For those keeping track, this is now (at least) the third wholly or partially AI-generated article on the theme of "taste" as enduring source of human comparative advantage to hit the HN front page.
Saying "more" seems generous. I don't trust Pangram at all, and I also don't trust the author at all. Both can be trustless charletans simultaneously, and I feel this speaks volumes about the state of digital media today.
Pangram is a complete scam. There is no durable way to detect whether something was written with AI. I've had folks say they ran pangram on my writing and it was "100% AI". Except I don't use AI to write or edit anything I post publicly.
So you would trust a known writers word over non deterministic software ? I read New Orleans will soon let AI handle 911 calls, Hope the AI believes the calls are from humans :)
This article is not AI generated. Everyone is asserting it is but the author is well known as a critic of these tools in his communities. Why would he lie and use it to write an article with his genuine views, edit it to assuage the feelings of HN commenters, and then double down? Speedrunning losing his credibility, or, his writing style got on some people's nerves?
I showed some old Olga Tokarczuk essays to my friends to pick what was ai slop and what was not as experiment, pre- 2020 ones from Czuły Narrator to be exact. Guess what, they picked the ai one as human and her as most ai. Maybe because she uses a lot of big words and em dashes.
We are at the point, when there is too much cognitive energy needed to discern writing from human, human ai assisted and all ai. Should we know make mistakes on purpose to avoid being labelled as ai slop? :)
This is why I'm tired of the "this is AI slop" commentary. I'm well past trying to determine whether a human wrote something, and have fallen back to my old ways of critical reading to search for value in the thing I'm reading. And if I don't find it, it doesn't matter who wrote it. Everyone is so excited to be clever and call out others for using AI, when it's far more useful to cite and elevate good work, regardless of tooling.
On the topic of taste, I think of this quote in Susan Sontag's "Notes On 'Camp'" often:
“For taste governs every free — as opposed to rote — human response. Nothing is more decisive. There is taste in people, visual taste, taste in emotion — and there is taste in acts, taste in morality. Intelligence, as well, is really a kind of taste: taste in ideas. (One of the facts to be reckoned with is that taste tends to develop very unevenly. It's rare that the same person has good visual taste and good taste in people and taste in ideas.)
Taste has no system and no proofs. But there is something like a logic of taste: the consistent sensibility which underlies and gives rise to a certain taste. A sensibility is almost, but not quite, ineffable. Any sensibility which can be crammed into the mold of a system, or handled with the rough tools of proof, is no longer a sensibility at all. It has hardened into an idea..."
An important feature of taste is also that it is highly variable between people, and because there is no system or proof of good taste, we cannot really say that one person's taste is objectively better than another's. This is what causes one person to build a large software system in a classical OOP language while another builds almost the same system in a typed functional language, each following their own internal logic of what is good taste. In the end we learn something from it, and maybe someone who worked on both projects refine their taste and starts preferring something that lies in between the two paradigms and comes up with a new interesting programming language.
One thing I fear is going to happen when more and more people use LLM's for coding is that they will default to the "average taste" of the LLM, and we will stop getting new interesting ideas about how to build things.
> One thing I fear is going to happen when more and more people use LLM's for coding is that they will default to the "average taste" of the LLM, and we will stop getting new interesting ideas about how to build things.
I don't think it will have any impact at all on creativity in coding.
Creativity is rewarded in humans, just not by their employment where both training examples and monetary rewards can be found, rather the creativity is rewarded by challenges such as code golf, demoscene, and so on.
Go to a job interview and I have to know the shibboleths of Clean Code, SOLID, an arbitrary one of five frameworks, three different ways to handle asynchronous code, and four design patterns, and the magic word "agile". Invent something new at the workplace? If you're in a small team, or high enough in the hierarchy, fine. Big team? People will reject it with reference to those same shibboleths.
I wrestle with this also. A colleague pointed out that LLMs reduce language choice to an implementation detail… don’t have to know assembly to write python, and now you don’t have to know python to write a functional app. So why stick with python when rust might be better? Leave that to the LLM and let your engineering effort go toward designing the app instead of operating a programming language.
In which case yeah, projects are likely to converge on whatever pattern LLMs broadly fall into. But maybe that part doesn’t really matter. Not every project needs the engineer to care about that. If I’m not writing hardware level code, I don’t care what instructions my python compiles to, either.
Meanwhile nothing prevents someone from caring about these details if wanted or needed. Someone who knows and cares might absolutely do a better job than an LLM. Those people won’t necessarily go away.
But there -will- be a whole new category of software contributors who have ideas they don’t know how to express in a software language. Those ideas might well be very tasteful. So it could be argued that software development as a field will gain access to a whole new range of tastes.
I don’t know the answers but it all seems plausible
To be a bit more precise, SE is the practice to organize large codebase in a meaningful and human friendly way. AI will attract more hackers doing fun projects without careing about implementation details under a much higher level of abstraction, but they will not introduce more variation in SE.
>because there is no system or proof of good taste, we cannot really say that one person's taste is objectively better than another's
We can't say it in times of a general culture of confusion, like now. In times when there was an established cultural and artistic hierarchy, they could.
> One thing I fear is going to happen when more and more people use LLM's for coding is that they will default to the "average taste" of the LLM, and we will stop getting new interesting ideas about how to build things.
It is a possibility. But I also think we're underestimating the human need to create cool, unique things. Like yes most of the projects in this situation will have "average taste" but most of the projects out there today already do. Very few people have true taste and I don't think we're just going to see them disappear or be satisfied with LLMs feeding them "average taste." They will either likely stop using LLMs if that is the case or find ways around it to make it do things the way they want to, hence preserving their taste.
But I think it is a bit of an overblown fear that LLMs will just kill creativity and innovation in coding. Hackers will be hackin
>But I think it is a bit of an overblown fear that LLMs will just kill creativity and innovation in coding. Hackers will be hackin
Hacking and a taste for hacking is a muscle you develop. LLMs will make most people not train it at all. Same way a huge number of people are obese and out of shape. For LLMs to be bad news in this regard, it doesn't matter if some people will still find a way to do something creative. As long as far more people (than without LLMs) don't, that's already bad news.
Yes, there are exercise nuts who are exercising like crazy, but the average population is getting weak and obese.
The same with mental muscles.
How many people can do mental calculations in their head nowadays?
What about remembering several phones numbers?
What about simply navigating with map but without GPS? (Or, heaven forbid, just by common sense and observation?)
Properly thinking about hard problems is exhausting. Majority will be glad to offload the hard thinking to an AI 'assistant'. The same way we offloaded computing to calculators, and memory and navigation to phones.
What remains is the easily manipulated lizard brain...
I also don't think it will kill creativity in the space, although my phrasing might have suggested that, just that it might diminish it and make implementation styles more conformative.
On the other hand, we might trade that for creativity in other more high level designs where things like coding style is an unimportant detail for the experiment. For example, I have started playing around with microcontrollers because I can have the LLM generate the software while I focus on learning how to solder and think about how to modularize my hobby automation system. That's creativity and taste at a higher level which I didn't have access to before because I didn't want to invest my precious free time in learning the software ecosystem.
I feel two ways about it. I've written some algorithms and architectures for motion control that were good, after many years of chewing on the problem through iterations. I don't think I'll replicate that effort or those results, now that I use LLM's so much.
On the other hand, I did that in pursuit of a goal, and LLM's take me further, faster, toward that goal.
"Consistent sensibility" is a tasty little phrase.
I think taste is developed from a combination of sensual attunement (clarity of perception and focus) and experience (exposure to relevant, high quality 'stuff'), but the catalyst is joy, or at least some kind of pleasure.
I'm a little confused. She follows describing people with good/bad taste immediately with the statement that there is no system for taste. If it is all relative, then how is she determining that someone has good or bad taste? Is there anything between moral relativism and mathematical proof? Just your own ineffable belief that you are right?
Even if something is relative, you can compare it to what other people believe. If most people who know person X think X has "good taste" (that is, that their tastes more or less agree, or at least that they aspire to having similar taste like how more people admire those who like classical music than listen to it themselves) then it is reasonable to say that person X has "good taste".
