Stop calling it vibe coding
davekiss.comVibe blogging about vibe coding, I don't know what I did to deserve being in this timeline but I want out of it.
Hard agree. AI Stans can’t help but get down on their knees so they can look cool for internet points.
The person who owns that website should feel absolute shame for posting that Trash. But they won’t because they’ll ask AI how they should feel and it’ll tell them they’re a super smart, good little developer!
hi! Dave here (a human, like you) — sometimes it's easy to forget there's a human on the other side, so I won't hold your thrashing too close.
1.) I've been writing both code and blogs for a long time. The feeling of having a skillset cultivated over a decade plus suddenly feel irrelevant is very real. Nothing about this post is unrepresentative of my perspective on what we're all going through. I don't think it's trash, but everyone is entitled to how they feel.
2.) I am shipping more code and words in the past month than I have in a long time, and yes, the process been AI-assisted — but _not_ AI derived. Anyone who has a personal blog can tell you first hand how it's the easiest channel to let slide, but I am of the camp that I'd rather have my thoughts exist in some form then die in the cavities of my head. This is true for both exploratory product ideas and recent personal writing.
3.) I'm no stranger to long-form articles. I can do it alone. Does that mean I should? I find it super interesting how those in tech are a-ok with AI-assisted programming, but other mediums are shunned. I'm a musician and write my music the old-fashioned way because I personally consider it my art. I probably wouldn't love listening to generative music. Some blogs I'd probably consider my art too. Others? I care more that the idea is represented in public than I do about the way it got there. My process has been dictate the idea -> work with Claude Opus 4.5 to structure it -> revise where the LLM misses the mark -> publish. It's been a force multiplier and I've had more traffic to my website in the past week than I had looking back at least 6 months. Clearly, the ideas resonate with others.
Does that mean I should stop? Is it more prudent that the idea dies due to its inability to climb the priority list? The whole point of these things are to use them as a tool. I'm not even defending this process, but I _am_ exploring what it means to use them effectively.
Good news; more on this will be posted soon, should you opt in(!) to read any further.
Here's my metric for "LLM-assisted" anything: if I, as someone who is well-informed on LLMs and pays close attention to the content I consume, can't tell an LLM generated the article/code/image/music/voice, it's fine and indicates that you as a person put enough effort into shaping the output to claim it as your own. If you can't even be bothered to do that much, to make it not completely obvious that an LLM generated it, you have no right to call it "assisted" or stake any ownership of the content. And I think it is frankly insulting to your readers that you would expect them to read something you couldn't even bother prioritising enough to write. If your ideas are actually worth sharing with the world, they would be worth making the time to write about.
And to be especially clear, my objection is not rooted in the moral aspect of LLM generation (although they are frequently used to plagiarise as well). My objection is that if I can tell the LLM generated it, it means the content is genuinely garbage. LLM output for writing is garbage. So too for code. So too for art. With enough human effort on top of it (usually more than doing it from scratch, IME), you can get it into a not-garbage state, but you didn't, and instead I wasted my time reading part of an article in which the writing was too bad for me to finish it.
From my perspective, the internet I once knew has been destroyed. The spambots won. Now spambots are openly accepted, promoted, reach the front page all the time. Somehow it is not a moderation policy to kill spambot posts and ban offenders. All because spambot technology advanced enough to fool people who are just consuming content passively. And now I have to sift through so, so much garbage to find the people who are still putting effort into creating content. Their work is being buried. It feels like living in a surreal nightmare for the past two years.
Footnote: Although I called the content you posted garbage, this is not an attack on you, because you did not write the content. I have not personally insulted you, so I hope that rather than taking this personally, you might reflect on whether this is really the way you want to express your ideas to the world. You might get views, but are your ideas really resonating? A lot of people mindlessly scroll the internet to kill time rather than critically engaging with content and digesting it. Is your ideal audience one who reads your idea and then forgets about it seconds later as they move on to the next shiny thing on the screen?
Simonw had the same thoughts a while ago:
I'm using "agentic coding" like the author here, I'm not interested in "vibes" being a part of the name for what I do because there is no "vibes" about it
> “Vibe engineering”, really? ... Is this a stupid name? Yeah, probably. ... I’m ready to reclaim vibes for something more constructive.
