AI Rekindled my Joy of Coding
When I first started writing code, I had a sense of joy in the fact I was learning the arcane incantations that brought modern day magic to life. Over time that feeling faded as I faced the realities of building code at scale. Bogged down by so many opinions of the “right” way to do things, instead of the joy of creation I now felt the dread of knowing I hadn’t tested the code well enough, or that dependencies weren’t injected, or that I had 327 npm deprecation warnings.
While I, like everyone in tech, got really excited about AI with the release of ChatGPT - so much so for me I made a language learning app for myself - there was a gap between the potential and the application. Github Copilot was interesting, but mostly only useful for parsing errors.
Recently, the use of agentic coding tools have rekindled that joy for me. I feel like I have a new type of magic, the ability to quickly write code that not only brings ideas to life, but does so in a rigorously correct way. While many decry the code created by AI as “slop,” I think that’s a matter of not using the tools correctly. I’m lucky enough to have had to make and review many, many product and engineering specifications, turn those into architectures and specs, and implement.
That same pattern yields incredible results with AI. Taking Cursor as a specific example, being able to create combinations of .cursorrules and notebook files to define what the project is and how the architecture should work: implement specs and then build against those specs has yielded functional software at 10x the speed it would have taken me to implement it before.
The website this is written on is brand new. I knew I wanted a new site, so I told cursor to start a new hugo project. It didn’t just write code files, it knew to execute CLI commands to scaffold it, so that the generated files were deterministically correct. I told cursor the changes I wanted to make to the layouts, and it worked in a matter of seconds. I, from past experience, knew I wanted it deployed to AWS via an S3 bucket with Cloudfront, and it spun up the terraform code to do just that.
These are all things I could have done before, maybe in 4-8 hours of remembering the AWS nuances, or working through golang templating syntax. But now, I can do it in a matter of minutes.
Right now, this sea change of productivity is what I want to focus on. It’s scary, and will affect jobs heavily, but imagining a doctor being able to focus more on patients and less on paperwork is exciting. Seeing what dangers a contract might have, or what an insurance policy covers, empowers everyone.
Like all new technologies, we need to be careful, but for now AI brings me joy.