Someone with questions about this excerpt of Sontag on taste would do well to read the rest of the work, which does a fuller job of drawing distinctions between sensibilities like taste, and systematic ideas like good/bad, morality, relativism, mathematics, proofs, and wrong/right.
> If it is all relative, then how is she determining that someone has good or bad taste? Is there anything between moral relativism and mathematical proof?
Social proof. Everyone has taste and everyone's taste in a given space is influenced and informed by the taste of others. It's mostly a consensus model. Newer members of a given community arise into said community and learn from it what "good" means, which can sound authoritarian but is really just... voluntary. If you like, for example, mens formal wear you will follow publications and noted figures in the mens formal wear space, you will learn what "good" means to them, and will emulate it out of a desire to fit in with them. As you get more knowledgeable in that space, you will develop your taste, enhanced by them, and ideally eventually contribute your own ideas to that mix. Some will be accepted, some will be rejected, again by group consensus.
It's a lot like language. If you want to speak a language you have to learn it first. If you go to France and simply blurt out French words at people, even with perfect annunciation, you will be incomprehensible to them. You learn structure, sentences, how to communicate ideas, how to inflect tone. How to tell the difference between telling someone to fuck off when they cut you off in traffic versus telling them to fuck off when they make a horrific pun. And, as people speak that language and use it daily, it will reliably change. The meanings of so many words in every language today are wildly different than they were even a hundred years ago.
i'd completely forgotten about notes on camp, had to read that in modernist criticism class. good stuff.
i see parallels to zen in the art of motorcycle maintenance as well, who's whole thesis if i can recall is something like, "quality cannot be put into words." Its a pre-rational thing that, try as you might, can't be pinned down.
This is an incredibly insightful comment, thank you. I had not read "Notes On 'Camp'" at the time of writing (and nor do much reading for the post in particular) but it seems very intriguing from what I can find on the internet.
I have mixed feelings... I stumbled upon, read a post-mortem that looks even more AI-generated than the article.
"That says something about my writing. It says nothing about who wrote it." - Two contradicting sentences, author's writing skills are part of the author as-a-whole
"The short sentences, the reversals, the one-word lines" and then "I will write the next one plainer.". - even more?
And... I wonder if this is like some agent wrote article, and, in the loop, added port-mortem, after visiting hacker-news comments and manipulate that this is human written. Or maybe author is just trolling us with AI-like prose...
> and this post will go down as my not-the-proudest, however, I take your criticism to heart—although not personally—and strive to improve.
I just do not understand why people - even when actively denying LLM usage - use the fucking em dash.
I don't care if it is the "right" thing to use there. The character has been associated with slop. Why do you keep using it without even just acknowledging that you did.
Just use any other character - like I am doing right now.
It must be AI indeed. Humans cannot be this dense.
___
Oh god I just saw the annotations on the side
> A fair pushback I got:
Dude. Why. Oh god. Why do you talk like claude.
I want to believe you. I really do, but hell. Even claude could likely tell you that you sound like claude.
I just do not comprehend how that can be the hill people choose to die on.
Like.. I get it. It's wrong. It feels wrong to give the AI nonsense just an inch and all.
But in the grand scheme of things, it's just highly ineffective dogmatism that makes people unheard, miserable and just further increases struggle.
Which people are free to do, in the same way that I am free to purposefully drop a massive hammer on my foot just to make some kind of statement.
Hurts me a lot, rest of the world stays entirely indifferent.
a) Using some other horizontal line is not in the class of "wrong" that would warrant this sort of reaction. You're debating the pixel count. Not the object. The aspect ratio of it.
b) If you're acting out of fear that's dumb, yes. If you however simply pick your battles for efficiency, that's not "the terrorists win". That's "you win" instead of "you loose due to attrition".
__
I've also consulted with the LLM regarding the "class of wrong" thing, because I thought it might be not quite clear.
So what I meant by that is that it's not giving up actual words or dumbing down your sentences.
Just swapping out a character that is practically indistiguishable for the untrained eye.
Hell, whole classes of social engineering attacks rely on that very property.
I know I know. "Wehret den Anfängen". Slippery slope.
But those are insane as a default reaction to everything.
___
I get it. SV sucks. HN sucks. The world sucks. Can't let it also take this little thing.
If it can take all the big things, then at least it won't take this tiny tidbit.
Emotionally, I get it.
But that's not how you make rational decisions. Nor is it how we would win.
Th emdash is not something I use ever (before or after AI slop), but telling people not to use it because it’s associated with slop is not the way.
It’s like when people tried to tell everyone to stop using the ok symbol with their hands because a small group of hateful fucks pretended it meant something else. It only switches meaning if you let it, and stop using it yourself.
What happens when AI adopts something else? Do we give that up too?
With all of that said, this guy is definitely AI lol. He’s gotta be trolling us by doubling down.
I mean you would stop using the OK Hands thing when trying to communicate with a group that has for (possibly arbitrary) reasons decided that that symbol is bad, and thus anything you say is bad, right?
At least temporarily until you got synced up with that group?
There is a place for the strategy of just continuing to use something and push through the noise there, but the em dash is not in that place.
Common perception of it has too much momentum you cannot overcome like that, which - at the time of writing - makes it a bad choice.
How it will look like in 12 months we shall see, but today, I would advise against using it (without acknowledgement) in texts that talk about AI.
No, that’s my whole point. I would not stop. If someone tries to attribute malice to something I do that has none, I might explain to them how misguided they are, but I’m not going to give in just because someone can’t be bothered to understand context.
I think it’s fine to use as one of many signals to identify AI (not the only) but, again, avoiding its use is not going to solve anything.
I'm being serious here: Looking for em-dashes is a terrible way to "scan" for slop. You need to look deeper than that and if you did, there's clearly a voice there. If you let yourself develop an "em-dash allergy" and think this helps you avoid LLM-slop, you're scanning for the wrong things and it's going to do the opposite for you. Actually stay on your toes and don't become this complacent.
sigh The em-dash is something we've read a lot from GOOD PUBLISHED WRITING so we as human authors tend to emulate what we read as good writing. It's a publishing artifact that's migrated. It's not a sign of the apocalypse. Neither is AI, but hey.
Em dashes are mine. I will use them whether they have been appropriated by LLMs or not. I used to use commas in their place. That resulted in a much more dense post layout and far less readable text.
FWIW I actually use three dashes (-) and my blog's build system transforms those into em dashes. This is not a defense, but nor is "this character is associated with slop" an argument. Yes LLMs use lots of words, phrases and patterns. Would you rather I stop using altogether? I am all for criticism, but please pick something I can actually change and improve without cutting a part of my style for no reason other than "Claude uses it too!"
> This is not a defense, but nor is "this character is associated with slop" an argument.
You're misunderstanding me. It's not an argument. I have no strong opinions about it.
I am telling you how people perceive it. You can (and should!) use that information however you like, but don't mistake this as some sort of opinion thing.
I've observed the world. I believe it operates like this. Feel free to discard if you have data that suggests it operates otherwise.
__
Don't mistake my delivery of that for personal outrage about you or that unicode character.
I just get very annoyed when I see systems not working optimally with failure modes that to me seem.. dumb.
If you allow me to quote the LLM:
> He isn't trying to police notashelf's writing. He doesn't care about notashelf's writing! He is just a guy who noticed a pothole in the road, pointed it out, and is now watching in pure fascination as people fiercely defend their God-given right to drive their cars straight into it.
Yeah, I am suprised this article ranks so high. It's either written by AI or it's poorly written. Very little to show on behalf of "taste".
"Taste" is one of these words that have been bandied about recently, in addition to "moat". I've never seen a productive conversation stem from either of those mentiosn.
yes this is 100% AI, the article and the 'post mortem'. People are so gullible. Anyone who has read a lot of AI can spot it instantly, but Pangram also says 99% of this is AI. So many signs, and no there were no human authors who wrote like this
Please, I couldn't care what Pangram (or any other "detector") says. If it feels like AI, does it even matter if LLM was co-author? It lands in the same "bad writing, I won't continue to read that" category.
That said I am recently getting paranoid about YT videos, I hear Claude traces in every 5th script. If people are going to spend substantial time recording and editing, is it worth "saving" time on most important part?
