Software engineers are having an existential crisis right now. For decades our value has been in knowing how to code, and now AI is allowing anyone to code by simply prompting an agent like Claude, Amp, or Cursor.
This thing we spent years learning to master is now available to anyone by typing a prompt. Does that mean we’ve lost our value?
The first time I used a really good LLM model to accomplish a complex task and watched it write code as good or better than me at 10x the speed, I thought the answer might be yes. Why does anyone need me around if the LLM can do it faster?
Then a couple things happened to change my mind.
First, in my personal projects I reached a ceiling where the AI stopped working so well. Building a simple web app or a prototype is easy, but as soon as you have something more complex it becomes much more difficult to make progress.
The AI starts making more mistakes, fixing one thing breaks another, and the agent gets into loops where no progress is made. It needs technical guidance and review to be successful. On top of that, at a certain point the real value is taste, creativity, differentiation, things that no AI prompt can help you with.
Then a second thing happened.
I use LLM assisted coding every day, after over a decade of writing code I now write almost no code myself. The crazy thing is, I am still busy, really busy. It’s not like I am sitting around twiddling my thumbs because I’m not writing code, I’m able to do things I had been putting off, or explore new ideas.
I built a product prototype in a few days that was not even on the roadmap while simultaneously working on a new agentic loop for a different product, having meetings, and reviewing PRs. I can add tests where I didn’t have time before, try out new ideas and maybe throw away the code. I can refactor and redesign and deploy new ideas so quickly it’s incredible. I can build things in days myself that used to take a team of people months.
I’ve always had to sacrifice things I wanted to do because I didn’t have time, but with the LLM writing the code, I now have time to do those things.
My value is no longer writing code, it’s at a higher level. It’s knowing what to build, what that should look like, how it fits in within our company, and what value it provides to customers. It’s having a great idea and guiding the LLM in the task execution.
The more I work with LLMs, the less scared I am for my job. Good ideas and technical expertise are more important than ever. AI is a force-multiplier, if you are a great engineer you will be 10X better with LLMs.
With AI, a company can now have 10 10X engineers, instead of 10 1X engineers. Most companies value growth over cost savings, so it makes more sense to keep all those engineers and create more code, as opposed to firing 9 of them to write the same amount of code.
Even if Jevons Paradox holds true, there will be job displacement for those who don’t adapt to this new paradigm. It will be more important than ever for engineers to lean into product + engineering, which I believe is the new full stack.
Read the full post here: https://scottbolinger.substack.com/p/dear-software-engineer-you-still
