Almost a dupe - it left out the bit about Godzilla.

No doubt that is some of what's happening now, but I don't think it's all about stock price. I'm pretty convinced the both the memory and logic companies are going to build a bunch of new fabs, make more parts than they've ever made only to have demand drop to more sensible levels.

While the current situation seems unprecedented, semiconductor manufacturing has been a cyclical boom-bust business for pretty much forever. Ramp up as fast as possible for a year, make a giant backlog of parts the next year, lay almost everyone off, and then in another year start the cycle over. I'm glad I was able to work on the edges of the field and avoid the worst of it.

It's not just fissile material - it's anything with high Z. The paper discusses how this technique would have trouble distinguishing a satellite with a warhead from one with about the same mass of another high Z material like tungsten or lead. I guess when someone launches a solid gold satellite, you could find that too.

Well the summary sucks, but this is not proposing to look at radioactive decay from fissile material. Do you actually think something so obviously wrong would get past peer review at a journal like Nature? From the abstract of the actual article:

A GeV proton kicks off a whole bunch of neutrons from high Z materal. Spallation using high energy protons is how things like the Spallation Neutron Source at Oak Ridge work.

Previous ground based measurements report an equivalent diameter at about 450 meters. I don't think they can obtain dimensions from their new images until they nail down the satellite to asteroid distance.

Very true. I've played around with freecad and others, but they are only good enough for little one-offs.

Interestingly, the real pro IC design software - cadence, synopsis are Linux only.

Just this week, I've seen a couple of news stories that said Samsung was using quantum computers to optimize it's lithography. Computational lithography - applying corrections to the mask for diffraction, resist non-linearity, and other optical effects - is critical for making small features and has been around forever. The article claims they are employing QCs for some of this. Not sure I believe a word of these reports, but I found them interesting.

These "generational node names" have been nonsense since 22 nm or so. They mean "this is the size we would have have to make the smallest critiical dimension if we were still using the same architecture as back when we were at 22 nm". While critical dimensions have shrunk some, it's really the density that has increased in recent years (mostly due to architecture changes). This stacking is really another density increase. The smallest feature the ASML EUV steppers can print is about 8 nm.

Are you sure you aren't mixing up your recollection of QC in the 80's with the other overhyped superconductor-based thing that was going to revolutionize computing in the 80's and 90's? There was a large effort at many companies and universities to develop a conventional digital computer using niobium Josephson junction RSFQ (rapid single flux quantum) logic. It seemed very promising in its day - low power consumption, super high clock speeds (50-100 GHz). Lots of people made impressive working gates, but it mostly died off because no sensible memory was developed and interfacing to room-temperature was difficult. There is some overlap between the RSFQ and QC crowd. The first superconducting qbit (NEC 1999) was a side project of the much larger RSFQ effort there. The non-quantum QC company, DWave, uses a chip fabrication process that is very similar to RSFQ. People are proposing RSFQ as a more scalable way to control qbits.

QC as an active experimental field didn't really start until after NEC's charge qubit, so I still contend that field hasn't been trying to make things work for 50 years - more like 25-30 years. The superconducting logic people have been at it for 50-60 years and there are still some people flogging that particular horse.

The hyperbole, over-optimism, and over-investment surrounding QC certainly seem fairly ridiculous given current results. Without a new direction or some big breakthroughs, QCs in there current form seem unlikely to be useful. However, your repeated "they've been working on it for 50 years" is equally annoying. The work in the 1980's was just a few theorists making proposal about how something might work. Small groups of experimentalists first made objects showing some degree of quantum coherence about 25 years ago. Serious efforts to scale those devices started around 15 years ago. For digital computers, it did take 100 years to go from Babbage/Boole to Eniac and another 30 years to figure out how to really scale them.

While this is generally true, there are some examples where the parent company really messed up. One notable example is ASML. Philips spun them off in 1998 partly because the R&D required seemed like it wouldn't pay off. By market cap, ASML is worth more than 20x Philips right now. That's not to say I believe is the slightest that Anderon will be an ASML-like success.

It could be worse. They could be using Excel with VBA macros.

On the side of the of the earth opposite the sun, ie the earth is always between L2 and the sun. Here is a nice diagram.

BAE bought Ball Aerospace in 2024.