What if the tool you use, and your ability to use it, is all down to somebody else? What you think you hold in your hand, that you’ve bought and paid for, can only be used if and when other systems are in place and up and running.
What if you bought a thing – a tool, a handsaw, a wrench, a hammer – that only works 99% of the time? The other 1% of the time, you have to put down your tools and wait. Wait until somebody else does something to fix the problem.
Because that’s where we are with AI and all this online automation.
The 99.5% Problem
Take today as an example. Claude AI has been hit and miss since 8.34 this morning. Intermittent service for two and a half hours. If you were trying to work, you couldn’t. If you needed information, you waited. If you’d integrated this into your workflow, you were stuck.
Over the past 90 days, this platform has averaged 99.5% uptime. Sounds impressive until you do the maths. That’s roughly 36 hours of downtime in three months. A day and a half where the thing you’re paying for simply doesn’t work.
And Claude isn’t unusual. Major cloud services promise 99.99% uptime – the famous “four nines.” But even that means 52 minutes of downtime per year. When you’re one of millions relying on it, 52 minutes can be catastrophic.
When the System Fails
When these systems go down:
- You can’t do your job
- You can’t get the information you need
- A diagnosis can’t be made
- An answer can’t be reached
- A solution can’t be implemented
It’s alright saying that tech is amazing. It’s alright having these absolutely incredible systems. But if they throw out errors, if they give incorrect information, if they’re only available 99.5% of the time, how reliable are they really?
The Medical Reality
Here’s where it gets serious. AI has been touted as this revolutionary tool that can assist doctors in providing medical care. Diagnostic support. Treatment recommendations. Patient monitoring.
Imagine a doctor relying on an AI system to help diagnose a patient, and the system’s been temperamental all morning. Do they wait? Do they guess? Do they make decisions without the tool they’ve been trained to depend on?
A hammer doesn’t need a data centre in California to work. A saw doesn’t require stable internet. If your traditional tool breaks, you fix it yourself or grab another one. You’re in control.
But if somebody trips over a wire in Silicon Valley, or knocks over a server, or accidentally unplugs something, or types the wrong command in the software, hundreds of millions of people are quite literally left in a difficult situation while they wait for someone else to fix it.
The Scaling Problem
Right now, these platforms are handling millions of users. In the coming years, that number’s going to multiply. More users. More dependency. More critical applications. More points of failure.
The infrastructure is impressive, don’t get me wrong. What these systems can do when they’re working is genuinely remarkable. But we’re building entire workflows, entire businesses, entire medical processes on foundations we don’t control.
When a traditional tool fails, you’re inconvenienced. When a cloud-based AI system fails, you’re helpless.
The Control Question
The real issue isn’t that these systems occasionally fail. Everything fails eventually. The issue is that when they do, you have absolutely no ability to fix it yourself. You’re entirely dependent on someone else, somewhere else, to sort it out.
You can’t pop the bonnet. You can’t check the connections. You can’t even see what’s wrong. You just have to wait and hope it gets fixed quickly.
Maybe we should go back to hammers and handsaws. At least when they stop working, it’s because you broke them, and you can fix them.
Or maybe we need to be a lot more honest about what we’re building, what we’re promising, and what happens when the 0.5% failure rate lands at exactly the wrong moment.
I’m still here, using AI to refine this after waiting for the system to come back online. You’re still here, reading it, probably after experiencing your own frustrations with technology that didn’t work when you needed it to.
We’re both still getting through it. Finding workarounds. Adapting. Making do.
And tomorrow, when it happens again, we’ll still be here, won’t we?
A Quick Note About These Nostalgia Posts
I’ve been digging through old notes, thoughts, and half-finished rants from years back, turning them into something readable. Some of it’s rough. Some of it’s raw. But it’s all real.
If you’ve found this piece useful, frustrating, or just a bit too relatable, there are more where this came from. Thoughts on technology, work, mental health, and everything in between. The unfiltered stuff that doesn’t usually make it to a polished blog.
Because sometimes the unpolished thoughts are the ones worth sharing.
FAQs about AI and Automated Systems
What does 99.5% uptime actually mean?
99.5% uptime means a system is unavailable for approximately 36 hours over a 90-day period. That’s a day and a half where the service simply doesn’t work, leaving users unable to complete their tasks.
Why is AI reliability important in healthcare?
Medical professionals are increasingly using AI for diagnostic support and treatment recommendations. When these systems fail, doctors may be unable to access critical decision-making tools, potentially affecting patient care.
Can’t cloud services just improve their reliability?
Even the best cloud services targeting 99.99% uptime still experience about 52 minutes of downtime per year. As user numbers multiply and critical applications increase, even small percentages of downtime affect millions of people simultaneously.
What’s the main difference between traditional tools and cloud-based systems?
With traditional tools, when something breaks, you can fix it yourself or source a replacement. With cloud-based systems, you’re entirely dependent on someone else to resolve the issue, with no ability to diagnose or repair it yourself.
Is this just about AI systems or all cloud services?
This applies to any cloud-based system we depend on for critical work. AI platforms, productivity tools, communication systems – anything that requires external servers and infrastructure beyond your control.