It's surprising that this restriction continues to linger at all. The newest nuclear warhead models in the US arsenal were developed in the 1970s, when supercomputer performance was well below 1 gigaflop. When the US stopped testing nuclear warheads in 1992, top end supercomputers were under 10 gigaflops. The only thing the US arsenal needs faster computers for is simulating the behavior of its aging warhead stockpile without physical tests, which is not going to matter to a state building its first nuclear weapons.
It wouldn't be a solution for a personal existential dread of death. It would be a solution if you were trying to uphold long term goals like "ensure that my child is loved and cared for" or "complete this line of scientific research that I started." For those cases, a duplicate of you that has your appearance, thoughts, legal standing, and memories would be fine.
I have been trying to catch up with recent OCR developments too. My documents have enough special requirements that public benchmarks didn't tell me enough to decide. Instead I'm building a small document OCR project with visualization tools for comparing bounding boxes, extracted text, region classification, etc. GLM-OCR is my favorite so far [1]. Apple's VisionKit is very good at text recognition, and fast, but it doesn't do high level layout detection and it only works on Apple hardware. It's another useful source of data for cross-validation if you can run it.
This project has been pretty easy to build with agentic coding. It's a Frankenstein monster of glue code and handling my particular domain requirements, so it's not suitable for public release. I'd encourage some rapid prototyping after you've spent an afternoon catching up on what's new. I did a lot of document OCR and post-processing with commercial tools and custom code 15 years ago. The advent of small local VLMs has made it practical to achieve higher accuracy and more domain customization than I would have previously believed.
[1] If you're building an advanced document processing workflow, be sure to read the post-processing code in the GLM code repo. They're doing some non-trivial logic to fuse layout areas and transform text for smooth reading. You probably want to store the raw model results and customize your own post-processing for uncommon languages or uncommon domain vocabulary. Layout is also easier to validate if you bypass their post-processing; it can make some combined areas "disappear" from the layout data.
Any country capable of producing nuclear warheads will also be able to toss up enough BBs and other small objects into LEO to wipe out most of Starlink and anything else in LEO.
Pakistan doesn't have a domestic orbital launch capability but it does have nuclear weapons.
Surprisingly, the United Kingdom doesn't have a domestic orbital launch capability at present though it has had ballistic missiles and nuclear weapons for many decades.
At present, I would say that building a basic implosion-assembled atomic bomb is easier than building a rocket system that reach low Earth orbit. It's a lot easier to build a bomb now than it was in the 1940s. The main thing that prevents wider nuclear weapon proliferation is treaties and inspections, not inherent technical difficulties.
You don't need orbital velocity to blow satellites away. Just do a well-timed suborbital launch against the satellite's orbit, and the satellite will provide most of the kinetic energy.
But I shouldn't have to. That's the issue here. I shouldn't need an ID of any sort. I shouldn't need to provide my name or date of birth. Either I have weapons or dangerous substances on me or I don't. That's all that should matter.
Twenty-five years after the ISS began operations in low Earth orbit, a new generation of advanced solar cells from Spectrolab, twice as efficient as their predecessors, are supplementing the existing arrays to allow the ISS to continue to operate to 2030 and beyond. Eight new arrays, known as iROSAs (ISS Roll-Out Solar Arrays) are being installed on the ISS in orbit.
The new arrays use multi-junction compound semiconductor solar cells from Spectrolab. These cells cost something like 500 times as much per watt as modern silicon solar cells, and they only produce about 50% more power per unit area. On top of that, the materials that Spectrolab cells are made of are inherently rare. Anyone talking about scaling solar to terawatts has to rely on silicon or maybe perovskite materials (but those are still experimental).
Crystalline silicon solar panels have about 95% market share, and "By weight, the typical crystalline silicon solar panel is made of about 76% glass, 10% plastic polymer, 8% aluminum, 5% silicon, 1% copper, and less than 0.1% silver and other metals."
Everything that is manufactured is made out of atoms, and you can say that any manufacturing requires some nature destruction in the aggregate. But solar electricity requires far less mining and natural despoilation than fossil-fueled electricity.
Solar panels contain quite a bit of lead, and small amounts of cadmium. Lead can be taken out if you're willing to pay a bit more, in other words it never is. Cadmium is required. Other metals are sometimes present.
