Civil aviation infrastructure has been tragically underfunded for decades. Maybe it kicked off when Reagan fired striking air traffic controllers, but a recent article was talking about how a lot of control towers still run off floppy disks.
But maybe that’s also partially explained by “if it ain’t broke, don’t fix it” and tried-and-true is worth more than the latest technology. I mean, jeez, imagine if two planes collided because Windows 11 insisted on installing an update…
Sure, Trump’s probably been the worst of the lot since he’s a Russian plant out to destroy the country by all means possible, but he’s only the head of the pack.
The floppy disk thing is 100% because they had a system which worked, and replacing it would be difficult and expensive to do properly and safely.
Air traffic control has to be designed and built just as carefully as the aircraft themselves, and if a mistake is made it could kill hundreds of people. In that kind of environment, “if it ain’t broke” is a really strong argument.
The floppy disk thing is 100% because they had a system which worked, and replacing it would be difficult and expensive to do properly and safely.
A thumb drive with an emulator of the floppy protocol would be dirt-cheap, and almost certainly more reliable, since floppies are among the shittiest storage media that exist. This is instead the result of deliberate resource starvation and consequent accumulation of decades of technical debt.
Air traffic control has to be designed and built just as carefully as the aircraft themselves
Should be. I worked on ATC systems for over a decade in an earlier phase of my career. Your assertion is only true if the aircraft you’re referring to are DC10s. The languages used, the methodologies used, the protocols, and the hardware and comms systems were no better engineered than anything else from that era. The big difference is process: they test the hell out of those systems before deployment, and build in several layers of redundancy and failover capability. They also regression test any new software releases to an extreme degree, give major attention to rollback, and extensively test that too. And that’s also what drives up the cost of such systems.
At least, that’s what they used to do, and are supposed to do now.
In this case, my guess is that Trump and/or DOGE have cut the staff that did that validation, or may havereplaced them with engineers who didn’t know dick about safety-critical systems. This also may be another case of sabotaging an agency to create a pretext for its privatization.
Also, if they were trying to fit LLM-based AI into the ATC system, the people responsible for that decision should be jailed for life. LLMs are inherently unfit to be used in any decision affection people’s lives, health or livelihoods. If it’s ML, there are some cases where that might be OK, subject to the rigorous testing and process controls I’ve already mentioned.
Civil aviation infrastructure has been tragically underfunded for decades. Maybe it kicked off when Reagan fired striking air traffic controllers, but a recent article was talking about how a lot of control towers still run off floppy disks.
But maybe that’s also partially explained by “if it ain’t broke, don’t fix it” and tried-and-true is worth more than the latest technology. I mean, jeez, imagine if two planes collided because Windows 11 insisted on installing an update…
Sure, Trump’s probably been the worst of the lot since he’s a Russian plant out to destroy the country by all means possible, but he’s only the head of the pack.
The floppy disk thing is 100% because they had a system which worked, and replacing it would be difficult and expensive to do properly and safely.
Air traffic control has to be designed and built just as carefully as the aircraft themselves, and if a mistake is made it could kill hundreds of people. In that kind of environment, “if it ain’t broke” is a really strong argument.
A thumb drive with an emulator of the floppy protocol would be dirt-cheap, and almost certainly more reliable, since floppies are among the shittiest storage media that exist. This is instead the result of deliberate resource starvation and consequent accumulation of decades of technical debt.
Should be. I worked on ATC systems for over a decade in an earlier phase of my career. Your assertion is only true if the aircraft you’re referring to are DC10s. The languages used, the methodologies used, the protocols, and the hardware and comms systems were no better engineered than anything else from that era. The big difference is process: they test the hell out of those systems before deployment, and build in several layers of redundancy and failover capability. They also regression test any new software releases to an extreme degree, give major attention to rollback, and extensively test that too. And that’s also what drives up the cost of such systems.
At least, that’s what they used to do, and are supposed to do now.
In this case, my guess is that Trump and/or DOGE have cut the staff that did that validation, or may havereplaced them with engineers who didn’t know dick about safety-critical systems. This also may be another case of sabotaging an agency to create a pretext for its privatization.
Also, if they were trying to fit LLM-based AI into the ATC system, the people responsible for that decision should be jailed for life. LLMs are inherently unfit to be used in any decision affection people’s lives, health or livelihoods. If it’s ML, there are some cases where that might be OK, subject to the rigorous testing and process controls I’ve already mentioned.
A couple of facts I can give you, interpolate what you will:
The FAA comms network has been in place since 2013. Those routers were working fine yesterday.
AI gets released, routers no work anymore.
The AI SMART system does not “do” anything, it makes recommendations to ATCs.
My pretty well-informed opinions:
The AI system most likely uses the comm network to distribute those reccos.
It was probably overloaded, with cascading impacts.
Edit: Oh look, I was right. https://en.wikipedia.org/wiki/2026_FAA_ground_stop