I’ve been thinking about how I feel about AI tools recently, engineering tools like Claude Code, ChatGPT health, or automated investing through MCPs.
I wanted to dive in and talk about how I feel about AI doing more of the work while making my individual productivity much higher.
I’ve been an avid Claude Code user, like most of you, for about a year now. I’ve done things in research and engineering that would’ve taken me so much longer if done by hand.
An example is that I recently reverse engineered all of Old School Runescape, then made a private server, designed the scripts and persistent states for about 2,000 bots, and now I watch an automated economy form with me being the only human player. Without AI that would’ve been an insane thing to do, even using pre-existing tools and just putting things together would’ve been a big headache. With AI though I was able to do it in a day and now it’s something I can enjoy.
On the flip side, when a project comes together that fast, it’s easy to end up with a system you didn’t build line by line. You know the architecture, but not every specific.
Which is why I think my stance has changed a bit recently with AI use.
I’ve come to the conclusion that when you make something it’s important to be able to maintain it yourself. I think speeding things up with AI is great, however you should still be able to say what went wrong and what needs to change to fix it.
The reason for me at least isn’t some moral stance, it’s because I want to be a contributor when I’m making something with AI. I should be able to look at the codebase as Claude is making changes in it and be correcting things in real time.
Learning is hard. Understanding is hard. But I think it’s worthwhile even if it’s tiring. Humans should have to struggle when mastering something, to master something struggling seems to be a prerequisite.