Being a software engineer, I will admit I was a little slow to really start and embrace AI, but lately I have been trying to use it more and more in my day to day work. I have been primarily using the Opus 5.5 model as my daily driver and I must admit I am pretty impressed with the code it actually generates.

In my typical fashion, when I try to learn something new, I usually create a blog engine (or some other well known application) and so as "proof is in the pudding", that is exactly what I did. You can find all the code that was used for creating this blog here. I will admit I did use a template of my own to initially kick off this project, but after that it was 100% generated by AI. As I learn more, I will try to keep this like a running dialog/journal of my learning experiences. Right now I can clearly say I still have a lot to learn about how to use this "new technology".

My current tooling

  • I primarily have used Claude Code for my work. I am currently running on a paid Anthropic plan (right now using the Max x5 plan) and while I have kind of gotten close to my weekly limits, I have still not hit them.
  • I was using the cli for the longest time but then had co-workers mention how nice the Claude Code App is, so I have been trying to transition to that
  • I have only been using Opus 5.5 on "High" thinking and haven't felt like I need to change that setting for the moment.

My current thought process when using AI

  • I start off with a general prompt that has a lot of what I want in my application, but then also end things off with "do research and find items that I have missed" and I also tell it to "write a detailed design document" that I can review and verify.
  • Based on that technical design document, I work with AI to walk through any open questions it had when it created the document.
  • After all opened questions have been answered, I prompt AI to create a detailed task document that can be used for tracking progress during implementation.

There is still a lot I need to learn, and currently I will probably continue to focus on the claude agent style of doing things since I have figured out already that AI is moving so quickly, it is tough to keep up with the models, much less all the tooling around the models. Trying to then learn the way multiple agents operate is a little too much for me at this moment. Maybe I'm making a mistake, but those are my feelings at the moment.