precision matmul. And that's all that these things are going to do.

The signature doesn't touch the image data in any way. It's just a piece of metadata attached to the image, like any other metadata tag.

They did, they contracted with a vendor to store and backup the data. If that is insufficient then just about every customer of AWS, Azure, GCP, OCP, and every other managed storage provider is guilty of the same sin.

I have to agree with GP here, this is pretty incredible. They contracted with a vendor to store their data, not to back it up. It's not a backup if you can lose it in the same instant you lose the data.

You're right, they are. If your data can't survive a single storage host vanishing from existence, you don't have a backup. If your data can't survive one predictable or regularly occurring catastrophic act of nature, you don't have a backup. If your data can't survive a piece of malware -- with all of the credentials you have -- erasing it, you don't have a backup.

EDIT: Also, if you don't regularly test that your backups actually work, you probably don't have a backup. Lots of companies learn that one the hard way.

If you do not have at least one, but usually two copies of a backup in your physical possession, you do not have a backup. |o-o|

> If that is insufficient then just about every customer of AWS, Azure, GCP, OCP, and every other managed storage provider is guilty of the same sin.

Yes, they are as guilty! You say that as if it's a ridiculous assertion. Have you not read the occasional HN submission where someone is locked out of their cloud provider for dubious reasons?

A single vendor to handle the primary storage and all backups is insufficient for your most critical data.

But where are you seeing this vendor was tasked with backups? This looks like a single copy situation.

That's not a backup, that was their only copy of the data.

2 is one, and 1 is none.

I wouldn't hire someone with this mindset of yours to run the IT at my company, we'd lose data for sure.

Then that is just the 1 in 3-2-1 backups, and neglecting the 3 & 2 is a sin (wrt backup strategies).

As a whole, I'd say it's for better. Obviously Nine PBS's situation is shitty and fuck "OSS" for putting them in that position, but they are not Iron Mountains customer. They are strangers to Iron Mountain who just demanded 50TB of customers data from Iron Mountain. The only sane response is to tell Nine PBS to come back with a court order, and (hopefully) to tell their lawyers to just make sure Nine PBS is in fact the rightful owner and if so, not to fight them on it.

Anything else is the sort of stuff that causes blog posts and news articles about data breaches to be written.

Yeah; I haven't read TFA but IMO especially if Iron Mountain is aiding the procurement of said court order through giving some (potentially redacted as required for privacy purposes) "sample" court documents that could be readily used as reference so the lawyer on the Nine PBS side (filing with the court to get a judge to order Iron Mountain to release that data to Nine PBS) doesn't need to waste their time repeating all the drafting from scratch for what I'm assuming would be a (from the legal perspective) straight-forward request (that nonetheless has to specifically account for the particular legal data possession/ownership/access-rights situation under which the data in question is stored at Iron Mountain), especially as relevant for the situation at hand allowing Nine PBS to put a temporary data retention legal hold on these 50TB giving them a fair deadline (not merely "legally 'reasonable'") either on charity (understanding the gravity of the "this Nine PBS data is a public good (at a cursory glance)"/"information wants to be free") or offering general reasonable list price of what they bill for retaining 50TB stored under the storage model at play for "just one more month" (or whatever fair deadline) paid by Nine PBS.