Hacker News Reader

Comment by Aurornis | Hacker News Reader

I think the bifurcation is between people who want to write code and people who want to have the end product of the code.

People who want to write code hate AI because it's doing the part they wanted to do.

People who want the end product of the code love AI because they want anything that helps them get to the end product faster.

The person who wrote this post feels oddly in neither camp. They like playing with the AI and seeing what comes out the other end. Some of the projects they boast about having built aren't even usable projects, like when they had it mock up a UI of a product and then got bored and moved on to the next before writing a backend.

View on HN

Agree this is largely the hidden issue much AI discussion misses.

AI industrialized a previously creative output. If you enjoyed the writing of code this is a nightmare. If writing code was a chore to solve a problem, this is a blessing.

I don’t think this is the full picture. Plenty of people who like writing code are happy to delete code too.

When you write the code you learn from writing the code.

When you have an LLM produce something and then delete, you didn’t learn much.

This is it for me. I can put up with a lot of struggle and non-sense if I feel like I'm actually developing some knowledge or skill. Vibe coding is a lot of the non-sense but little of the system knowledge actually sticks. Unless you very diligently try to understand every line, you're just as oblivious about the problem as you were when you started. At that point though the AI isn't making you any faster.

you didn’t learn much.

And that's entirely your fault, not the LLM's.

'Entirely' is unfair. I do learn from working with AIs. Yet in the production pipeline their output can be so entangled (especially for non-text) that it's difficult to decompose and adjust without great effort.

Just today I was toying with AI to make some bumper music. It came up with some great phrases and fragments. But its 'song' output is a hilarious mess, and feels like I'd be better off starting from scratch and taking only the bits that work.

Then there's the ethical question of where those clever lyrics even came from. Perhaps just lifted from niche works I never heard before.

You learnt (or should have learnt) that the strategy taken by the llm was a poor one. You then explore why it didn't work, and what would have been a better approach. You think about the structure of your code base, the issues you still need to solve, and then you roll back and start again - this time prompting a different approach.

I love coding. I've always loved coding. But coding with llms has helped me step back and see the overall design of a project better. And llms know tricks I didn't, and some of these are neat, and now I know them too. So, personally, I feel like I'm still learning, a lot. It's just a different way of learning - and one that makes coding interesting all over again.

(That said, I cannot use agentic systems and embed myself fully in the loop, and still code myself. Sometimes its faster to code the next change manually, sometimes it's faster to guide an llm. But it's still interesting, and it's even more fun, at least to me.)

I learn what is possible.

I know, for example. That I can take a Cheap Yellow Display from Aliexpress and have it show the train timetables and weather. BUT I also know that the timetable public Web API requires encryption - which is a pain with the small chip in the display - but the private API with an API key works via plain http.

I didn't write a single line of code, but the display is on my hallway cabinet, working 100% the way I like it.

> I think the bifurcation is between people who want to write code and people who want to have the end product of the code

I think most developers are both! Depends on the task. Sometimes I want the result, sometimes I want the process.

Also sometimes, if I want the process, it’s because it’s something I want to have intimate knowledge of. There’s a practical benefit to writing stuff yourself, even if most of the time that benefit is tiny.

We should brace ourselves to find less and less satisfaction from the process as we become more comfortable embracing AI. At least, that's what the first chapter in the evolution of the AI-assisted software developer role has taught me.

I'll only trust that if you're the one being oncall for that product. Or at least is directly responsible if it ever fails.

I agree but I think its also more nuanced. I am also happy to use AI to operate at a higher level faster. I think that is somewhere in between. For instance, maybe I have some specific ideas and architecture I want to try for building a durable workflow engine. I dont just say 'make me a durable workout engine'. I'm very intentional about what it's doing at a system level but I am happy to cede the low level implementation details. If things work out, refactoring those details to my liking is also easy.

Well, you are learning something, just the thing you’re learning has an even shorter usable lifespan than programming languages, namely you’re learning what works to get useful responses from ai agents. Whether or not that has value to you is a different matter, but it’s worth bearing in mind something is being learned, even if it’s not engineering or programming.

You're learning to manage idiot savants, which is a very useful skill.

> You're learning to manage idiot savants, which is a very useful skill.

I think the real bifurcation is whether you will settle on that belief.

Some of us are settling on the belief that the idiot savant, lacking the coherence of a functional mind, cannot be managed. It's essentially a chaos agent masquerading as something more cooperative.

The thing is, LLMs are more like the opposite: Sophisticated ignoramuses.

> namely you’re learning what works to get useful responses from ai agents.

Having worked a lot with AI agents, I don't agree.

AI agents are amazing at producing response and results that look correct as long as you don't look too closely.

Even when I try to write extremely detailed specs and test harnesses, even Opus 4.8 and GPT-5.5 on max will find creative new ways to write code that breaks under real use cases.

Doing throwaway LLM output, playing with it a little bit, and then calling it done will create a false sense that you're really good at getting LLMs to produce working things.

I’m not sure that’s true. All my side projects exist to scratch my own itch, so the appeal to hop straight from design to done is really appealing.

But it’s never really that straightforward.

There is some truth to the idea that some people enjoy it and others do not. I haven’t seen a pattern between them.

> All my side projects exist to scratch my own itch,

That's exactly what the second group in my comment was meant to address. You enjoy the end product, therefore being able to skip the code writing is appealing.

The blog post is about someone who was having AI write a lot of side projects that they weren't even interesting in using. The post directly states that they were not useful, they didn't need them, and they weren't interested in maintaining or even finishing them.

The bifurcation is probably mostly just along the lines of slop tolerance than whether they "like" to code or whether they're a boomer or whatever.

There are a lot of people with high slop tolerance and who are seemingly prepared to endure the side effects of that.

Reply on news.ycombinator.com