Safety first Link to heading
There was this promising stock, Monday.com, with positive metrics and well-known—at least to me. It had been on YouTube ads for what felt like millennia. I bought in on a dip, but unfortunately the dip persisted. Since I was just getting to know investing and was overconfident in my analytical skills, I had no automatic loss protection configured. Long story short, I lost about five years’ worth of liquor money in that week.
Who what and why Link to heading
After the initial shock, I looked around to find out what caused the crash even after an earlier price correction. It seems the two reasons below were the culprits.
- Investors fear that AI can write software faster and make pre-built software irrelevant.
- Search traffic is decreasing because of AI.
Since my money was on the table and there is an ever-growing need for code and script troubleshooting in system administration, I tried GitHub Copilot.
The fifty thousand packages Link to heading
I asked the AI to build a simple project-management application, and it gave me a complete list of requirements. As usual, it was full of nicely named components that required installing many packages with numerous dependencies. There may be advantages to that setup, but as a sysadmin who wants to build software for my company, I wanted a very minimal setup. So I asked for a PHP-based setup, which I’m very familiar with. The AI provided setup instructions and code, followed by steps for security hardening and other next steps.
I need a clone Link to heading
Because I wanted to create a kind of clone of Monday.com, I provided the AI with an image like the one in this post, and it produced the result in a few minutes. Although it was just a simple PHP page replicating the provided image, it was a fully functional web page. I had used Copilot for code writing before, but it often had issues like missing code segments and logic errors. This time, however, the output seemed more polished and the code more intuitive.
Integrate or isolate Link to heading
I installed Windows 2000 in my home lab back in the day so I could use Active Directory to create user profiles and passwords. This was suggested by the AI as the next step, and it reminded me of the plethora of complex compliance requirements a company might need. To keep things simple, I went through Monday.com’s feature list and provided it to the AI. The result felt like it came from my mom—very detailed about every aspect of the software-building process, but still requiring a lot of thinking and planning on your part.
Final headaches Link to heading
I love building things myself, but there is a limited amount of time you can allocate to certain projects. For that reason, my home lab is built with open-source software, and I use Google and Microsoft for the rest of my needs, like email and AI. I could certainly build a simple project- or work-management application myself, but I’m not sure it would be suitable for commercial use unless it’s heavily tested and used for a couple of years. Regarding support, owning and maintaining critical software for your company is usually a nightmare, so that’s a big no. As a final note: AI is good—really good. But in an era where technology is supposed to ease your burdens, being the person responsible for software that you and AI have built can become too much of a burden.