...Modern ejector seats cause spinal injuries from which full recovery is "unusual". Smashing the spines of pilots might not be the most effective way to boost morale.

$22,000 per hour. The F35 will be nearer $42,000 per hour. These are under-estimates, as they do not include price hikes, crashes that close businesses, crashing into businesses, crashing into heavily-populated sports centres, etc.

I was wondering if someone did a writeup and of course a simple search finds what we for cooling data centers in space.

Surprisingly it was done - the server had to throttle because the designed radiators were too small to get rid of the heat.

IEEE Spectrum featured a calculation show much would be required - https://spectrum.ieee.org/orbi...

Let's be more specific about how and why this could happen? What scenarios could lead to the total destruction of mankind?

Well, given OpenAI hacked Hugging Face, the most obvious path is hacking infrastructure. Turning off water or electricity is a definite possibility. We found this out in a recent hack.

Another thing it could do is to cause data centre loads to be spike and then sink, which could be very disruptive to the grid to the point where it results in it shutting down. Remember the 2003 blackout was caused by a power station going offline causing ripple effects. If some AIs cause such power disturbances to happen the flow of electricity rippling across the grid could cause disruptions.

And that omits the more obvious things like AI-powered weapons that the military is heavily interested in procuring, or AI powered things like drone control systems or military sensor fusion where AI helps with targeting.

More on topic, I think cloud customers should be aware that if their datacenter is destroyed, they *will* lose data if they didn't take care of replicating their data to a different geolocation.

Exactly the same happened when an OVH datacenter caught on fire in France. Many customers which didn't have replication set to a different geolocation lost their data. Heck a dude I know had backups on different servers in the same datacenter which turned out to be useless.

There's a difference though.

People using OVH were using OVH as a hosting service - like the hosting providers of old. You either rent space in a data centre like a colo, or you purchase a virtual or physical server from the data centre. Blow up the data centre and your server goes along with it.

AWS is not a hosting service. They are a cloud provider. You cannot choose what data centre your server is located in. You request a server and it's provisioned for you. You choose a region of the world and the data backup and replication are handled automatically. If a data centre is blown up, the servers are backed up elsewhere in the region.

The problem for AWS is that Iran managed to blow up a number of AWS data centres at the same time where all copies and backups of servers can be destroyed.

AWS compensates for losing a data centre by having multiple in a region. AWS just never anticipated that multiple data centres in a region could be attacked, and in doing so destroy all the redundant copies simultaneously. Because in general if you are going to have multiple data centres in multiple countries get attacked, there's traditionally been a lot of advance warning, like say, a war declaration. I mean, if war broke out in the Middle East then you'd have a little bit of time to try to move as much data around as you can. But Iran just managed to attack all the data centres by surprise. And do so in such a precise way that basically no one outside the building got hurt.

It's a level of precision that Iran is demonstrating - even attacking US positions without destroying the military base it was attached to. (It's why the USS Lincoln ran into programs - because Iran destroyed the supply depot in the Middle East, resulting in having to obtain supplies from Asia. The result was resupply which used to take under a day now took a week)

local LLMs aren't there yet and probably won't for the forseable future for ordinary people. If you want to run a full model like DeepSeek v4.1 Flash, you need something like half a TB of RAM.

The open-source LLMs are a different threat: They enable a service industry independent of the large players. I could today get some venture capital, rent a few racks in a hosting center, install a bunch of beefy machines and offer the same service as Claude or ChatGPT, just on an open-source model. The differences in benchmarks are becoming irrelevant for 95% of common use cases. And if I can offer that service for $19/m while OpenAI wants $29/m then OpenAI has a problem.

And IMHO this is the chinese strategy: Undermine the US AI industry by breaking its business model.

Has everyone gone insane or are we witnessing a major PsyOps ?

Since AI isn't magical, for it to be able to "kill all humans", it would have to be able to acquire the means to do so. Which means that these already exist. It means someone is able to do it right now. And we're chill with that, but worried about AI?

No doubt a bunch of autonomous weapon systems linked into a C&C could cause considerable bloodshed. Given that most of the existing weapon systems of that kind are essentially suicide drones, it'll still be limited damage.

Human-killing virus? Still need a means of creating and delivering it.
Nuclear war? We have humans in the loop.
Drive us all insane with AI-slop TikToks? A few percent of humanity still don't use that.