"it's not just ...., it's ..." is in EVERY youtube video I watch. There's got to be a reason AI uses it so much. Maybe it was already prevalent in bad scripts and we never noticed it before. Sort of like the "Webster dictionary defines ... as " cliche.
An antithesis is a completely valid stylistic device. It's just very easy to overuse or to use as a cliché when you're not using language as a tool for a specific purpose, but just want to sound smart or think "fancy stylistic devices = good writing".
I really don't know, their GH profile [1] and their projects, even the recent ones [2], seem pretty non-LLM.
IMO the post-mortem is more a satirical response to the HN criticism, doubling down on the "AI slop" writing style.
> IMO the post-mortem is more a satirical response to the HN criticism
They have edited it again to be slightly more emotional. It's why their "On Taste" closure doesn't actually pertain to Taste anymore. Of course, they wouldn't notice the mismatch - they didn't write it in the first place.
They aren't being satirical. They're pissed they got called them out. Were they not, they'd not bother with edits.
It's neither satirical, nor by an agent on the loose/in a loop. I read the comment section with good intentions, assuming good faith---and in their defense there is plenty---but the amount of misattribution is counterproductive. Nothing about the post-mortem section is "pissed off". I wrote it in complete serenity, like I am writing this comment to explain. There are three groups of commenters:
1. Skeptic, but engaging in good faith - thank you, much appreciated
2. Skeptic, not truly engaging with the post - that's fine, can't win 'em all
3. Completely disengaged with anything in writing. They're under the impression that I am criticizing their taste, and felt the need to "strike back" - what?
The post-mortem is a response to all three groups. I've explained the situation, apologized for what I agree I did wrong, and clarified my boundary. It's not worth pointing time and time again it was not LLM authored, and I do not have anything to add.
Fine for me, I didn't ask any justification from you, but the post-mortem does/did sound as LLM-speech, and presuming it human I also presumed some kind of humor in it (I was probably projecting).
I'm a writer with what may be an unpopular opinion. I have a college degree in creative writing; I do technical writing and programming for a living, and I used to publish technical articles on Linux, Python, Elixir, and other topics—complete with em dashes—long before AI was a thing.
When confronting AI as a writer and deciding how I feel about it, I've landed on the conclusion that I've never accused mathematicians of churning out "calculator slop." A calculator is just a tool to make the complex work of mathematics easier. I now view AI as a calculator for writing, among other things. If the idea is good and the information is sound, I don't let the occasional awkward phrasing bother me. The truth is, most people can't write very well anyway, so if AI helps them communicate better—even if it is obviously generated by AI— more power to them.
- At least to me, AI writing (especially from Claude) is actively painful to read. It's worse than the average person who has flawed English (or just sucks at writing) because of how verbose and stylistically repetitive it is.
- AI writing is a strong probabilistic signal that what I'm reading is going to be crap in terms of the actual ideas and content too, even if it isn't the case 100% of the time. I just don't want to waste my time on that kind of thing.
When I use it for writing I usually have an idea or thesis to explore already. I typically write it out stream of consciousness. The core idea is mine.
I'm not using LLMs to think for me, instead just edit the writing so it has proper grammar and the run ons are reduced.
Even better for something that has data. You can pull PDFs with encoded tables and extract the data into consistent units to compare. You can have it write code to do it and cross check the outputs. You can do that in a minute. Would take hours without an LLM.
Calculating is not math. Calculators make mathematicians better mathematicians by allowing them to spend their time on actual mathematics.
Writing is thinking.
LLM writing makes authors worsr authors and even dumber people overall.
Also a calculator gives the same answer as a human (except for human errors where the calculator isa strict improvement). LLMs absolutely do not give the same output as humans.
There's a bit of a difference to me between a calculator and writing.. Creativity. Words aren't just a vessel for ideas. If there were different calculation structures etc and space for people to develop their own styles and approaches to calculate things, then sure.
A person's writing style is developed through what they read. And if human's writing style becomes more homogeneous through AI then that's sort of a shame.
Somebody at $formerwork wrote a "skill" that takes a poorly/briefly specified Jira ticket, and "fleshes it out", purportedly so that when an agent picks up the ticket it has more context.
Really at that point, with head in hands, rather than try to explain the foolishness, I decided I was better out.
"Zoom and enhance" used to be a joke/meme, apparently it's now "engineering".
One line of idea spread into 30 is the worse way to convene an idea. Only people terrible at writing think using AI for prose is good. Most of the times, they also hate reading it, but love asking people to read their AI slop.
I don’t like the term taste, but the problem that I have is that LLMs don’t seem to work “good enough”. They seem to be able to solve the immediate problem, but stacking this on the scale of 3-4 devs over 6 months or so doesn’t seem to produce anything.
One thing that I’m particularly frustrated with is the writing quality of LLMs. Like this is the thing that they should be able to do, but I would say almost everything they write has almost no signal.
Over a mid sized AI generated codebase that means I’m reading like 500 words to figure out what a module is even doing.
I’ve started requiring that unless absolutely necessary, all comments and docstrings be one-liners in our code 10k+ codebase(s). LLM-generated code has a tendency to “over-justify”, which is reasonable when first reviewing the code as a human.
But then, once it’s passed the first human it should be considered human-to-human communication rather than LLM-to-human communication, and the long, winding, and often repeated justifications can be significantly condensed.
Also you can’t let it run loose with unit tests unless you want to be blasted in the face with hundreds of lines of absurd test fixture preparation.
I'll risk sounding antagonistic and ask you this: if LLMs are not "good enough", why are they still around?
Obviously "good enough" was a poor pick of words of words on my behalf---and I'll gladly own it---but they should be "good enough" for something if a significant amount of resources keep getting allocated to them. Yes on a very personal level you look at LLM-generated code and think to yourself "wow this is garbage" but what about the people, as pointed out many times in this thread, that simply do not care? Does that not count as "good enough"?
The market logic isn’t going to be defensible. The market is highly irrational and prone to be influenced by those with the most capital, not the, “invisible hand.” Cars are freaking everywhere not because they are the best solution of transportation but because people with vested interest in the sale of oil and cars made it so by buying politicians, starting political and marketing campaigns, and on for generations. It took decades for the car to become the de-facto mode of transportation in North America let alone the rest of the world. It’s not good for us. It’s highly irrational to prefer cars when freaking wildfires and climate change threaten our very existence. In a similar vein, LLMs as they are, are not around because they are necessary and superior.
You need roads, traffic lights, drivers and pedestrians that respect rules, not too many holy cows in the way, not too much CO2 in your quasi-self-sustaining spaceship, lungs and livers that can deal with the soot, a tolerance for high-velocity impacts risks, not too many other cars around, ...
> Cars are actually the best solution to transportation
"the" is the error here.
There's places and people for whom cars are so. There's other cases where public transport covers the people while business-owned vans cover deliveries of bulky items, which is essentially how I've lived the last 8 years, though for "bulky" I should note I've seen someone take an actual kitchen skink on public transport here in Berlin.
I've also found some islands here Berlin with no bridge connection to the mainland and hence no practical use for a car at those locations; that said, my main reference for "car free" domiciles would be Sark: https://en.wikipedia.org/wiki/Sark
The current kind of car is extremely good at doing what it does. It's just that most of the benefits are short term, and for the car user (and the oil salesman), while the costs are borne by everyone else, and some of the worst ones are long term.
> Yes on a very personal level you look at LLM-generated code and think to yourself "wow this is garbage" but what about the people, as pointed out many times in this thread, that simply do not care? Does that not count as "good enough"?
Aside from the "amateur" qualifier, the Booch quote below provides insight to those questions:
The amateur software engineer is always in search of magic,
some sensational method or tool whose application promises
to render software development trivial. It is the mark of
the professional software engineer to know that no such
panacea exist.[0]
"Code" is shorthand for "encode", which can be defined as "encoding a solution to a problem", which implies knowing what problem to solve, and thus ultimately implying understanding of same.
So "good enough" depends on the stakeholders' definition of that valuation.
IOW, if a person tasked with delivering a solution can instead deliver garbage and still get paid for it, then it is "good enough" for some definition of same. If the people approving a person's remuneration are not satisfied with this type of work product, then the answer as to whether it is "good enough" or not will be quickly answered.
