It's crazy. I starting learning to code last august. I flipped flopped from web to software development. I finally setting on software and apps. I'm in the midst of trying to build a tech start up using nothing but AI. I picked an existing space with the idea of making it better, be it esthetics and/or performance. I've learned how to do product research, map out features, put together an overview, create a file structure and create mockups using excel, all before any code is written.
And from my experience, I will say this about AI. It's cool if you're creating basic web pages or basic software tools. But if you're building complete builds, it's going to make the same mistakes humans make. I've had to make corrections or call out the AI for giving me the wrong shyt multiple times. In the beginning, I would just ask it to write a whole file. When it didn't work, I would show it the errors and tell it to write it again. Then I started doing my own investigations and asking the AI questions. Basically, you have to use it as a collaboration tool and not something you can expect to produce everything. I initially went from using Claude, ChatGTP, Cursor, and Bolt, to just ChatGTP and Claude. And the only reason why I use 2 is because 1. chat limitations, and 2. They both give different perspectives.
