After using some cursor.ai

beginning

As the final vacation started, I felt some pressure to restart my developer’s life. From high school, creating something with computer was my hobby. I always worked on some project just for fun. Therefore, my major became programming related.

Joining a Korean army after first year of University was really natural decision. As a Korean citizen, serving for Korean Army was mandatory, and unavoidable. Everyone says it is better to finish the military service as younger age, and I was asking myself what should be my specialty. I did like programming and computer since elementary school, but I haven’t thought about detailed specialty. Not because I don’t want to do anything, because I liked everything. Also, after Chat-GPT came out, the world took too much changes. That is the background of early military service, and I have no regret about it.

More than a year passed, my final vacation has started. Final vacation means I got enough vacation to use every day until the discharge date. Which means I got no more military life any more. While the final vacation, I stayed home and enjoyed life of civilian. I came back to my original life and programming was part of it.

At the beginning, I didn’t know where to start even I prepared a lot of plans and ideas about projects during the vacation. The AI went so big, and role of programmer has flipped. I couldn’t even choose my IDE. But I have enough time for trying various methods, so I decided to do what I’m good at. Learning by doing

Using AI tools

The first method I tried was my old bread and butter, VS code. I had a student account for GitHub copilot, so VS code + copilot chat was my first method. Its auto complete was pretty okay, but I know that is not enough to follow the world. I heard too much reviews of Claude code, which is a cli version of Claude ai. I didn’t have Claude pro plan, so I used a alternative, which is Gemini CLI.

The first project after the military gap was personal webpage renewal, which seems really natural choice. It was totally interesting to run a project with command line LLM. I renewed my personal website with Astro, TypeScript, and Tailwind CSS. Why? Just because AI told me they are good for building static personal website. After finishing the project, I realized something. I know nothing about Astro. Gemini CLI did every Astro job for me, and what I know is how to edit /src/data/posts folder so I can create new blog post on my website.

This is the way I’m not familiar with. Using some tech stack from a project was the way to learn and get familiar with that technology. However, I didn’t thought this was the loss of opportunity to learn new technology. Technology will get old soon, and it is meaning less to learn every framework that is good for only few years. Instead, AI gave me some kind of freedom to focus on real problems, such as “How to plan a structure for a website?”, “How to brand myself?”, and “How to focus on the value of the project?”.

Those question was also important before AI, but technology took those effort like taxes. With AI, we encountered duty-free area of career, so we don’t need to spend our effort on something that will vanish someday.

Blind point here is that it is still possible to run a project without such deep consideration. If I just say “do this” or “fix this” to ai, I will earn nothing from the project. And if you don’t build yourself, congrats, you become a replaceable.

Using cursor

Programming with cursor was even more interesting. Unlike running only one agent, (that’s how I used Gemini CLI), Cursor enabled me to use multiple agent at once. The first project with cursor was creating a working webapp with REST API and simple database. Therefore, I used agent called ‘project manager’, ‘frontend director’, ‘backend director’, and ‘QA helper’. I set the clear rule between the interface to prevent conflict. The Project went well, and I could finish the first MVP(Minimal Viable Product) in a single day. It had so much to fix because I haven’t explained in max details, but I can see what it can do with clear and complete Instruction.

I realized that I’m not a programmer anymore. While the project, I didn’t do a single programming. all I typed on the codebase was just a API key. All I did was explaining about project to agents, learn how to try the completed product. I was more like PM or QA. I must use more to fully understand what is my role and goal, but I already got shocked.

So What?

There are two perspectives still fighting inside my head. One is acknowledging my new role and new world. Unless something bad happens, I will be able to use AI agents while coding. Which means I don’t need to think about actual programming anymore and only focus on learning how to exploit those agents. Another perspective is believing I’m still a software engineer, and knowing the technology is essential part of solving problem. Even you have a slave that does everything for you, you still need to know the technical process.

Whichever side is right or wrong, all I can do is just trying and learning.