How AI spotted my most embarrassing oversight

Published May 15, 2026, 11:15 AM EDT I am an author and features writer at Android Police. I primarily writes guides, how-tos, and roundups on the latest smartphone apps and features for Android Police since joining the team in early 2022. While my work often diversifies depending on the topic, you'll often find me writing about the newest entertainment or hidden utility Android apps or discussing my paranoia about digital privacy and poor smartphone practices. I will occasionally dabble in Samsung's latest features in One UI. On the games side, my area of expertise is in action RPGs and gacha games, but I will play and study the occasional competitive shooter. But most of all, my appetite for new stories still goes unquenched — as shown with my personal love for the Trails series. Before joining Android Police, I studied chemistry and graduated with an honors specialization in Chemistry in 2016, leading me to spend many hours toiling around the lab as an undergraduate. Eventually, all those hours spent at the lab led me to develop my analytical mindset. So now, if you give me a problem, I will relentlessly tackle it to find a solution. My favorite pastime as a student was always writing reports, presentations, tutorials, and literature reviews, which guided me into completing a graduate certificate in technical writing in 2019. Thanks to my current role with Android Police, I learned to appreciate user security and privacy, leading me into studying Cyber Security. My first Android phone was the Samsung Galaxy Note II in 2012, which gave me a taste of how a small piece of powerful hardware can open up endless opportunities for my favorite hobbies. Though if you ask about my purchasing regrets, I will always say missing out on the Google Nexus still stings to this very day. I've also been a gamer for over 20 years, starting with Super Mario Bros. on the NES. Thanks to those early influences, I now own over 15 devices for gaming, ranging from handhelds to consoles. Nowadays, you'll find me studying spreadsheets and assembling data to theorycraft new teams and builds for Genshin Impact, Honkai: Star Rail, Wuthering Waves, and Zenless Zone Zero. You'll also see me digging deeper to uncover the truth behind AI as a field: which part is actually useful, or it just a bunch of bloat being sold with your phone? I've been using ChatGPT almost religiously when I'm troubleshooting my desktop computer. I learned a lot about how to be careful with my prompt-writing so it doesn't hallucinate or introduce bias in its answers, and how to properly redirect it toward a logical, objective answer that pertains to my situation. It's been a journey so far — nearly four months since I last worked on my computer. I have taken it to a professional for thorough hardware health checks, and some other known issues have been fixed since, but my desktop is still not where I want it to be. Since I have had a few crashes, I have been using ChatGPT as a debugger to figure out exactly what is happening at the software level that conflicts with the programs I run. Now, I feel like I am getting really close to the answers I need to optimize correctly. As a result, it has helped me expand my knowledge base, making me more comfortable learning about what's happening on the backend. While I may have been specifically dependent on ChatGPT, mainly because its memory retained information about the issues I encountered, you can technically do this with any LLM-based AI, such as Claude or Gemini. Thankfully, even when my computer is acting up, I can still access it from any platform, including my Android phone, which feels reassuring. Related Debugging with AI reminded me how much I didn't know I learned a lot by parsing my crash logs with ChatGPT Credit: Lucas Gouveia / Android Police | witsarut sakorn / Shutterstock Lately, I have had a lot of trouble understanding some terms when I asked ChatGPT to check my crash logs. This is where I asked the AI to put it into a language that made sense to someone who isn't as familiar with coding as an actual debugger. For example, I did not understand what "heap" or "racing" meant in the context of how data was being treated. But I did learn that this was a behavior and a type of data structure. That helped me narrow down to the fact that whatever was happening was performance-related, which led me to do a bit of work on the GPU while checking the CPU. With ChatGPT's help, I ended up experimenting more with my Nvidia Control Panel settings to reduce stress on my VRAM while ensuring nothing else was capping its performance. At first, I was a bit nervous about changing too many settings, so I asked the AI exactly how each setting would help with my current crash issues. And as it turns out, helping me understand what I am changing and why made me feel better, since I didn't want to necessarily try to overclock my GPU when my system had been unstable before, but none of my applied changes, like setting low latency mode on or adding a power saving to maximum performance, really changed the temperature after trying it. After trying different performance profiles and tweaking my GPU to improve performance, I still didn't get anywhere, which led me back down another rabbit hole: my installed drivers. Understanding GPU usage and driver compatibility I did not know the difference between Nvidia's drivers Credit: Nvidia Thanks to ChatGPT, I only recently learned the difference between Nvidia's Studio and Game Ready drivers. But basically, my computer is sensitive, and I need long-term stability. Since I purchased a gaming rig that I also use for a mix of other tasks, like GIMP editing, writing, and some gaming/emulation, I realized very late that the Game Ready drivers weren't going to be the best fit. Finding the sweet spot for the GPU driver to consistently perform is very important, since I only use an Intel F-type processor with a discrete GPU and no internal GPU to fall back on. After my chat with the AI, I decided to dig a bit deeper with these driver types. I learned that if your PC is primarily for playing new AAA releases, the Game Ready drivers are designed to optimize them on day one. The problem is that this set of drivers is updated more frequently and often includes experimental features that can help or hinder your GPU experience. As a result of my learning experience, I ended up using DDU to remove Nvidia's Game Ready driver and opted for the Studio driver released in April 2026. When ChatGPT implored me to try it, I was skeptical at first. I thought maybe it was out to lunch because I sometimes do around 70% gaming on my desktop and 30% work-related tasks. So I prodded the AI for a while, asking it how it reached that conclusion and to break down logically why it would help my situation. Mainly because I didn't want to go through another round of installing different drivers, since finding one felt like playing roulette. However, this time, the Nvidia Studio Driver 596.32 has given me the best experience so far. It is fairly new, but this driver has been in testing longer than the Game Ready drivers, which are released much earlier in their testing cycle. I haven't considered Studio drivers because I am not a creator, I don't do 3D development, and I don't use the applications this branch supports, such as Adobe Premiere Pro, DaVinci Resolve 21, or Wondershare Filmora. But given my issues, I was wrong to ignore them. But I have tried around four different Game Ready drivers in the last year, using the DDU (Display Driver Uninstaller) method to remove my existing ones, and nothing has helped stabilize — it only made it worse. This time, I found a driver that I really liked. While my computer still doesn't feel perfect, it's much more stable than it was last week, so that's a plus. While I definitely agree that AI is not yet in a position to replace a professional, it has still done wonders in expanding my thinking and has been a great emotional support. I have spent months bouncing ideas off everyone I knew about what could be causing my computer issues, and at some point, people give up on helping me because they are at a loss for what's happening. It's been discouraging, but I can't blame the people around me when I'm struggling to diagnose exactly what's happening and have paid professionals to fix it. I also didn't necessarily want to nuke my $2,500 setup, since it was only slightly out of warranty, but I've been really tempted to. Though I still have more puzzle pieces to find, at least this time I feel much better about where I am now than a few months ago — my computer is actually in a usable state, and now it is more about removing the final hiccups. So if you're in a similar situation, I recommend working it out with the AI first, like I did, before spending a lot of money on an unnecessary fix or replacement.