Hello there! 👋
AI has become an integral part of software development in the current time, with all these intelligent models and productive coding agents, it's important to keep a balance and not overuse these tools, or you have a different take on that?
Do you think we should use AI more or less in our day-to-day work, and when to limit their usage?
Share your thoughts in the comments section, and I'll go first!
Top comments (1)
I have personally learned how to code a year before ChatGPT was released, which happened in November 2022, definitely to me learning prior to these LLMs was totally different, now it's much easier to build and learn things much more quickly. However, now we've got all these agents, and agent-packed coding editors like Cursor which I personally like, it's unarguably smoother to get something rolled out, maybe a POC, even a full blown project with these AI tools that we have today.
I used to always think like I need to limit my use of these tools cause that would make me less smart, and that I wouldn't actually think through code and solving problems, but recently I have shifted my mind on that. Many may not agree with me on this, but I feel like these tools should become a skill to master and use properly in every-day tasks. I have had this thought always that if I don't write everything on my own, I wouldn't learn it or understand it, which did reduce my productivity to actually accomplish my work or something I'm working on. When I started actually thinking through solutions in a high-level way, and focus on solving the problems on paper, or architecting some designs on paper, and then have these agents help me out in the implementation, I did realize that I don't necessarily have to do everything myself.
These tools are here to stay, and make us way more productive, I even got a subscription for all major models, GPT, Gemini, and Claude (My favorite for coding), and I do use them in every task, now I'm way more productive, I can deliver more, and even learn better by leveraging these tools correctly, which leads me to the usage limit.
Don't ever use an agent and go like, hey, build this, and spam it with a feature requirement. Don't get to the code of it until you have understood what's needed, what's the problem you're solving, and at least devise a simple solution, and then coach the agent or model to achieve it in an efficient way based on the requirements.