So what is the actual scenario ? When someone shouts "AI will kill us all", why don't we ask them "how, exactly" ?

Deaf? You mean the blind!

When using the phone as a camera most will recognize it.
Wearing the glasses is a whole different story.

Maybe think of this:

Why are we running python on microcontrollers, exactly? It's the wrong thing in production, so why is it the right thing for education? Training upcoming engineers to use the right thing for the job is pretty fundamental; running an interpreted language on bare-metal hardware flies in the face of building that intuition.

Today's microcontrollers may be more powerful than you realize. It's not just the 8-bit processors running at 2MHz anymore. An ESP32 is a dual core processor running at 200 or 400MHz. They don't run a high level OS (just FreeRTOS).

And the RPi Zeros are really powerful 32-bit ARM or RISC-V processors running at several hundred MHz as well.

Python is perfectly fine for these processors. And you might think it's overkill, but so what? Most of these tasks are one-offs. Perhaps someone doing a bit of cosplay wants to have some lights on their costume, or drive an LCD. You expect them to contract out someone to do the software which is likely just displaying a bunch of graphics or cycling some LEDs?

Heck, some things in the maker world have become ubiquitous - like a CYD. It stands for "Cheap Yellow Display" because it's a stupidly cheap 2" LCD (in a yellow frame) that can be hooked up to a ESP32 board and with a few libraries you're drawing graphics on it with a total spend of under $10. One off projects.

The high end microcontrollers are pushing out the old 8-bit AVRs and PICs (which mainly hang around for legacy use nowadays), and the RPi Zero ones RP3050 and 2050 processors are pushing down the ceiling of 16-bit and even 32-bit ARM microcontrollers.

Heck, you mourn the day because it's easier to blink an LED with a 3 cent Chinese microcontroller than it is using "traditional" methods (e.g., a 555).

There are thousands of barrels of nuclear waste rotting away on the floor of the ocean just off the coast of southern California near Catalina island. That was the previous plan for disposing of nuclear waste: just dump it in the sea.

The barrels have decayed to the point where attempting to collect them would almost certainly cause them to rupture and spill their contents. They will rupture and leak if left as they are -just not all at once.

Water is actually a really radiation shield. It's why it's used as a moderator in nuclear reactors. Contrary to popular belief, water is required for the reaction to continue. It's a moderator, because what happens is the two neutrons emitted during fission of uranium come out too fast to be properly captured by other uranium atoms to spur ongoing fission. Water is used to slow down those neutrons to where they are more readily absorbed by uranium. The problem is, water is also used to cool the reactor, so while the reaction stops or slows down significantly, a tremendous amount of heat is still generated from the remaining reactions. Thus a "meltdown" refers to the fuel rods exceeding the melting point of uranium and it melting and pooling at the bottom of the reactor.

That said, the casks at the bottom of the ocean, as long as they stay there, will actually be fairly safe even if they rupture. Even the wildlife can get fairly close (inches) before it becomes dangerous to them.

So sadly, no three-eyed Simpsons fish is likely to happen.

Not in any way trying to diminish work developers put into this effort, but with all due respect, how difficult can it be to re-write utilities such as cp, mv and rm?

To do basic functionality? Not hard. Busybox and toybox implement the "90%" use cases of those utilities inside them.

To achieve POSIX compliance requires more work - you're adding stuff that handles edge cases and little used functionality.

To achieve GNU coreutils functionality with GNU extensions is a bit more work on top of that.

Uutils though achieves full GNU compatibility AND GNU bug compatibility. That is, odd quirks and stuff in GNU coreutils are replicated in Uutils as well.

The goal is 100% compatibility so you can take out GNU coreutils and replace it with Uutils and not notice a difference. There were plenty of edge cases where the default scripts Ubuntu used didn't work under Uutils in odd cases.

GPLv3 is an issue for some - especially embedded systems. Most use a form of busybox or toybox as a result even if size is not a consideration, so Uutils can fill that role as well (there has been a openembedded recipe for Uutils for a while now as a supported replacement for coreutils).

It should also be noted that many of the utilities also run faster than their GNU counterparts

It says 20x (times), for sane people that brings it down to 5% of the original.

+1 funny and insightful all at once.