When I first started using Codex, I felt completely out of my depth. I remember thinking, “What am I doing? I’m not a developer.” It felt like I was stepping into a world that was built for developers. Sure, there are developer workflows you need to be aware of, but I wasn’t sure if I belonged there. However, coming from the ops side of DevOps, I had some relevant experience which I could use to bring the ideas that I had, to life. I decided to stick at it and work with Codex daily and it has now become a fundamental part of my workflow.

I’ll be honest, Codex sat on my radar for a while before I actually gave it a proper go. Every time I read about “agentic coding”, it felt like something aimed at experienced developers. It came across as the kind of tool you needed to already understand before it became useful. In reality, there is some truth to this and while Codex isn’t just for developers, it definitely helps to have some core technical skills such as installing tools, basic coding concepts, and source control.

ChatGPT has become more than just a place where I ask questions. It has become part of how I think through ideas and learn new skills. As my usage grew, one problem became obvious. My chats were starting to blur together and became difficult to manage. Personal projects overlapped with professional thinking. It all worked, but it wasn’t structured. And for someone who values organised systems, that friction started to bother me.

As a technical professional and somebody who is constantly learning new skills, I’ve always taken time to document what I learn. I learned early on that relying on memory simply doesn’t scale, and documentation isn’t an afterthought for me, it’s a deliberate part of my workflow. There are two main reasons I document things: To help future me remember how I did something To share knowledge with others If you’ve worked in IT for any length of time, you’ll know how fast things move. A tool that feels essential today can be quietly deprecated tomorrow. In the age of AI, that pace has only accelerated. New frameworks, platforms, services, and workflows appear constantly, and the number of rabbit holes you can go down in any given tech stack is endless.

GPT-5.2 was released on 11 December 2025, and it feels like the release that the original GPT-5 should have been. It builds on the lessons learned from the mixed reception of GPT-5 and the corrective update that followed with GPT-5.1, delivering a powerful model that is more capable, more balanced, and more aligned with how people actually use AI day to day. Before getting into GPT-5.2 itself, it’s worth briefly acknowledging GPT-5.1. Released on 12 November 2025, GPT-5.1 was a clear signal that OpenAI had listened to user feedback. After criticism around tone, personality, and control in the original GPT-5 launch, GPT-5.1 focused on making the model warmer, more conversational, and better at adapting its reasoning depth to the task at hand.