Embracing AI
Starting my AI journey and embracing it.
AI is everywhere, and, like most developers, there was an initial fear and scepticism. AI came quickly after the crypto bubble, so naturally, I thought it was a similar fad. However, it looks like AI is here to stay, and it is where the industry is heading.
I have started to fully embrace AI in my daily workflows and have started to use it to improve them. I started using chats as a better Google when I ran into coding issues or knowledge gaps, still writing 90% of the code myself. I then found myself relying on it for writing test cases and doing tedious tasks like reformating csv or research tasks or light web scraping. These all proved useful but were very manual: copy from 1 tool to another and back again, tell it it’s wrong, try again, etc.
Claude Code has completely changed how I work, and I am now fully locked into an AI workflow. I find myself wanting to change the fundamentals now. I use VSCode as my primary code editor, and I am now in the terminal more than ever. I am experimenting with many different approaches, tmux for example, and I have been getting Claude to write small CLIs that work the way I do, things like managing Jira sprints, checking out git worktrees based on GitLab MRs. I am still finding my feet with all this, and perhaps I am a little late to the game. But there is so much rich content out there, and I am finding it really fun learning a new skill. I was finding that I had started to plateau, my knowledge of the front-end had hit a limit, and I was doing more and more back-end. But now I have a new angle to approach front-end work, and there is lots to discover.
My current research is looking into custom Skills. I have built an /atlassian skill for interfacing with the acli for reading and writing Jira tickets and a /refine skill that builds on top of that and asks me a bunch of questions about a given ticket, producing a high-level technical architecture and questions to ask with the team. Yes, skills like this exist in the wild, but I like to understand how things work, so building some simple skills like this from the ground up is deepening my knowledge. I don’t like black boxes, and AI is definitely prone to lots of them.
Next on the agenda is building a custom DESIGN.md file and looking at how that interacts with Claude and how I can leverage Figma, perhaps working with MCPs.