Crazy Time To Be A Software Developer

It's crazy to me how we as developers are using AI today. Two years ago, using AI was opening GitHub CoPilot in Visual Studio, or maybe using ChatGPT for answers to common coding questions. Four years ago, AI was a mystery to me. Recently, I've moved on to the next level of using AI: agentic coding. I use a terminal app called Claude Code which does the coding, troubleshooting, and development for me. I simply prompt it to fix things, update styles, look into errors, and so on. This is a huge change to how I work as a developer, and while making the switch to agentic coding wasn't my decision to make, I actually don't hate it.

Don't get me wrong - agentic coding takes the fun out of coding, period. It atrophies your critical thinking skills as well because you're not the one solving problems anymore. However, this is where the tech industry is going, and shortly after stepping into my current role as a Full-Stack Web Developer, I was told to use Claude Code to do all my work. Why? Claude Code can do it much faster than I can, the project I was working on was moving too slow for senior leadership's taste. Thus, a process change was made to our development team: use Claude Code to write everything, don't write code manually. That new process started 2 months ago. In this blog post, I will share my opinions about this change to my development style.

I don't hate agentic coding

With the requirement of using Claude Code in the office came the end of writing practically all code at work. Yet, I still feel a sense of achievement when I make a fix or implement a feature. I'm completing tickets way faster than I was before, and I feel accomplished every time I submit a PR. While I'm not the one doing the actual coding, I am the one that pushes it, and I'm responsible for the end result.

I've become a bit of a QA/QC professional now, seeing how a good chunk of my job is testing the code Claude writes. I don't hate that either. Working with Claude Code has felt like working with a junior developer, and there's definitely some comradery at times, in addition to the AI being an excellent sounding board for ideas, design directions and UI/UX recommendations.

I do miss coding - thankfully, I still get to do it on personal projects - but overall, I haven't hated this change. I don't love it either, obviously, but it does feel like those that don't embrace this change may be left behind. I'm simply not willing to take that risk, with the job market where it is today.

AI hasn't changed how I do everything

Despite this massive shift to agentic coding, here are some things that have not changed how I do my work, and probably never will:

• I don't vibe code. I define vibe coding as prompting AI to implement features and fix issues without reviewing the code it pumps out. I don't think I'll ever agree with developers that do this. You are responsible for the code that AI is creating because you are the one submitting and merging it, so read what you're submitting. To me, vibe coding is as nonsensical as not reading your own PRs before submitting them, which I've seen way too often in my career.

• I test every change AI makes. When Claude fixes something, I reload the page and make sure it works. I make sure it looks good on mobile. I check for console errors. These are basic expectations for any full-stack developer, even if they use AI.

• I don't feel like I'm forgetting how to code, because I read everything Claude writes. How could I, say, forget how to create a custom React hook, or implement a .NET Core API endpoint? I see it in the PR diff constantly.

• I still review my colleagues' PRs. AI hasn't changed my passion for finding issues in people's pull requests, believe me! 😂 I read PRs carefully - the same way I've always done - so no, AI isn't dulling this skill either. I haven't yet worked on a project with AI code reviewers, but I'm open to experimenting with that when the time comes.

Conclusion

I truly never saw software development going down this path. AI is a useful tool as it makes it easier to solve problems, but as a result, it does make us worse problem solvers. It takes the fun out of coding, and we all know it's affecting the job market. Still, it allows us to complete our work much faster than ever before, and it still gives a sense of achievement when you finish something. I'm interested to see where we end up in 5 or 10 years, and what working with AI looks like then.