So solar panels are classified as hazardous waste.
Cadmium is only required in cadmium telluride solar panels, which have less than 5% global market share. Lead solder is still common in crystalline silicon panels, though not universal; modules built with heterojunction cells typically avoid solder because the cells can't tolerate temperatures that high:
The difference is that the Allen Institute models have open training data, not just open code and weights. Meta doesn't share the training data you would need to reproduce their final models. For many uses open-weight models are nearly as good, but for advancing research it's much better to have everything in the open.
Reading their paper, it wasn't trained from scratch, it's a fine tune of a Qwen3-32B model. I think this approach is correct, but it does mean that only a subset of the training data is really open.
The massive build-up they have is mostly renewables. Surely, you see the problems with that, right? Georgia is a red state, so it's political suicide to even hint at proposing that.
Large scale solar power generation has more than doubled in Georgia since 2020:
Texas is number 2, behind only California. Solar power is popular in sunny states even if they're "red," though the most heated political rhetoric doesn't reflect that.
Huh, I didn't realize how far the build up had gone.
Your second link is interesting, though, because it shows solar in Georgia took a nosedive in 2025. I've got a feeling that that year's data is much more representative of what it will look like in the next two or three decades than any historical trend might be.
I still really dont see how solar or wind power the future needs at all. surely nuclear is the only solution longer time. obviously it has to be made safe but why are wasting so much time and money on solar and wind that are demonstably not good for the environments they go into. at scale that is going to be felt because no, actually deserts are not "just empty spaces doing nothing" they have a huge knock on effect when changed either life within them, or how they feed the surrounding non-desert environments. Why is nuclear still the bogeyman when the sun is a nuclear event. cut out the middle man. surely.
There’s movement around nuclear but it takes 10-15 years to build one plant and that’s for plants that are already tested. 15-20 for something new or experimental. Even China with all its rapid construction can’t build one in less than 8. We’re not offsetting anything with nuclear anytime soon. Solar plants take 3-6 months to get up and running.
A combination of solar/renewables with nuclear is the best strategy over the long term.
Some people say that human jobs will move to the physical world, which avoids the whole category of “cognitive labor” where AI is progressing so rapidly. I am not sure how safe this is, either. A lot of physical labor is already being done by machines (e.g., manufacturing) or will soon be done by machines (e.g., driving). Also, sufficiently powerful AI will be able to accelerate the development of robots, and then control those robots in the physical world.
I would like to believe that we're about to see a rapid proliferation of useful robots, but progress has been much slower with the physical world than with information-based tasks.
After the DARPA Urban Challenge of 2007, I thought that massive job losses from robotic car and truck drivers were only 5-8 years away. But in 2026 in the US only Waymo has highly autonomous driving systems, in only a few markets. Most embodied tasks don't even have that modest level of demonstrated capability.
I actually worry that legislators -- people with white collar jobs -- will overestimate the near-term capabilities of AI to handle jobs in general, and prematurely build solutions for a "world without work" that will be slow to arrive. (Like starting UBI too early instead of boosting job retraining, leaving health care systems understaffed for hands-on work.)
> But in 2026 in the US only Waymo has highly autonomous driving systems, in only a few markets
10 years ago I predicted that the uptake of autonomous vehicles would be slow but that it would be because of labor protections. While those have had some impact, that isn't really the issue: it's that the cars just don't quite work well enough yet and that last ~20% of function turns out to be both incredibly difficult and incredibly important.
One thing that I've not quite been able to sort of get my head around about the whole AI and future of work thing ss the view around work in the physical world being safe. I don't particularly buy the rationale and not from the position of robots are going to do the work. I don't know much about robots really but from what I've seen from the more viral stuff that breaks through to mainstream internet from time to time, it still feels that we're some way out.
But that feels like the least of the worries to me. There seems to be an implicit assumption that those physical lines of work don't get eroded by the higher proportion of able bodied people who are suddenly unemployable. Yes there is some training required etc. but the barriers to entry aren't so high that in the shortish to medium term you don’t see more people gravitating to those industries and competing wages further down to not make then sustainable employment long term. I'd even think that having LLMs that can recognise photos or understand fuzzily explain questions about some blue collar skills many have forgotten actually reduces the barrier even more
reply