I Needed an Internet Speed Monitor, So I Built It Using AI Coding.
I was tired of guessing whether my internet was the problem — or if it was just Zoom being Zoom. When my connection lagged, my ISP always said everything was fine - I needed proof. Hard data. Real numbers over time. So I built Internet Speed Monitor - a desktop app that quietly tests, logs, and charts your connection speed.
But here’s the twist: I didn’t really code it myself. AI did.
Why I Needed It
I needed evidence - not a one-time speed test.
Sites like Fast.com or Speedtest.net are great for a single check, but they don’t show patterns or history.
I wanted something that would:
- Run tests automatically in steady intervals.
- Save results locally (no cloud, no tracking).
- Visualize trends so I could show when my connection dropped.
- Export data to be able to analyse further and share it.
In short: I needed a personal internet telemetry dashboard.
AI is My Development Partner
I could’ve coded it from scratch, but I didn’t want to spend days fighting Electron boilerplate or debugging cron jobs. So I let AI write it.
I used an AI coding assistant (Copilot) to build nearly every part of the project — structure, features, bug fixes, even documentation. Here’s how that unfolded.
1. Start from scratch: “Create an Electron app…”
I began by asking Copilot: “Write an Electron app that periodically runs a speed test and displays results in a chart.”
It generated a complete project from boilerplate; main.js for Electron’s main process, an HTML renderer, IPC messaging between them, speed-test scheduling modules, local files for data storage, a build script and packaging configurations.
Pretty soon, I had a runnable desktop app window — something that would’ve taken hours to scaffold manually. I barely touched the first version; I just let AI fill in the blanks and ran it.
2. Automate build and release
Next, I wanted an automated build and release proccess. I asked Copilot for Github workflows to detect when we (it, usually) changed the version in the commit, then it would create a new git tag, build for all platforms and attach binaries to the new Github release.
3. Test and Improve
When running the application I started testing it manually. Working with the application myself has quickly created new feature requests and bug reports, where Copilot and I would debug together; sometimes I’d suggest where to look, and sometimes it would generate more log output to inspect itself.
4. Document and Present
I asked Copilot to create a website for this application, which we publish with each version from within the repo using Github Pages. I only had to provide the screenshot, and specify the areas I wanted it to focus on.
What I Learned
AI coding is real, now.
I built a fully functional cross-platform Electron app almost entirely through prompts.
It’s like pair-programming on steroids.
I didn’t stop thinking. I did stop Googling.
It’s not perfect, but fast iteration beats perfection.
When AI got something wrong, I just re-prompted until it worked.
The role of the developer changes.
I was no longer writing code; I was orchestrating code.
The Result
You can check it out here: 👉 Internet Speed Monitor
and its open-source code: 👉 github.com/omrilotan/internet-speed-monitor
It runs quietly in the background, logs your speeds, visualises trends, and gives you evidence when your ISP insists “it’s fine.” All data stays local. No servers. No tracking.
But to me, the bigger takeaway is this:
I didn’t build a speed monitor.
I built a workflow where AI builds software.
What’s Next
Of course, I plan on continuing to maintain and improve “Internet Speed Monitor”. But more than that, I can’t wait for my next project. I’m excited to try and build an AI-driven app using test-driven development, to see if it can improve on new releases process, trigger development via user issues, and explore even more automation possibilities.