Simon should have stopped after the first answer imho, we can find a better name for what we do as our profession increasingly adopts agent assisted development
Stop calling them vibrators, they're agentic lovers.
> Here’s the question a lot of developers are sitting with: do we lose some of our cognitive abilities if we’re reduced to reviewers? If we’re not writing the code, are we still thinking deeply about the problems? Or are we reduced to pattern-matching against LLM output, slowly atrophying the muscles we spent years building?
No, absolutely not. Please don't start to think that by not doing stuff on your own, you miss out on keeping yourself sharp and trained and on improving in this stuff.
If developers who have adopted LLMs for writing code start to think this, that might scare away some from using LLMs at all, and this weakens my evil plan to make a lot of money fixing their code with my intact coding abilities, especially when these LLM tools are going to get crazy expensive because at some point they will need to be sustainably funded, when I'll still be somewhat affordable in comparison. I'd rather get rich quick by doing the same thing I've been doing for years, don't take this from me please.
> especially when these LLM tools are going to get crazy expensive because at some point they will need to be sustainably funded
You say this with such certainty when progress is actually proving otherwise.
LLMS are only going to get cheaper.
There will always be expensive models, because they use the latest tech and infra, but that doesn't mean we need them for everything...
But year after year we see free or local LLMs become more powerful.
Haha. I can totally see a future where business hire outside consultants to fix in-house AI developed code and to train developers on AI pitfalls.
May be the design will be outsourced to consultancies and then implementation using AI is left to in house talent. May be new recommendation and lists like OWASP top 10 and 12 factors will emerge for AI too.
Author is conflating vibe coding with LLM-assisted development. Many such cases.
The author spent much time defining the differences. Where do you see conflation?
They explain it as if one has transformed into the other. Vibe coding has always been where you ignore the output.
The author describes how their use of agents has transformed, not how "vibe" coding has transformed. See these passages for example
> But a year of daily use changes things. The way most engineers I know actually work with these tools now—myself included—has evolved into something different.
> What we’re actually doing # So if “vibe coding” doesn’t describe it, what does? I’ve been calling it agentic coding. The distinction matters: it’s using AI agents while maintaining the expertise and judgment that keeps the output good, rather than letting it rip with zero validation.
See also their passages around how using vibe coding for side projects is different from agentic coding for professional projects
> Vibe coding has always been where you ignore the output.
Does this definition apply to (say) Fortran compilers from the 50s?
The difference is the author has zero clue what they’re talking about and likely vibe coded this whole waste of time post.
lol thanks
> We’ve all collectively figured out what sustainable AI-assisted development looks like, and it turns out to be more structured than those early vibes.
Is this satire of AI hype? Genuinely can't tell if he's being serious.
A few days ago, I commented on a post. My comment was that vibe coding should no longer have a negative connotation because it is quickly, if not already, becoming the primary way to write software.
Soon we'll have GPT 5.2 Pro performance at Cerebras inference latency. Then it'll get even better.
It would shock me if humans are still writing code in 5 years, unless it's for fun.
If nobody does, what will these models be trained on?
I guess only people writing code manually will be "frontier".
> I guess only people writing code manually will be "frontier".
Personally, I believe that seems almost inevitable. Ever since Sonnet 3.5 came out, my assumption has been that most devs will either need to become largely product people, or find a new career.[0] I mean, most devs have been implementing known patterns most of the time, right? That seems on track to be completely replaced by agentic dev tools, does it not?
The best terms for this new role I can think of are "Product Developer," or "Software Product Developer." This is a product-minded person who is able to create non-frontier software using agentic dev tools.
What I am really curious about is if devs at the frontier will be more, or less likely, to publish their work as open source in the future.
[0] I think there has been a clear divide here in excitement regarding agentic dev tools along these lines. Product-minded devs are really into the new tools. Non-product minded, more code-driven devs seem to be far less excited.
Synthetic data pipelines, which is already in use for like over a year now.
Can't tell if your text implies a '/s' or not.
I mean, a lot of code is written using the NPM ecosystem, that doesn't make it good though. There are tons of objectively bad things that are popular. "Tracking personal data on webpages is already becoming the primary way of advertising" "Touchscreens are quickly becoming the primary way of interacting with one's car"