Grief and the Nonprofessional Programmer: when AI does the coding, what’s left?
MISRYOUM Tech News — There’s a kind of silence that shows up after a win. You prompt an AI coding assistant, it produces the working thing you wanted, and for a minute you’re just pleased. Then, later—maybe after the laptop fan fades or the meeting is over—you realize you never actually internalized what happened.
The essay at the center of this debate, “Grief and the AI Split,” is written by someone who doesn’t quite pretend to be a professional developer. They say they analyze spreadsheets with Python, sometimes “hack something together,” and when it’s for fun they gravitate to prime numbers and numerical analysis. The familiar frustration is there too: getting stuck in pandas, hunting documentation, tutorials, and—yes—several incorrect Stack Overflow answers. But the deeper tension starts when AI enters the picture and makes the hard part easier.
The author describes experimenting with AI-generated animations of algorithms, initially after reading about AI-made sorting animations. The spark is simple: if an AI can animate algorithms, maybe they can build an interactive one too. So they ask Claude Code to generate a web animation for the Fourier series, and later, using Algorithms in a Nutshell as a kind of menu, they push it toward less common animations—ending with Dijkstra’s algorithm for shortest paths.
Here’s where the story gets interesting and a little messy, in a human way. The AI has trouble with a few algorithms, but the problems are handled once they ask it to generate a plan first, then follow up with a second prompt asking it to implement the plan. That two-step workflow—plan, then build—makes the end result work. And it’s genuinely fun. The author leans into the thrill of controlling machines, and mentions that they could even spin up multiple independent runs in parallel without fancy orchestration. No Gas Town, apparently—just straightforward parallelism and prompting while life keeps happening around it.
But then the grief hits. They say they don’t really understand Dijkstra’s algorithm anymore—not in the way you can confidently rewrite it from scratch or explain the “why” behind each step. They know what it does and have a vague mental model of how it works, sure. Yet using an animation as a replacement for reading and thinking through the algorithm feels like cheating their future self. It’s not that the computer failed. It’s that the author bypassed the slow, stubborn process of understanding well enough to write the code.
They connect this to Fourier transforms too. They’ve already vibe coded an application that recorded audio from their laptop’s microphone, ran a discrete Fourier transform, and displayed the result. The detail is oddly vivid: they move the laptop over to the piano, start the program, play a C, and see the fundamentals and harmonics. But even with that moment—almost satisfying in a tactile way—they realize they wouldn’t “need” to write the code again. So if they learn Rust, would they implement Fourier transforms from scratch, or ask Claude and inspect the output? They already knew the theory, but the sense is that an era ended, just slowly enough to notice later.
Underneath it all is a broader worry about how AI changes learning. In the past, the shift from assembly to higher-level languages still required you to know what you were doing; you couldn’t learn signal processing purely by writing assembly. Early languages like Fortran, Lisp, Algol, even BASIC supported gradual understanding. The author argues this is the real source of grief: understanding how things work gets traded away when “getting stuff done” becomes the default mode, especially when the screen is blank and your brain just wants motion.
So what does this mean for creativity? They float the idea that delegating understanding of old problems to AI might delay or hollow out human creativity in computing. Maybe creativity shifts up the abstraction stack—toward specifications, problem framing, and writing the detailed instructions. But they don’t fully believe that solves the grief for the programmer who loves coding, or loves the understanding coding brings, even if they don’t love the coding itself. And the ending lands a bit open-ended—like they’re still checking whether the feeling is real, or just the cost of moving faster than their own curiosity.
Microsoft patches Copilot Studio prompt injection anyway data leaked