Over the past few weeks, I’ve noticed something strange. My apps are freezing mid-scroll, features quietly breaking after updates or my phone just needs to be entirely shut down. The tools I rely on every day suddenly feel broken or need to be rebooted more often.
If you've noticed this too and if you've been paying attention to the news lately, you know exactly what I mean. Just days ago, a routine Windows 11 update broke the Remote Desktop feature for anyone using multiple monitors with different scaling, a bizarre UI glitch that resulted in overlapping text and hidden buttons. Worse, that same April update was linked to sudden boot loops on Dell and HP laptops.
Then there was the massive global Microsoft Outlook outage at the end of April. As millions of users were locked out of their emails, Microsoft pushed a fix, only to publicly admit a few hours later that the fix "did not appear to have provided the intended relief," and they were still trying to find the source of the errors.
I've also noticed that just before or after a major model update, chatbots go down. Earlier this week, Claude was down after some feature updates.
Is AI to blame? Well, there's a growing theory, that has been covered by the New York Times and in studies, that while AI is helping build software faster, it's the very thing making it more fragile.
The speed boom no one is questioning
I recently read an article by developer Mario Zechner that helps to put all of this into perspective. Essentially, AI coding tools like ChatGPT and Claude have completely changed how software gets built, making the process faster than ever.
Developers can now generate entire functions in seconds, ship updates faster than ever and automate large chunks of their workflow. In many ways, it’s a breakthrough because what used to take hours or days, now can happen almost instantly. But speed comes with a tradeoff, and we're starting to feel it.
Faster code, weaker understanding
The problem is, AI is replacing human intelligence for speed. AI writes the code and developers don’t always fully understand it. That’s not a knock on developers, it's the nature of the output.
AI can easily produce code that looks correct and introduce subtle bugs that are hard to trace. It can also generate logic that technically works, but isn’t well structured. So, multiply that across thousands of updates, and small issues start stacking up. The result is software that works, but not very well, then ultimately breaks.
The rise of “slop” software
Instead of big, obvious failures, we’re seeing something random glitches, inconsistent behavior, features that break after unrelated updates. Sure, these aren't catastrophic, but I don't know about you, but I like my software to be reliable.
That's why some developers have started calling this "slop." Essentially code that is generated so fast by AI agents that it becomes a bloated, tangled mess. It holds together on the surface, but cracks under pressure.
And if you’ve caught yourself thinking, “Why is this app suddenly so weird?” well, this might just be the new norm. The reason this is happening now is because we're seeing how AI changes behavior. When building software becomes faster, more code gets shipped and less time is spent reviewing it. Fewer bottlenecks slow things down.
In the past, human exhaustion acted as a natural speed limit for bugs. We get tired, we sleep and we learn not to make the same mistake twice. But the mistakes AI makes are compounding at lightning speed. Bottlenecks used to be a form of quality filter, but that friction is disappearing along with the oversight.
Bottom line
I’ve spent a lot of time using AI to speed up my own work, and I've felt this firsthand. And the faster I move, I've noticed it's easier to accept an answer without fully checking it or move on before fully understand it. Building on top of something slightly flawed is also more possible than ever. Speed isn't the real advantage all the time and we're seeing that.
But to be clear, this doesn't mean AI is bad for software. Far from it. It means we’re in a transition phase and the industry is still figuring out how to maintain quality without the traditional human bottleneck.
What do you think? Have you seen more glitches lately? Let me know what you think in the comments.
