Disclaimer: In my day job, I work for big tech and have been writing code professionally for more than 10 years. The monolithic code bases I toil in are different than what I’m discussing here.
We’ve been working on an algorithmic trading platform over the past 8 months. I’d estimate that LLMs have written virtually all the code (vibe coding ftw), with the remaining bits being copied from previous projects. I can’t think of any meaningful feature I’ve built purely on my own. Here’s a brief architecture diagram to give a sense of complexity:
In the beginning, I spent more time fighting the LLM, than the LLM actually being helpful. I’d reject multiple suggestions in a row, reiterate myself often (while occasionally shouting at my laptop), and very carefully try to prevent the LLMs from going down a rabbit hole I didn’t want to go down. This was across a wide variety of tools.
Claude
ChatGPT (web)
Cursor
Exponent
And then, over time, things changed. Models got more performant and I learned my own tricks for guiding LLMs towards the intended destination. As a result, I drastically changed the way I think about my code. I used to care deeply about code quality, readability, and maintainability. Now I find myself doing a cursory scan of the diff/changes and going to the next prompt. I’ll paste a full stack trace into Cursor go brew another cup of coffee, and move on. My codebase started morphing into someone else’s code base. While I still understand the high level architecture and the various interactions between microservices, I find myself ambivalent to how it’s actually implemented.
Does it function? ✅
Are there tests? ✅
If there’s a problem, have our LLMs progressed to the point where I personally don’t have to debug it? Generally ✅ and that is where the game changes.
The net result is I don’t care what the code looks like. I care that I can turn my ideas and dreams into reality in a fraction of the time it took before. We can debate if that makes me less of a software engineer later — right now I’m having more fun than I’ve had in years.
