My Personal Philosophy on Vibe Coding
Published · last edited · 487 words
Vibe coding is dangerous. It lets me build faster at the speed of light, and lose control of what I build even faster. I strive to leverage that speed without ever losing control of what I do.
The source of this philosophy comes from the article, Why superintelligence != superhuman.
AI knows a million more libraries, coding languages, and techniques than I can even name off the top of my head. But right now, I am still the master of my own destiny.
It's a fact that AI can do a LOT, and I take full advantage of that. It can read 500 papers in the time for me to read one. It can summarize and aggregate information from hundreds of sources it's seeing for the first time. It can write thousands of lines of code in the time it takes me to write 100. It can surface details that I haven't thought of. It's POWERFUL, and I will never deny that, and I don't think anyone can really put it back in Pandora's Box and lock the lid.
But none of that replaces years of human intuition, knowledge, or creativity I've gathered. These facts AUGMENT me, not replace, not overtake.
I trust it to surface facts; I don't always trust its interpretation of these facts, and often double-check and confirm them. It can guide my decisions, but not make them. It can help me understand, build, and control greater and bigger things, but I will always maintain control at the levels I care about.
When doing the things I do care about, I make sure that I can explain the code produced at the level that matters. Maybe when I build an application, I won't examine every single line in detail all the time; but I will make sure I can explain the purpose of the functions that comprise each class, how each class interacts and builds off one another, and why it is implemented in that way.
I believe in this age of automation, the most valuable is whatever’s authentically YOURS. I do not let AI directly write or edit things I care about. I never let it have the final say in any decision I make. Because whatever it is — it might be messy, it might be ill-informed, it might not be optimal — but it’s mine, well and truly mine. And if I make the wrong choice, then I’ll be damned, but keep in mind…
LLMs learn patterns from data, but they don't own my goals or bear the consequences of my decisions. They don't see what I see and think the way I do. I'm a person, and I can learn and grow as we have for millennia before the thought of a transistor existed. In today's era, our tools let us accomplish much more than ever before -- but you'll still find me at the helm of it all.