The use of “amateur” is indeed a very poor choice. An amateur is someone who seeks to master a skill or perform a craft for the love of the art, not for money. It’s the professional who is looking for the shortcut, not the amateur.
The amateur is the one who hones a tool’s edge until it’s razor sharp, and cuts away all unnecessary detail until they’ve found the essence of what they’re trying to achieve. The work of an amateur is often highly impractical and uneconomical, but sublimely beautiful.
I think you are torturing this comparison by using different meanings of amateur and professional
A professional does mean "someone who does something for money" but it also often just means "someone who is very good at something"
Similarly, amateur does mean "someone who does something out of passion without wanting/needing to be paid" but also often just means "someone who is just starting out, is new at something, and is probably not good enough to be a (paid) professional doing it"
You're talking about the "passionate" amateur and comparing it to the "getting paid" professional. I think many people talking about amateurs and professionals are mostly talking about "amateurs (newbs)" and "professionals (very experienced)"
An amateur is not a master nor a disciple of a master (e.g. future master).
The vast majority of amateurs have no ambitions nor passion and simply give up early on, that's a rose tinted view.
Since they don't get paid to do it, they have to allocate their precious free time toward the skill, which means they don't have those 40 hour weeks where they spend all their time honing their skill to begin with.
Even someone who does a skill half heartedly without passion nor ambition is going to learn faster with those 40 hours at his disposal.
>The amateur is the one who hones a tool’s edge until it’s razor sharp, and cuts away all unnecessary detail until they’ve found the essence of what they’re trying to achieve.
That's not an amateur and you know that. By your logic Japanese master craftsmen doing this are amateurs.
They're here, because even "bad" code has uses. We can have quantity over quality. We can generate throw-away programs (where the lack of maintainability doesn't matter). We can solve some problems by brute force (keep trying until tests pass). For people who can't program, even bad code can do more than no code.
LLMs are also useful as a natural language interface. They're pretty useful for search (as RAG and autoresearch, not the glue-on-pizza Google search crap). They're good for filtering and classification against rules that can be specified in natural language instead of special syntax or by training dedicated models.
> They're here, because even "bad" code has uses. We can have quantity over quality. We can generate throw-away programs (where the lack of maintainability doesn't matter). We can solve some problems by brute force (keep trying until tests pass). For people who can't program, even bad code can do more than no code.
If this is an acceptable definition of software engineering productivity, where "quantity over quality" is prized, then hire me.
Because I guarantee I can produce dozens of PRs daily, each having tens of thousands of LoC deltas (pick any language you desire), none of which having research and/or understanding underpinning them, and all easily quantified as being "brute force."
I'm not happy about the drop in quality either, and I'm not advocating for just making software shittier.
However, look at history of DSLR cameras vs mobile phone cameras. Mobile phones were always worse than large format cameras, and we've got a massive quantity-over-quality explosion in photos taken. It didn't replace professional photography, but let people upload photos of their lunch to social media.
I think in programming we're getting frustrated because we're trying to naively use high-volume low-quality code generators in our previous low-volume high-quality workflows. We're still discovering what's the programming equivalent of bathroom selfies, photos of receipts and QR codes.
There's a quote I forgot that says: "There is a Quality to Quantity".
I think it was from a Chinese person, as they took manufacturing of cheap things, low quality stuff, but managed to scale it to mass quantities.
I agree with that, it may not be reliable, full featured, well designed, ergonomic, etc. But quantity enables things and thus is a quality in itself. It allows use-cases that benefit from accessible cheap software. Those use-cases were not viable prior, because cost/time was prohibitive.
> Because I guarantee I can produce dozens of PRs daily, each having tens of thousands of LoC deltas (pick any language you desire), none of which having research and/or understanding underpinning them, and all easily quantified as being "brute force."
I can guarantee that you cannot produce such PRs in the same quantity/quality than a modern LLM. The speed at which today's LLMs output code is superhuman.
It's textbook historical revisionism. It's not like we didn't have mountains of bad code in the pre-LLM era. I'm old enough to remember endless complaint posts about having to wotk with legacy codebases full of horrible spaghetti code. Somebody took a long time to write it by hand.
Have you realized you are attacking a straw-man? Nobody said all “hand written” code is good. I _think_ the implicit argument is: “all vibed code lacks cohesion”.
Breaking down long term problems into small shorter problems we can solve one after another until the long term problem is solved is the essence of engineering.
No, the "essence" of engineering is having a cohesive vision for a larger plan.
No quality large scale project can exist without it. And any seasoned engineer should understand by now that no SOTA LLM can produce quality engineering at scale.
Long term projects that survived on short term wins are still around. But you're going to want to check how big the graveyard is of "project collapsed under tech debt from short term decisions" to see whether it's a good strategy.
Nah. You can succeed in every single individual thing but fail at the overall project because the parts don't align into anything that makes sense.
If you need examples, look at game dev. There's plenty of games that have good execution but aren't fun to play or they're a confusing mess because there wasn't good overall direction.
Videogames are a weird comparison because they are basically the ultimate culmination of the combination of both engineering and art. The two can crossover in weird ways but still have distinct aspects that do not really compromise the other.
Just because a videogame isn’t fun doesn’t mean there wasnt a cohesive engineering vision. God of War: Ragnarok is a good example of this. Excellent technical execution and (arguably) great direction but a terribly boring game.
And many games that are terribly engineered are also incredibly fun. Dark Souls is a good example - runs like ass, looks very rough even by the standards of the time when the game released, with terrible enemy AI in a combat focused game - one might even call it a confusing mess (and I laugh at the notion that it doesn’t have good direction)… despite all this it’s a classic and arguably the most influential game since its release.
Elden Ring carries a lot of legacy baggage from years of iteration on the same engine as ~~Dark~~ Demons Souls - and it’s widely considered From Software’s magnum opus and a masterpiece. Yet considering the scope of the game it is a great example of how small wins added up can solve large problems when executed with the skill and judgment gained from experience. So one might still consider it a success in terms of “engineering”.
Would From Software have been able to make the same game using Unreal Engine 5 built from scratch with dime a dozen “Unreal experts” brought in off the street? Not a chance. Want to know why western game development studios are suffering right now? Compare American layoff culture and race to the bottom economics to the retention rates of Japanese game studios like Capcom, Nintendo and other heavy hitters.
For many of the most marvellous things humans have ever built, if you take a peak behind the curtain you’ll still probably find some amount of duct tape and popsicles sticks somewhere keeping it all together.
Fun, however, is a matter of taste and you can’t engineer that.
But that means they're good enough for the short them. This is why I find it such an interesting question to wield, but also, it is why my wording was poor in the post.
Short term can, IME, be very short. I've seen people generate, say, a bash script with an LLM. It's generated: short term, the problem is "solved": we've generated a bash script.
… but does it work? Someone comes along, reviews it, "this is garbage, and does not do what it says it purports to do". Perhaps it even gave an output: the script computed … something, but it's just GIGO.
But that "check if this works" friction is the same friction that is what people try to avoid by generating it with an LLM in the first place. If you're too lazy to write the script, you're practically by definition too lazy to verify it.
If this is your working environment, it sounds like quite an unusual place.
I literally can't imagine generating a script with an LLM without testing it at all.
Bash is one of those situations where LLMs can do really well. No human on Earth can remember all of the commands and even fewer humans can remember all of the switches for all of the commands.
This kind of remembering, searching, and assembling is exactly what LLMs are good at - as long as you're not writing a gigantic build system with hundreds of moving parts, in which case you should probably be using something more streamlined anyway.
You can have another agent write the tests and verify the former agent ? This is pretty basic stuff. Makes me question if people are actually trying to use AI
They're "good enough", but they're not good. If you're trying to run a business, programmer skill isn't worth the money you spend, LLMs are cheaper. If you care about the quality of what you produce, they're trash.
Programming is no longer skilled labor. Quality is a hobby.
If your business involves storing any kind of customer data or handling payments, you probably should still hire a programmer who at least reviews the results of your vibe coding. Getting hacked is bad for business.
