The la(te)st iteration of the todo app

(I tag everything #rambling, huh?)

I wrote about about my blog software iterations. That is not the only piece of software I iterated over quite often. The other one is your typical example app: a ToDo management software.

That one started around 2005 when I learned about C++. The only realy reason why I did a todo program was because I wanted to write some software and couldn’t really think of anything else I could use linked lists for :)… It got confusing quite fast, so I abandoned ship and learned Delphi.¹

There were a bunch of attempts over the years with different tech stacks. One of the wildest ones was with a microservices architecture. You can imagine how that one went.²

I got inspired to try it one more time for two reasons:

1. I needed some project to try out working with (coding) agents
2. What’s around, and what I can do on paper, don’t satisfy me

Yes, I know it’s a cliché. I’d really love to justify myself now, and write a few paragraphs about my unique approach to the problem almost every software developer at least once solved one way or another, but if I’m honest I didn’t even do a proper “market research”. I knew I needed a todo backlog, and some way to plan the next week in a very lose kinda way. I also knew how to do it (as much automated via agents as possible), and had a good idea of the architecture needed.

Setting up opencode was very confusing (do I really need different agents? Why not tell one but more detailed? What the fuck is a subagent? How can I make the agent not ask for permission to commit? […]). I gave it quite a detailed description of how the app should work, what tech stack to use, how to structure it, etc. Been a while since I wrote that much text without getting stuck or sidetracked. I wanted to see how fast I was.³

It was around 16:30 when I fired up the engine. It’s definitely scary to give up control over the code. Sure, I checked how it did things, but I was mostly interested in seeing how little I’d had to interact with the code. Let’s not get sidetracked into my experience with that (maybe another time, when I experienced more), and my thoughts about AI (again: maybe once I know a bit more about what I’m talking about).

The software was up and running in production at around 21:30. Funny enough: I’m quite experienced with DevOps, and I’ve built a lot of CI/CD Pipelines over the years. I still suck at getting that right fast, because there’s always something. That took me like an hour. Another half an hour was lost to traefik not seeing the new container or reading the labels or something like that - yes, restarting traefik solved it. I also got sidetracked by a colleague, which took like 15 minutes, I did that during agent work so it doesn’t really count. Around 45 minutes were spent on talking to a friend about this, which I count as working on it.

So, overall it took around 5h, in which I got the latest iteration of the todo app in the latest iteration of how software can be written. I still did a fair bit of manual labor, but that was mostly getting the deployment job right. I also won a round of spider solitaire during “thinking”. Does the software work?

Lol fuck do I know.
I mean I still don’t really know what I want. But I have a nice starting point I got to extremely fast. And I did enjoy the process.