---

title: Engineering with AI created_at: 2026-05-23T14:52:16.823Z updated_at: 2026-05-23T14:52:16.823Z tldr: It is still complex is_published: 1 star: true category: Technical share_type: add

When I started working on software projects, I wrote code by hand, letter by letter. To be honest, back then I could only type at the pace of 10 WPM. In that process, I really thought about the system as I typed through; in a way, I was forced to think about the boundaries of that piece of execution. Also, we had to write tests by hand - sometimes I had to go back and modify the code to add the boundaries I hadn’t considered before.

Now I can type at 60 WPM. I still have to loop through the bounds of the project as I type. Honestly, if I could have 4 hours of focus time, I could spin out around 1000 lines of code, with the same process. Writing code was never fun for me, but I had to.

With AI, this process has thick boundaries. AI is writing most of the code, but I cannot inspect the bounds as it writes. I have to read through all the modifications, validate, reject and add bounds to my initial prompt. If I don't like the way it codes, I have to set the context on how to write code or define the initial models the way I want. For long-running projects, like migrations from one to another, this loop becomes even more tedious. That is the same as before, the pain points have shifted before AI writes - Plan and after AI writes - Review. The only thing that has become faster is the pace at which AI can type. Like the Kid who would take 1 hour to finish colouring 1 page of a colouring book can do that in 1 second. If I don't like the way the colour was filled, I have to ask the kid to do it all over again!.

All the complexities I had when I first started still exist today.