我使用 AI 的經驗

When AI became more than a search tool

Hi! I wanted to share how AI has changed the way I code, come up with ideas, and learn.

When ChatGPT first came out, I thought it would be like a search bar or Google Translate: a handy tool for finding information faster, getting ideas, and having a little conversation using a improved chatbot... But I was wrong.

As AI tools kept improving, I started using them to learn new things, understand what concepts meant, and figure out how to fix problems. AI began to feel like a real teacher. It opened my eyes to how much I could learn and how many things I could do faster.

Bringing AI into my code

Then I tried using AI directly in my code. Using only AI Chat for guidance was great but I still had to copy and paste errors back and forth and spend time figuring out what to change.

When I tried GitHub Copilot, I was surprised by how quickly I could get a working MVP. I could describe what I wanted, allow the changes, and watch the project take shape.

At first, that was impressive. Later, I started getting bored with parts of the process: clicking “Allow” again and again, and explaining how I wanted the AI to code for each project.

Moving to the terminal

Then I learned about coding tools that run in the terminal, like OpenCode. The workflow felt familiar, but I could work from the terminal instead of switching to another window.

I also found the opencode --auto command, which made the process feel faster and reduced the number of times I had to approve actions. The agents also helped with some of the initial back-and-forth about what needed to be done.

What I think now

These tools have changed how I approach coding. I don’t want to give up learning how the technology works, but I do think AI can help me get from an idea to a working MVP sooner.

I can see how the pieces fit together, learn from the result, and then try building or configuring things manually in a future project.

Of course, speed isn’t everything. AI may keep working toward a goal without always choosing the best or safest approach. Developers still need to understand how the technology fits together, review the code, and think about security. AI can help us build, but we’re still responsible for understanding what we’re building.

That’s all for now! I’m planning to experiment more with AI and share what I learn. 😄