AI is just fine at security audits these days. It's better at finding bugs than it is at writing code, overall. You need to prompt for it, but that's not skilled labor.
I'm also not sure getting hacked is so bad for business. Most businesses that got hacked seem to still be around just fine.
I'd agree with the OP you are responding, what he means is that this isn't just about taste. It's that the code, while it works, and the text, while it has the ideas in it, is still not to the level of a human author. It's not that it isn't tasteful, in the sense of having the flair of specific author X, but it is verbose, goes around the point to get to it, etc.
Taste seems to imply an art, but I don't think we're at the point of cleverness, of LLM innovating a new beautiful programming paradigm or pattern out of immaculate taste for code. We're just talking more basic, even bad devs have more terseness, more clear intent in their prose and code, etc.
It's "good enough" for many use-cases, but it could be better, even before expecting the most beautifully clear code from the best devs out there, it has not reached the level of the average one in readability and clarity of expression (both code and prose). That annoys the author, he has to intervene manually to trim all the fat, and therefore not "good enough".
The market desperately wants to replace wages and the idea of a machine doing the work you previously needed to do before is so enticing, and the initial results so promising, that the rest doesn't matter.
Look at how many AI lay offs there already were a year or two ago - back when almost everyone would agree they weren't good enough. If good enough didn't matter then does it matter how when AI is better?
And despite what you might be trying to prove, it's obvious I'm replying to an LLM (a green account at that), but it's not for you, I don't expect you to bother reading it.
It’s a good question and I don’t doubt they’re useful.
But I mean they’re not good enough to do the things I need them to. Not enough that any difference can be chalked up to “taste”.
Put it this way, if you hired a very good handyman to build you a cathedral, you wouldn’t stand looking at the smouldering wreckage of the construction site saying “oh well that’s just a matter of taste”.
One part of the answer is marketing it to people looking for an easy way out without having to do the work or thinking at all. There's value in that as shown in everything like things for the home that purport to make your daily life easier. LLMs market themselves in much the same way
Not disagreeing with you but the "resource allocation" which I presume to hint at the fallacy that capitalism is somehow great or optiomal at allocating resources is not really true. Just look at any big company such as meta and see how much inefficiency and bad choices there is.
Stock market as a measuring stick is completely irrational.
There's the hope that they'll make all employees redundant and that businesses that adopted AI quickly will be able to exist as a pure money-printing entity without the liabilities of products, customers, or employees.
Put another way, LLMs are just another part of enshittification.
> if LLMs are not "good enough", why are they still around
They are not good enough yet, the people using these tools every day, and with a high bar for quality and design 'taste' notice the rough edges.
They are still around because LLMs are plenty good enough for casual consumers, but where 'taste' really starts to matter in design or in writing, the rough edges show up much more obviously and users who have been working with these tools feel it immediately.
This is just a problem of articulating what you want. Most people can't write for shit, I'm sorry. Because they maybe took one writing class in college. It's a skill in it's own. LLMs write better than anyone I've ever worked with and the reasons are obvious. If it rambles or is too "purple", that's on the person prompting it, because they can't write for shit to begin with.
> LLMs write better than anyone I've ever worked with
And yet every time I see AI-written documentation, there's paragraphs worth of throat clearing and space-wasters like the word "genuinely" for things that could have been outlined in a couple of bullet points.
Since they can summarize, you'd really think they would be better at condensing their own output and cutting out the filler after. I wonder if you could use a specialized second pass for it or something?
In my recent experience it seems to be a contextual issue. Llms are constantly being dropped into new situations where they have to rediscover high level and cross cutting information about the system they're working on from contextual clues.
The internal representations of this state and its projection back out to human language wouldn't be as concise as that of a practitioner or team that develop their own verbiage and ontology over time molded to their system.
This verbosity might get better as we figure out better ways for agents to learn long term and use that knowledge to adapt to the users and projects over time.
There might also be some good harness improvements we could consider like forked output streams or multiple long lived filter subagents to ensure that output appropriate for thinking is separate from code output and separate from output given to the user driving the session.
I think there are really two types of use cases for LLMs here, and people that get value out of them.
First are those that care very little about the quality of the work, or the process behind it, and just want a 'finished' product no matter what. These are many of the folks vibe-coding everything without even looking at the output, and a depressingly number of those getting hacked or what not.
But hey, if you've got an idea for a complex app or website but no interest in actually building it, an LLM will get you... something vaguely like what you wanted.
Second are small projects and small sections of existing ones, most of which easily fit inside the LLM's context window. If you're making a fairly basic WordPress plugin or React component or one page website, an LLM will probably be able to handle it just fine. Heck, it might not even look all that different from what a human might have created code wise.
One of the big issues we have those is that plenty of people and companies are using these tools despite having requirements that aren't met by them in the slightest. If you're working on a Google/Microsoft/Meta scale product where performance and security and code quality are of the utmost importance, then an LLM isn't going to be a great fit. Unfortunately, that's exactly where this technology is being used, and the results are becoming more and more obvious.
For me taste is judgement about things that don't matter. Like you have taste in clothes. But a judge doesn't have bad taste when he makes a wrong decision, he has bad judgement.
We've been talking about taste for a year now as it relates to AI, so I'm going to counterpoint it: if competitors can reproduce your features, UX, and visual decisions within days, your taste is not really an advantage. How many apps implemented pull down to refresh once one person did it? How many people even remember (or care) whose taste spawned that UX pattern? [1]
AI is not inaugurating an age of taste. It is shortening the half-life of taste and commoditizing software. What happens when software features, UX, and visual fidelity aren't competitive advantages anymore? If you ask me, it looks like business did in the 90s. We'll probably see the return of department store software in the next few years.
Microsoft was superior in both in the windows 95-2000 era, and with Metro (on phone, on desktop it was half-assed). The problem was when they started to try matching Apple, instead of staying true to the core of their own aesthetics.
Both the Fisher-Price XP style which was bot zeitgeist, and an attempt to answer for the skeumorphic tragedy apple was pursuing, and the recent floats, spacings, central taskbar mimichking the dock, etc...
I'm not really convinced Metro was better than anything Apple at the time, even though it was refreshing and interesting. Was it actually easier to use? I didn't really find that nor see it with others.
I used Windows Phone, and it was great, minimal, intuitive, and prodcutive for most usecases I had (email, texting, calls, maps/navigation, banking, listening to music, light productivity, taxi). It did not fit the gimmicky ad ridden hellscape we live in now, I'm not sure how it got so far in Microsoft. Possibly from some old memories of living of selling products instead of creating addictions and selling data of the addicts.
Version 7 and 8 were great UX-wise, I have mixed memories about 10, mostly performance regressions bugged me, but was still faster than most non-flagship androids of the time... yet sometimes it felt sluggish to WP7 on potato hardware. The typography based UI was both fast and very readable and focused for well thought out apps. The standardized navigation buttons (on android button order may vary be vendor/model), and the clear gesture based navigation instead of hidden intearctions everywhere (long press, 3d press, app switching gestures etc), but using gestures for in-app navigaton were all easy to understand and grasp even by the elderly.
It was readable, well organized, nobullshit, but that is not what the masses crave. Metro was like how a "dumb smartphone" should be, in my opinion.
On desktop ot didn't work well. The live tiles were still a good idea, and if the taskbar supported them, it might have been a good idea (or not, was not tried, but could have be a bit like the old NeXT style refined, not stagnated and skeumorphized like on the mac).
That's because Microsoft understood taste as aesthetic - "we can put it in a pretty box", rather than ethic - fundamental to the process of making the product.
Most of their endless UI/UX fails are organisational in nature, no amount of design can create beauty from an org structure which generated stupid requirements.
And nobody there cared enough or had enough power to force the org into a shape compatible with Apple-level design outcomes.
Microsoft's taste was "good enough". They lost because of an app ecosystem, and because of Apple's brand power, the same way beautiful Chinese shoes will lose to Nike.
Nokia was able to match and surpass Apple in taste on all fronts with their N9 phone and MeeGo OS. So it is possible, but how to do it I couldn't tell, since all large organizations are saturated in corruption.
