Hacker News Reader

Comment by rf15 | Hacker News Reader

This is pointing out one factor of vibecoding that is talked about too little: that it feels good, and that this feeling often clouds people's judgment on what is actually achieved (i.e. you lost control of the code and are running more and more frictionless on hopes and dreams)

View on HN

It feels good to some people. Personally I have difficulty relating to that, it’s antithetical to important parts of what I value about software development. Feeling good for me comes from deeply understanding the problem and the code, and knowing how they do match up.

I agree with you. I had done a bit of vibe coding over the weekend. Not once did it feel good. Most of the time it produced things which are close to what I needed, but not quite hitting the mark. Partially probably because I'm not explaining myself in sufficient detail to AI, but the way I work is not working very well with super detailed spec ahead of development. I used to always develop understanding of the project while working on it.

I feel more lost and unsure instead of good - because I didn't write the code, so I don't have its internal structure in my head and since I didn't write it there's nothing to be proud of.

Yep, I agree 100%. People have described AI coding as "delegating". But there's a reason I'm an IC and not a manager. It's because I don't want to delegate to someone else who does the work, I want to do the work. There's no joy to be had in having someone else do the work at my behest.

If the future of the technology industry truly is having AI write software for you, then I will do what I have to do. At the end of the day I have to put food on the table. But I will hate every second of my job at that point, and it sucks ass.

Definitely. And it’s hard to separate out whether the person is actually more productive or feels more productive.

Yes, this (higher perceived vs. lower actual productivity) was probably at least true for early 2025.

https://metr.org/blog/2025-07-10-early-2025-ai-experienced-o...

For me, it feels good if I get it right. But unfortunately, there are many times, even with plan mode and everything specced, where after a few hours of chipping away and refactoring the problem by the agent, I realised that I can throw the whole thing away and do over. Then it feels horrible. It feels especially horrible because it feels like you have done nothing for that time and learned nothing.

If you succeed you will have also learned nothing, but would be blissfully unaware of it. (that was my reason to quit AI coding)

But why does it "feel good", if in fact it does?

I tried writing a small utility library using Windows Copilot, just for some experience with the tach (OK, not the highest tech, but I am 73 this year) and found it mildly impressive, but quite slow compared to what I would have done myself to get some quality out of it. It didn't make me feel good, particularly.

It _does_ feel good, I know what you mean. I don’t understand why exactly but there’s def an emotion associated with vibe coding. It may be related to the feeling you get when you get some code working and finish a requirement or solve a problem. Maybe vibe coding gives you a shortcut to that endorphin. I think it’s going to be particularly important to manage that feeling and balance with reality. You know, I wonder how similar this reaction is to the endorphins from YouTube shorts or other social media. If it’s as addicting (and it’s looking that way) but requires a subscription and tied to work instead of entertainment then the justification for the billions and billions of investment dollars is obvious. Interesting times indeed.

https://www.fast.ai/posts/2026-01-28-dark-flow/

It's just like gambling addiction.

I like "vibe doc reading" and "vibe code explanation" but am continually frustrated with vibe coding. I can certainly generate code but it's definitely not my style and I feel reluctant to put my name on it since it's frequently non trivial to completely understand and validate when you're not actually writing it. Additionally, I find vibe coding to generate very verbose and overly abstracted code that's harder to read. I have to spend time pairing the generated code back down and removing things that really weren't needed.

Conversely I feel like this is talked about a lot. I think this is a sort of essential cognitive dissonance that is present in many scenarios we're already beyond comfortable with, such as hiring consultants or off-shoring or adopting the latest hot framework. We are a species that likes things that feel good even if they're bad for us.

We don't stand a chance and we know it.

> We don't stand a chance and we know it.

Drugs, alcoholism, overeating, orgies, doom scrolling, gambling.

Addictions are a problem or danger to humans, no doubt. But we don't stand a chance? I'm not sure the evidence supports your argument.

Yeah I get a lot of value from vibe coding and think it is the future of how we work but I’ve started to become suspicious of the pure dopamine rush it gives me. I don’t like that it is a strange combo of the sweaty feeling of playing StarCraft all night and finishing a term paper at the last minute.

I think it feels like shit, tbh. That's my biggest problem with it. The feedback on moment to moment is longer than building by myself and the almost there but not there sucks. Also, like the article states, waiting for the LLM is so fucking boring.

I'll also say that vibecoding only feels good until it doesn't. And then you realize you don't understand the huge mess of code you've just produced at all.

At least when I write by hand, I have a deep and intimate understanding of the system.

it feels good because we've turned coding into a gacha machine. you chase the high from when it works, and if it doesn't, you just throw more tokens at the problem.

> you lost control of the code and are running more and more frictionless on hopes and dreams

Your control over the code is your prompt. Write more detailed prompts and the control comes back. (The best part is that you can also work with the AI to come up with better prompts, but unlike with slop-written code, the result is bite-sized and easily surveyable.)

Also, you’re still in control of your code. It’s not an xor thing, the agent does its thing but the code is still there and yours. You can still adjust, fix, enhance etc. you’re still in control. The agent is there to help as much or as little as you want.

Control over code requires understanding it. That's what letting an LLM wrote everything takes away from you

I see what you mean but I haven’t had an LLM produce code that I didn’t understand or couldn’t follow. Also, if it does it’s pretty easy to ask it to explain it to you. I’ve asked for explanations for ridiculously long lists of tailwind css classes but that’s just a pet peeve really, I mean, I understand what they’re doing.

You what code is? A very detailed specification that drives a deterministic machine. Maybe we don't need to keep giving LLMs more details, maybe we could skip the middle man there

The gravity well's pull toward a management track, or in the very least the desire to align one's sympathies with management, is simply irresistible to some people. Unfortunately I do not think Hacker News is the best venue to discuss solutions to this issue.

Reply on news.ycombinator.com