Advertisement
•

BREAKING: Everything

We are seeing the XKCD dependency domino tower, but for everything.
BREAKING: Everything
Image via Unsplash

A few hours ago, I went to rent a Lime bike to ride the ten minutes to my therapy appointment. I opened Lime, scanned the QR code, and double clicked my iPhone’s side button to authorize Apple Pay. The app hung for a few moments, then closed. I tried several more times and the same thing happened. In the back of my mind, I wondered what was going on — was the Lime bike broken? Was my cell phone service out? Was it something else? I opened the Uber app and, similarly, wasn’t able to pay. I opened DownDetector and realized that the entirety of Apple Pay, a thing I thought existed only on my phone and not the cloud, was down. I texted my therapist, put my phone in my pocket, and started jogging. 

Not being able to rent a Lime Bike for several minutes is an extremely minor inconvenience barely worth mentioning. And yet the mental gymnastics my brain did in those few minutes — what’s broken, why is it broken, how is it broken — is something that everyone I know have been repeatedly doing for the last few weeks as seemingly every service and system crashes down around us. Over the weekend, people who I do not normally talk to about my job asked me: What is going on with the airplanes? Why is my bank down? What about Microsoft? 

The default assumption that many people are making is that these are either being caused by cyberattacks and/or “AI” broadly defined, whether that’s AI gone haywire, “rogue” AI agents, or something else, which is indicative of just how fearful people are of AI. What I have been telling people is, basically, that shit is broken and things are going to keep breaking. For lots of different reasons. It’s often not a hack and it’s often not explicitly AI-related, but everything now sort of looks like that XKCD comic of the dependency domino tower. 

While not exactly true that AI and cyberattacks are causing these outages, it’s also not entirely wrong, either. What we’re seeing is — I think — a series of infrastructure failures (and some cyberattacks) that are not directly connected but are broadly a symptom of the same general trends and systemic problems we’ve seen across society, the government, and the tech sector over the last several years. (WIRED has a nice piece on how you can assess any given outage.)

First, it’s worth mentioning that anytime there’s a string of outages, there’s a tendency to tie them altogether. And outages or hacks that previously wouldn’t have been noted at all suddenly get more attention and become part of the narrative. Take, for example, the fact that the International Ice Hockey Federation has been targeted by a cyberattack; you probably would not normally hear about this but it has suddenly become part of the broader story. 

The simplest explanation I can come up with for the broader disastrous trend we’ve seen is that since the pandemic we have seen: 

  • mass layoffs of tech workers and software developers 
  • mass deployment of AI-assisted and vibe coded products, software updates, and patches
  • mass layoffs in the government sector and outright destruction of government expertise and oversight
  • the loss of institutional knowledge associated with those layoffs and funding cuts
  • a failure to invest in upgrading infrastructure/redundancy
  • government contracts handed out to the President’s friends and political allies
  • AI and LLM-assisted/iterative exploit discovery, coding, and deployment
  • companies responding to the AI hacking threat with AI-coded and rushed patches

When Elon Musk took over Twitter and named it X, there were many people who thought his layoffs and the associated loss of expertise would result in the site failing basically because of an infrastructural failure. Twitter would basically “go down” because someone would break something and no one would know how to fix it. We haven’t really seen that happen to X, but that same philosophy of firing the experts and YOLOing AI into everything while hoping shit doesn’t break is now the explicit strategy of our entire economy and our entire government. 

What we are seeing, then, is a mix of tightly clustered attacks, outages, and fuckups that you would expect to see when people and companies are careless, don’t know what they’re doing, or are working very fast to respond or duct-tape something together in the face of new threats. 

"We are Ship of Theseus-ing our entire economy and infrastructure"

We know that this is happening to some extent because, in the aftermath of the PHP2Shell exploit, which was discovered with the assistance of LLMs, we saw various websites go down as they attempted to push updates faster than they normally would have to, because the time between a vulnerability being discovered and it being exploited is faster than ever. (As I was writing this, Calif, one of the companies that used LLMs to find PHP2Shell, announced that it had used LLMs to find apparently serious vulnerabilities in Java.)

Companies and agencies are pushing updates (including to, for example, our air traffic control system) on live systems, in part because of the threat of AI, and things are breaking. The National Security Agency, for example, said earlier Tuesday that some internal systems went down because of “temporary degradation during scheduled routine maintenance.” 

Nick Andersen, the acting director of the Cybersecurity and Infrastructure Security Agency, said Tuesday that we need to build more resilience into our infrastructure, but building resilience and redundancies into software requires people who understand how systems can possibly fail, planning for it, and mitigating against it. 

“Ultimately being able to prevail by being able to demonstrate resilience against disruptions, whether they occur because of things like an IT outage or a bad patch, or a modernization update gone awry, or because of a malicious malicious cyber incident — we want to make sure we’re engineering in the resilience that is going to be able to anticipate for and ultimately overcome those outages and those impacts,” Andersen said at an event called FedTalks. 

We may eventually get to a place where expertise and understanding highly complex systems is once again a priority, but the last few years have been full of companies and governments moving incredibly fast to reduce reliance on humans and offload it to AI systems and AI-generated code in order to ostensibly save money on labor costs and/or carry out an antihumanist political project. 

I thought this piece, called “Software is breaking down at a systemic level,” written by a software developer named Jimmy Breck-McKye, was quite good. He points at Kristen Maevyn’s The Decoherence Trap, which argues that AI-written code is causing a situation where people and companies do not understand their products. 

“Most software engineers now are unable to tackle engineering problems without some form of AI assistance. That’s not because they’re incompetent — they just can’t keep on top of the complexity,” Breck-McKye writes. “Even if they could once, they now face a tsunami of AI generated code. This is affecting everyone, at all levels — we are all fighting the tide of small, incoherent changes that look rational in isolation but slowly lose sense as a whole.”

People increasingly don’t understand or have any idea how their code works, people are moving very fast, and bad things are happening. Any individual outage or breakdown is likely, to some extent, a coincidence. Many of the outages we are seeing have nothing to do with AI at all, but do have something to do with the broader lack of investment in infrastructure, redundancy, and expertise. We are Ship of Theseus-ing our entire economy and infrastructure and so, I guess, we must simply continue to expect bad things to happen.

Advertisement
•
•