Funny you mention Apple because they are proof that taste can be king. Yeah they also kidnap their users into their ecosystem and make smart business decisions, but what really differentiates them is taste. And Microsoft is sort of the counterpoint. There's more than one way to succeed I guess.
It's not that the machines were ugly or badly designed (albeit not as good as the 80s heyday), but they were unable to compete on price/performance and OS tech stack, and that in turn was making them uncompetitive in application support, which would likely have killed them.
It's an interesting alt history - who would have won with no Apple around? I reckon we'd all own a lot more Sony and Nokia gear, it's possible the Cambrian explosion of weird featurephone form factors would have carried on a lot longer, perhaps the black rectangle form factor would never have won out.
Because let's admit it, later it was a taste of one guy who got another "chance" or the company got a "chance" to be lead by him. Now as he is gone the company still carries that taste legacy with better or worse results.
I think the point is without taste you wouldn't even know what things should be copied out of the countless competitors. How would you even know in the first place if a competitor did a particular thing tastefully without a good sense of taste yourself?
Also, there are a lot of areas where taste will absolutely reign supreme (and has been), especially anything related to media or entertainment (books, articles, videos, games, music, etc.)
I can see a possible future where taste is automated. If you can effortlessly make 1000s of iterations then you can test it on real people 1000s of times. The ones that do better you keep.
Another possible future is that taste is not automated but its not part of the designer's job (if designers still exist). Instead some AI will generate 1000 of iterations and people will be employed to evaluate 100s of software UI each day for taste. These will be taste farms if you like. These people full time job will be saying yes or no.
I found this post yesterday on this exact same topic. As someone who is interested in the space of creating visual things and its design choices, it's a bit disheartening to see the current situation, and brings you to the point of "why even try doing anything anymore" rather than just follow the copywagon.
I feel this will be a similar effect to the current slowdown of hiring of juniors: you're killing the pipeline of future seniors that lead development.
Couldn’t disagree more. AI is a button that you press and magically get results. The ability to judge the quality of those results is “taste”.
Anyone can push the button, but not everyone can discern whether the result is a good one or not. That skill is either learned or to some extent inherent in one’s abilities.
My example I’ve been thinking about is: what will be the first new, big programming language in the age of LLMs? Theoretically, anyone could go out and make it right now. The skills and expertise required are well-encoded within all of the frontier models. The thing is, without taste, the language devolves to a useless state rather quickly. On smaller scales, I’ve seen this with API design and even object oriented design. I’ve also seen it in AI generated images and music: devoid of taste, and numbingly facile and generic.
Edit: not to toot my own horn, but “numbingly facile and generic” funny enough is probably a phrase that an LLM would never create without significant prompting to refine writing style.
That's not really what I think of when I think of taste in software development. I think of taste more in the sense of how to go about architecting something, what problems are worth solving, what technologies should be used, how to structure the code and documentation, how the APIs should function, what the scope of the thing should be, knowing what to do and what not to do, yada yada. That's all taste and LLMs suck at it (noting that taste is very subjective).
I'd argue that a well thought UX isn't easy to copy in the first place because it's not something that can easily be put into word.
It's the same for a lot of things, I have some clothes that I love because of their, design, quality and more ... something intangible that I have hard time describing and that can easily be explained with: It was done with taste.
It was always easy to copy software (but yes, even easier now). The hard part is competing with the software that you're copying. Software does not exist in a vacuum. They usually have an already established user base, relationships with partner businesses, more domain knowledge due to having to actually think about the problems they are solving and so on.
You can start by copying but at some point you'll have to differentiate, and you might not be able to do so if you don't understand how your sloppy vibe-copied software works.
It's not about being the only person doing it, it's about choosing to include it. Many don't know what works and what doesn't. Being able to discern a set of features that are cohesive to a smooth experience is taste. Innovating and establishing oneself as the first person to deliver those experiences is a moat in itself.
Because reproduction is reactive. If an organisation only clones others' style then they'll always be behind the curve and will be widely recognised to be derivative and lacking their own 'taste'.
Look how long it takes Chinese brands to transition from being clones to establishing their own corporate image.
Dumbest idea ever. You are composing some text in a form, or filling in fields. You want to scroll up just a little to see something out of the viewport. Oops, your little swipe is misinterpreted as a refresh gesture: the input is gone.
I think it's impossible for it not to be poorly implemented because the gesture is inherently ambiguous (or becomes so when you overload it).
I think what "poorly implemented" actually means for touch screen gestures is to be lacking the customizability to reassign or disable infuriating behaviors.
Gesture configurability is nonnegotiable: when implementing gestures, have the configuration system in place first and make sure every single behavior is tied to a preference.
Imagine if artists were embracing AI the same way coders were being forced to. All that would be left there is taste, and your argument applies as readily there as well.
This article resonated tremendously with me, speaking as someone who has been coding since the 1980s and had to develop my own taste the hard way, through lots and lots of mistakes. Whenever I see a software demo built by someone mostly using agents to do the work, I have to question whether it's any good on the inside - or whether any real intuition or judgment went into its design. On the flip side, maybe none of that matters anymore - as long as it works, who cares how it was built.
For majority of people, including business executives it's the latter. No one cares about artisan code. Software is just a means to an end for them and not an art form.
People treat their computers like any other appliance. They don't care if it's riddled with bugs, they'll gladly restart their computer or refresh the page at infinitum to get it to work.
I think it's also why marketing and designers get more of a say in the software than engineers. UI polish matters more since it increases sales.
This is the hardest thing to come into terms with in this industry.
I'm quite sad that I have to disagree with you and agree with GP: if quality siftware mattered to people, Microsoft would have gone out of business a long time ago.
If quality software mattered, javascript wouldn't be the most popular way to write desktop software.
If users cared at all about quality software, we wouldn't have 800MB chat apps.
Of course yes, and it's even easier in some cases. Since there's a language that defines the functionality (e.g: formal methods) that the agent can use/veriy against
Yes but that 10% is all that matter to the people that matter to the business. Like a car, most people care about how it looks and how they interact with it... the only people who care what is under the hood, for the most part, are other engineers.
It resonated with me too, and even as a once fairly prolific coder that now uses agents to do all the coding, I feel like this is all relevant and does matter. The agent does whatever you ask and provides thin judgements based on what it thinks is your motivation and context, even feeding it the context it will often make bad generic choices. I think this is the part hard to optimize because it's not a verifiable reward, it's the feel of the software. I think Herdr is a great example, I've tried several dozen terminal multiplexers / session aggregators and they are all vibed and different, but this one felt so much better to use because of its choices and actually increased the number of parallel sessions I was able to keep going. My own projects I find I have to stress every detail, that is what makes the software good now is really diving into it, the experience feels like just as much of a craft as coding was and 'taste' captures the concept for me.
“as long as it works” is doing a lot of work here. Did the people building it say “no, again” enough times, and at the right times, in order to make something that _actually_ works instead of _merely_ works?
For what it is worth, I appreciate you sharing your experience. Looks like you have more programming experience than I have in life, so your comment matters to me. It also gets rather close to something I wanted to evaluate in the post: "whether any real intuition or judgment went into [its] design". I think that's a big question to dwell on these days.
Ironically, though, I had written a post that had people debating whether I put any intuition or judgement into. Would've been real ironic to instead add a post-scriptum that said "Yep, none of it matters. I just slopped something in 5 minutes and it easily went to the frontpage. Look how easy it is!" but unfortunately that is not why I write, not what I intended to achieve, and not something I am equipped to do. I care about my work, which is why I care about your comment as well.
People who care about their health aren't ignorant about that. I think that can apply to software too: if you care about safety or privacy you might think twice before using vibe-coded software.
That is a significant leap in logic. A vibecoding tool is not a McDouble. Nor is Chris Kempczinski the same person as Dario Amodei.
The biggest difference is that there are only so many ways to eat a McDouble, but you can actually wield a vibe coding tool with skill. There's nothing stopping you from looking at the code and asking for structural reworks or refactors that please your code taste. In fact one of the best use cases for AI coding tools is vibe refactoring old apps where the code is straight up horrible.
> but you can actually wield a vibe coding tool with skill. There's nothing stopping you from looking at the code
It’s hard to give something like “vibe coding” a rigid definition, but I think we can all agree that it’s inversely correlated with "looking at the code". The more I examine the code while building, the less the activity is about vibes.
