It's that day again when that one task needs to be done. That task you can finish regularly in a short amount of time. That task where you ask yourself every time:
Should I really automate this now?
💡TL;DR
In this article, I’ll show you a tool that lets you calculate when automation becomes worth the time investment. I also talk about how I implemented it and what I learned in the process.
The Motivation
As developers, we probably tend to automate activities very quickly. And why not? After all, we learned to make the PC work for us. But we've probably also learned that automation can sometimes take longer than hoped. That's why I dare to say that almost everyone has faced the decision: Is it really worth automating this?
Many years ago, I saw exactly this problem in an XKCD comic and thought it was great:
Today, we are going one step further. We will translate this comic into a website. Because never again should this question simply remain unanswered.
The Implementation
Randall Munroe already broke it down very well in his comic: Cumulative duration of repetitive tasks vs. one-time automation duration projected over 5 years. Now all we need is the calculator that takes our inputs and totals the values.
The finished calculator looks like this:
The source code is available on GitHub, and the page can be accessed standalone here.
In addition to the core feature, the following functions were included:
- Maintenance duration per year
- Visualization via a ChartJS
- Inputs are transmitted as GET parameters so you can share the result as a URL
Lessons learned
I already programmed a similar tool in 2017. Even back then, the comic motivated me to do a small side project. Back then, it was an exercise to try out AngularJS.
Now it is 2026, and I wonder if AngularJS even exists anymore? Regardless, nowadays we let code be written for us, so I used this project again to learn something new: how would Google's Gemini implement it?
So I cobbled together a prompt and just let the servers work. For the first attempt, I used Gemini's Canvas. About 80% of the application came directly from the AI that way. Everything was in a single HTML file and wasn't really easy to edit further. So I set up a local project and opened it with VS Code. There, the Copilot Agent mode was immediately pushed onto me. So I continued working with that.
➕ Positive Experiences
- Google Gemini
- great basic structure for a simple website
- knows Tailwind and CSS much better than I do
- great integration of ChartJS
- great implementation of localization via data-attribute fields
- Copilot Agent
- Extension of the GET parameter function went without problems
- Adding a reset function was no hurdle
➖ Negative Experiences
- Google Gemini
- Gemini Canvas only works in one file (As of 2026-01)
- Tends to make comments instead of assigning new functions with clear names
- I moved the Style and JavaScript to separate files manually instead of letting the AI do it 🙊
