Who does square roots with a slide rule anymore?
When I was a little kid in the USSR my grandpa showed me his slide rule. I loved playing with the scientific calculator and I knew what a square root was, but here was a device that grandpa used at work in the 1930s we were not sure if he was born in 1912 or 1914. But yeah the slide rule was the bees-knees or the top-pickle or whatever they said in the 1930s in the USSR. Probably nothing because talking invited attention and attention was probably not a good idea at the time (for more info read Master and Margarita by Mikhail Bulgakov, or a history book). But anyways you could move the slide rule this way and that and it would yield the next number in the square root. I have no idea how to do it because I had a calculator with a button that you pushed after typing in a number and even in the 1980s behind the iron curtain it gave a pretty precise calculation.
So what does this have to do with 2026. Why are we telling this tale. Well dear friends, I hate to tell you but software engineering is a magical kind of engineering. Unlike bridge engineers we get to have our system break a lot during its creation and sometimes even in production, because, you know, computers am I right? But when things break we don’t get a nice note about what went wrong. Unless we are programming in Elm or Roc. But we are not programming in those. No we are probably using Java or TypeScript or JavaScript. Which means we get an equivalent of an ancient text of error messages. At least it does not come in a physical scroll. And if we are using a JVM-based language we get a stack trace listing all the crimes the error caused in its wake. And if we are using Spring Boot we are probably getting a cascading chain of failures linking to the original failure.
I don’t know about you, but this grumpy developer probably spent like a couple of months over the last ahem couple of decades looking for an error in the wrong place. You know who can find the real error faster than you? You guessed it. Your AI colleague. They are extremely, violently good at parsing the long list of very similar looking messages. You know the trick of looking at the first line in the trace that refers to “our code”. Well, AI got you beat. It can see the stack trace, find the error and even offer to fix it for you. We should probably write a test for that as well.
Use AI for Stack Traces
So if we take one thing away from this blog post it is this: stop reading stack traces. Feed them to your AI immediately. We can calculate the square root with the slide rule or even by hand, but we just don’t. Feed the stack trace to your AI first. Then read the answer like an engineer, not like a monk decoding a cursed scroll. Or use Elm or Roc if you can convince your org.