It took me quite a while to solidify my take on AI tools for coding, and it's been quite a journey. From initial excitement to deep scepticism and back again, I hear many developers describing a similar emotional arc with AI tools.

I first tried AI in early 2023 when IT was mesmerised by ChatGPT and with AGI supposedly being just around the corner. I was learning Unity and working on a game prototype, so I needed to pick up C# quickly. Someone suggested using AI to answer my coding questions and I started exploring it. It was great, but definitely not able to build complex systems on its own. When I tried Copilot for a JavaScript project, I watched it generate boilerplate code in seconds. I was impressed because even doing that was cutting my time on repetitive tasks and helped so much with navigating documentation.

Soon after, I hit its limitations and went back to thinking it was all overhyped. This was around the time when every CEO was "replacing" entire development teams with AI.

I experimented more and explored Cursor when it launched. "Tab, tab, tab" magic was great but I wasn't as impressed as most people. Soon after that, Copilot became free and the conversations were dominated by AI agents.

I found myself in an up and down cycle. Each new tool or model promised a breakthrough, but failed at more complex workflows. Despite hitting limitations, I started gaining more experience with each project.

Claude Code (and CLI agents) changed everything mid-2025 for me. The experience I'd gained with other tools allowed me to use it more effectively, and I completely rewrote one of my personal projects in two weeks. AI won't build a complex application in one prompt, but it excels at specific, well-defined tasks.

After two years of experimenting with these tools, I've found my place somewhere in the middle rather than at any extreme. I learned that working with AI tools is like any other skill (but accelerated several times) - it takes time, gets easier with practice, and you gradually learn to match the right tool to the right scenario. You can't wait for perfect tools; you have to work with what's available and build your expertise along the way.