So, used this way, it's true that it's not a McDouble anymore, but it's also not so much of a vibecoding tool anymore either.
i have never had a "mcdonald's mcdouble", whatever that is. i cook my own burgers because i am a human being and i want to eat real food, not nauseating slop that should only be fed to livestock
The LLM has its own intuition and judgement which at this point is quite good. I’m not just throwing that around, I use LLMs 12-14 hours a day and for a long time reviewed atleast 75% of the code.
I've spent a day fixing a feature my very junior colleague pushed, and which was merged by my should-have-known-better colleagues who have almost as much experience as me.
We have access to the same models. Why did her version suck and my features don't need fixing?
Generating the first version is almost free now. Understanding why that version is subtly wrong, debugging it in production, maintaining it six months later and deciding whether the abstraction was a mistake are all still expensive...
Yeah. The Fable oneshot is a myth, the software will still require constant refinement and maintenance. I've been daily driving my "vibecoded" projects and they're never really finished, there's always something to do.
Very nice article. I agree with it, call it taste, call it judgment, call it quality, I want things done the way I see as satisfactory, and it requires me to make LLMs do things differently several times. Like exploring a tree with a heuristic they don't have and need to rely on us humans to give.
Hard disagree on "Good enough", most things are mediocre at best. Despite the massive AI boom I can at best count on one hand "fresh" products made with AI that are meaningful in day-to-day usage. Not to mention most of them seem to be targeting AI-space itself instead of bridging some actual business use case. Just look at all the products on ShowHN, you can "tell" a vibed product in seconds.
I do notice and realize, and I've gotten to the point where I had to fork and patch most of them to fix the shitty AI generated code.
It's so bad, and so everywhere, that I'm genuinely considering switching to a pre-AI stable distro and waiting 10 years for the AI hyped to be over before updating any piece of software ever again.
I'd believe it, my spotify breaks all the time now and I seem to remember them announcing none of their devs code anymore.
not 10 seconds ago I had to force close it to get it to stop playing music on my phone.
I think you agree with what is in the essence of the post, but has been worded poorly. I do agree that most things are mediocre at best. Hell, I'd argue that most of my comments on everyday observations are negative. Ranging from "this sucks :/" to "holy fuck". They are, however, "good enough" for some and despite how clever I must to notice how much something sucks, that does nothing to the fact that it will receive adoption. Turns out I'm not clever. Huh.
In hindsight, I have worded this rather poorly and let some phrases do a lot of heavily lifting for me in the post. Ultimately, I agree with you. Knowing, for this instance, is not half the battle.
I've run this thought experiment a number of times with different sections of my (relatively) diverse friend circle. I ask them a version of this question on repeat: "what if AI automates X aspect of your work?", and the final answer (either because of frustration with the loopy nature of the question or a genuine stopping point) arrives somewhere along the lines of "my judgement," "my knack," "my embodied nature," or "my quality of decision making."
I think the core still leads back to human agency and the ability to consider aspects that wouldn't fit into an LLMs limited context window or be able to be vectorized into a DB, including ultra-long-term consequences (especially those with great thinking abilities).
I still happen to think that AI/LLMs will never be able to "fully" replace humans, because the evolutionary process that led to our cognitive abilities and the way we train LLMs are vastly different thanks to different pressures, but maybe that's just me defending the last bastions of our collective humanity as a human; I don't know what else to root for.
Right now we're all training AI models on our taste. All of our chat logs are being uploaded, and the AI companies are training the models on them to get them right the first time.
It will soon no longer require taste. Or at the very least be 95% of the way there on the first run.
Google also uses ML to train the Waymos, they just have better sensory capabilities. Lidar is much closer to how the human eye functions (inverted, I suppose) than cameras are anyway. But FSD is still really cool, you just have to make sure you don't take your hands off the wheel!
> somewhere along the lines of "my judgement," "my knack," "my embodied nature," or "my quality of decision making."
Do your friends actually believe that AI can't automate these parts of their work? I think my answer would be that I'll just have to learn to do something else.
> I still happen to think that AI/LLMs will never be able to "fully" replace humans
Do you mean replace humans completely or just replace all the tasks humans don't want to do? I would hope that humans will still have a reason to live even if all the work is done by computers, and that the computers wouldn't intentionally wipe us all out.
Judgment was all there ever was to begin with. It's why it's what God himself does, although I'd be curious if there are deities that don't involve judgment in some form.
Anyway, I completely agree and the past several years have been an absurd journey into borderline psychosis with some people.
If you have taste in writing code, you spend less time on bugs and can implment features more quickly. This is still true with LLMs if you learn how to direct them instead of having them direct you.
Someone else without taste can re-implement what you have done with AI. But it will take them more effort, more time, and more money.
But that's always been the value in writing code. Any MBA can end up producing software to accomplish something by hiring other people. But their company might not get off the ground because of the expense of not having technical competency (or engineers with taste) at their core.
Taste was all that was left when product design peaked in like 2010-12 too and it was a great time.
But the advent of design systems and software like Sketch and Figma pretty much automated the job away. You didn’t need the raw Photoshop skills anymore - any CMU grad could be a top tier designer now.
I see the same thing basically happening here. Taste will matter for a while but eventually people are just choosing the most accepted/expected optimal choices and anything that falls outside the trend will be “not good” in the same way skeuomorphism in design is no longer good.
So we’re lucky we’re in a taste era - it’s a lot of fun. You (yes you!) can still change it, shape it, establish the status quo of tomorrow
Design is not moving pixels around in Photoshop or Figma. CMU grad could be a designer, but to be a great designer you do have to have... well, taste. Figma didn’t change that one bit.
I never quite fell for the scam of “designer who can’t photoshop”.
You can’t do product design without visual design - and many designers don’t. They’re picking from UI libraries like Tailwind or MUI, which is really something any developer could do, and many do. That’s why the job mostly died.
Even the most abstract of painters, fine artists, and so-on should have good fundamentals. They’ve gone through the discipline and developed taste, they don’t choose it from a library.
The CMU grad (using this as a stand-in now for any so-called educated hack, sorry CMU) is skipping all that and asserting personal authority without proving or living anything. And that shows IMO.
My probably unpopular opinion is that using AI can be very creative and you can get into flow states, and the quality of outputs vary by skill of the user.
> Taste was all that was left when product design peaked in like 2010-12 too and it was a great time.
I'm not sure I agree. I spent a good chunk of those years converting designer's PS sketches into CSS. Bootstrap and friends were a godsend, because even if standard design systems over-optimize for mobile, they are still far more functional than most designers's mediocre and incomplete views of user experience.
I have yet to see a good design falling out of people building flying castles in Figma, disconnected from the realities and basics of web development's languages HTML and CSS.
Taste is something you use to build the guardrails of AI over time and many paper cuts endured. Once you have a big enough set of these (call them rules) recorded, they become your development substrate. You narrow the paths the agents can take to arrive at a coherent system. If you don't do that, superficially it may look similar, but it's not and users will be able to tell.
"De gustibus non est disputandum"
("In matters of taste, there can be no disputes")
One thing I'm excited for is the expanded ability to modify and create my own interfaces and software so I don't have to be subjected to the bad taste of others. An example of this is something like Chrome reading mode that removes all the stupid decoration from what I'm trying to read.
My workflow has become a sort of "living process". When I encounter friction in some part of it, I focus my attention on it, it adapts, grows from from an annoyance to something that I delight in when using.
Some small examples: Google Chrome doesn't have default shortcuts for accessing bookmarks so I made an extension for myself that does this. I have a SketchyBar setup that I'm constantly updating. Recently added my CC/Codex plan usage to it so I'm no longer constantly running /status in the different CLIs. Etc, etc.
> Suppose you have taste. Suppose you paid the full price and you can feel the sag in the sentence and the wrongness in the function.
>
> Congratulations! You now ship at exactly the same speed as the person who cannot.
Yes, previously the person who didn't understand shipped faster. It was a long standing problem. The people who crapped, got the bonus, and ran were much faster than the patient engineer of quality.
