I don't actually like coding. When I chose a major, a huge part of it was the appeal of the salaries in software engineering. There was a little optimism that I would enjoy it in the beginning. There was something exciting about learning how computers worked, the math that goes along with it. By the time I hit my third semester, one summer internship already done, I just wanted that six figure salary.
I became disillusioned with it earlier than most of my peers I would say. For them, there was a glory in "building" and an excitement to "solving problems". Admirable, but not for me. They really are doing great things out there, spurring the economy, boosting the average American's 401k portfolio.
It probably has to do with something about how I grew up, but work was always a means to an end. A software engineering job is one of the nicest jobs in history, good pay, using your brain, working with a team, entire companies built around your skillset.
Being good at coding was a skillset that helped land a job, but I think it was the rest of the skills that I had outside of coding that helped me in my career. I became an engineering towards the end of the summer and beginning of fall 2024. For various reasons, I was still coding quite a bit, if not more than prior to being an engineering manager. But recently, as more folks have been joining my team, I am taking my hands of the keyboard more and more and focusing on the rest of the job.
Luckily, LLM tools got really good this year. It was about February when I started diving into Cursor more heavily, and then sometime not long after Claude Code. People debate the value of these tools, are they increasing productivity, are they more hassle than they are worth, etc. I actually don't care about any of that. When measuring the productivity of these tools, no one is measuring the "unpleasantness" of coding and what AI tools do for this.
Unpleasantness is not always valued in measurements around labor and producitivity. We are concerned with return on investment, increase in GDP, time saved, etc. Unpleasantness can be measured in two categories: a job being physically demanding or repetitive and boring. Coding has nothing to do with the former, and people might think that it is has nothing to do with the latter.
There is a repetitiveness to many problems of software engineering. Most software engineering problems are "solved". We already would read documents, stackoverflow, or other examples in our internal codebases. Netflix, Google, and Meta, to name a few, have blog posts over a decade old that you can still read and apply to your own problems when you face that same situation.
I am not like the superstar coders out there, and most software engineers are not. I know the superstar coders. I went to school with them. They truly enjoyed coding and building things, and much of them make 2 or 3 times what I do now and have net worths that are already in the 7 figures if not higher for some. Those superstar coders will progress rapidly through their careers, they will hop jobs every few years, go to graduate school, and always be pushing the boundaries on solving problems in industry or academia.
AI tools free us up from much of the repetitive and boring work of coding. You can describe that you want to implement some pattern but in this language, or ask it to write tests for your existing code, ask it to rip out 15 API endpoints for you. Yes, there were code gen tools and other more deterministic ways to do this, but they are not as good as the AI tools as of 2025.
More and more people who are superstar coders are finding AI tools to largely solve most of these repetitive and boring tasks, the actual process of 'coding it up'. There are people who I know have high quality standards of coding and will honestly say they are not writing any of the code themselves anymore. I know they are not shipping slop.
Whether or not there bubble will burst soon, and it will, whatever the space of coding looks like a year, five years, or ten years from now, the unpleasantness of coding might be entirely gone. What will remain is the problem of working with others on a team, delivering projects on time, choosing the right architecture, and creating new ways of solving problems.