logoalt Hacker News

altairprimeyesterday at 8:34 PM1 replyview on HN

AI coding versus AI writing may be a useful lens to focus through; while I personally abhor both, HN seems extremely positive about the former and (now) extremely negative about the latter. I hope that policy is extended to all YC startups someday :)


Replies

alteromtoday at 12:15 AM

>AI coding versus AI writing may be a useful lens to focus through; while I personally abhor both, HN seems extremely positive about the former and (now) extremely negative about the latter. I hope that policy is extended to all YC startups someday :)

Coding is writing though.

Somehow, HN can say that "code is written once and read many times", and insist that code isn't writing at the same time.

All programming languages were created with the express purpose of allowing humans to express their ideas in a way that other humans can understand while simultaneously being convertible into machine code in a precise enough way.

Code has style, code has readability, and when it comes to algorithms, code is often the best way to communicate them (I haven't seen a CS book without at least some pseudocode in it).

Code is supposed to tell what a program does, and what it's for— to a human that wants to understand or change that behavior.

A human who doesn't have this need has no need for the code.

Programming languages make coding less tedious and more efficient (compared to writing assembly) as a side effect.

The primary purpose is facilitating communication about what the machine should do from humans and to humans.

Sure, the scope of ideas computer languages are tailored to facilitate expression in is not universally broad. But that doesn't mean we're not writing when we write code. Lawyers writing a legal argument are still writing, even when they are doing so in very specific, formal language. Mathematicians are still writing papers.

It takes extreme mental gymnastics to consider coding (which is universally an act of producing text) to not be a form of writing.

To that end, having a negative view towards LLM writing while cheering on LLM coding seems (to me) to be borderline schizophrenic.

The people that advocate AI coding for throwaway projects, or using LLMs as a tool to get more insight into codebases make points that I can understand.

But a day or two ago I've responded to a person that argued that Open Source is no longer necessary because you can just vibe code anything. Many others advocate for using agentic coding in production religiously.

Apparently, this is not incompatible with rejecting AI writing at the same time.

I'd be very curious to hear about how people are overcoming this sort of cognitive dissonance.

show 1 reply