By the way, what were the essays that formulated Claude Style? if the author was indeed formed by them, I'd like to read them - or at least enough of them to understand.
Also just a shout out to the author - you're getting a lot of flak here, but your article hit me at the right time in the right way. I trust you're telling the truth about writing it yourself, and thank you.
I keep seeing people trying to define judgement and taste and then they don’t actually like cite Kant’s Critique of Judgement where this is basically the entire concern.
I'm not only referring to that, I'm also referring to a reductionist view when I read "Taste Is All That's Left" - which is a statement that has been mindlessly parroted around as of late.
Yeah even if Dev Schumer is some absurd hype artist he was correct to note that it is judgement and taste, not taste alone, that is most key now. Although he attributed it to the models own subjectivity.
I'm not disagreeing with you at all, but I have personally found most philosophy "classics" painfully hard to get into. That doesn't mean they shouldn't be respected, but the truth is that sections of society (like myself) do rely on the people specializing in philosophy parsing it into current language and metaphor.
Look I mean, many famous artists don't understand their work at all. You don't have to have read philosophy to create something meaningful or make an insightful comment. But if you are going to start using terms like Judgement or Taste, you really ought to know where they come from.
There's a lot here that I enjoy muddling on. And I think taste is very vital. Taste, ambition, seeing possibility: these resonate to me, long resonate, as what I've wanted to see.
And honestly: I've been tired of an age of engineers that are beat down, that only want simple, that finding the most blunt approach is the only way. It has felt tasteless. Anti-ambitious. It often is still very sensible and practical and what you should shoot for! But there are also people wandering around now trying much harder! And I am excited for those frontiers! I think it will challenge and shake the foundations that we've accepted as true, as There Is No Alternative, in exciting bold and fun ways.
I am however pretty unconvinced by the article. There's a lot that doesn't quite work for me, that isn't building the case I'd like, whose takes are off from my read.
> The output is good enough—that is the problem. Good enough is a solvent. It dissolves the reason to do better. For as long as making things was expensive, the expense did quiet work on our behalf. It rationed output.
Agreed about the rationing of output. I do think that the article though continues to show a magical thinking. That we have these things now and they autonomously do the thing. That the LLM's have solved it all.
The article itself goes on at length about how nebulous and abstract taste is:
> Taste is that. It is the compressed, wordless verdict you reach faster than you can justify. It is the “no, again” you say to yourself with total conviction and no available argument
But this contrasts so distinctly to me against "the output is good enough". Is it? That depends. That depends on your taste. The proximate first results come quick. But the technics underneath? Those themselves, in my view, rely enormously on engineering taste to support and advance. I think we see a very similar sort of magical delivery thinking, very clearly on display here:
> When the factories came, they could suddenly make everything—cheaply, uniformly, by the thousand.
As if there was some magical "good enough" transition where suddenly the aliens came and gave us this box that just does the thing. As if we discovered the right formulas and math and now: we had production. Again I think there's just an enormous amount of work and taste that is still actively required inside the factory box, that building the industrial processes is incredibly intense & difficult, even though we have reliable industry-line production and now robots doing the labor.
There's still so many gems, so much lovely material for thought. I love the provocations here, and I think there's a lot of great calibration.
> The friction was not an obstacle to developing taste. The friction was the curriculum.
I agree with your criticism of "the good enough" and "done with taste" dichotomy.
I would like to comment on this, though:
> that only want simple, that finding the most blunt approach is the only way. It has felt tasteless. Anti-ambitious.
I am always aiming for simple. If it's simple, it's maintainable and can be easy to reason about. Doing that is hard work, though. I used to spend three iterations to achieve it: functions are simple, easy to reason about, properly named and composable.
It's the opposite of blunt, though, so it might be that you meant something else with the term "simple" :)
This post is awful! Annoyingly written and zero insight. Most posts about this taste concept are bad (and there have been so many), but this is below average.
You've read many of them? Then you'd be the best to review even more of them! I'm ready for a raw and truthful review of my very own edition of the "taste" blog post, posted mere ~6 months ago, on the same topic: https://emsh.cat/en/good-taste/
Ha! I have a heterodox view on the taste thing, which is that it’s mostly cope: the models apply something like taste in nearly everything they do, and these taste posts will disappear as that becomes more obvious. However, besides the substantive disagreement, I think your post is fine since it appears to be your actual point of view, written down by hand. It’s not particularly novel or insightful (most of these takes are basically “there are still some things the models can’t do” — no shit!) but it’s measured and reasonable. Fine.
yeah the skill du jour keeps shifting. Everyone wants to say they have something the llm doesn't. The more subjective and ill-defined that "something" is the better.
I used to think the average HN reader probably did have good taste, at least above average. After all HN is a place for builders, making things. Most people never make anything. Making things means deciding what to make and how it should be, and also what not to make instead, so I think charitably HN probably does have higher average taste than random people or most other web forums.
These artsy fartsy takes on a subject that would be better studied as a science are fun but are they useful? I am not saying there is no such thing as taste and neither will I attempt to define taste except to say "tasteless" is probably the opposite. "Judgement" seems more valuable as some have mentioned, maybe judgement and taste together identify something we want?
Still, I like better when we actually do try to name what we're talking about. We've probably all read the motorcycle guys take on quality and things like that right? I guess the theme some are going with is that quality can't be objectively quantified and to define it is futile but I've never liked that take. If we think LLM generated code is low quality then it certainly makes sense to start our attempts to define or at least describe what makes the quality needle move, because language is how we encode lots of information and information is what feeds LLMs.
People also like to rag on uncle Bobs takes, clean code, etc but I like how at least he tried. I've been a fan of his for awhile but I'm not big on objects so don't care about his pushes on that sort of thing. The best take that offered hints that I would call quality is "A Philosophy of Software Design" by John Ousterhout, there is plenty of stuff in this short little book that was very satisfying to me, like for instance where Bob wants short/small functions, Mr Ousterhout was much more flexible on this, like for instance maybe some functions can be bigger if the need arises, things like that.
This brilliantly articulates what is required for best practices in the AI era. I am putting this article on my required reading list for Software Engineering. On behalf of my students, thank you for writing this!
> Everyone can make anything. Almost no one can tell you what is worth making. That was always the harder skill. It is now the only one left.
But people can and do tell you what is worth making. Taste is largely set by class and power. The “skill” is having the right context (wealth, education, network) and then pattern matching on it.
When resources are scarce, you are forced to use them in a meaningful way. You learn to do only what matters most: what is right, minimal, essential. When you need to choose wisely when resources are scarce -- this is how you learn the taste.
Now that creation is cheap, the scarcity has moved to adoption. Things built without taste are less likely to be adopted, to survive. At least that is what I'd like to believe.
Absolutely not. Scarcity forces you to do with what you have. Only in abundance, only when you can have everything, you have room and reason to develop taste.
Who had historically the most taste? The rich or the poor? And who had the worst taste? The people who grew up in poverty mindset that got rich. These are widely shared, common cultural observations.
You need a long term, persistent abundance to develop and even have a use for taste.
Is taste really the one thing LLMs can't take away or do we talk about taste because judging has become the cheapest move available? I think focusing too much on taste is how we lose focus on the two questions that actually matter: what do I actually care about, and why do I make things at all.
You cannot separate "taste" from the question of "why do I make things at all?" because they are inextricably bound. The reason for your making something is where taste is expressed.
I find this post disheartening. To my eye, it seems to be the author's first blog entry with obvious LLM contributions.
Given the recent "moving to a farm" ... 7 days later ... "leaving the farm" posts, it's hard not to read this as someone clearly struggling with the burnout which is all too common in our industry, and resultingly giving in to the false promise of a machine that can just handle the thinking for you.
From my interactions with him in some nix forums online, he is not in the industry. This is hobby work (not to devalue it, as his software is generally high quality).
> No, because the writing is objectively bad in ways that LLMs excel in producing.
Author does not dispute that. Yet they claim to not use AI. So your preferred belief is that they are lying, rather than a thing that looks AI written might have been actually written by a human and you are wrong thinking that it can't be true in this case.
> I do write like this sometimes. The short sentences, the reversals, the one-word lines—all of it. It’s just the way it is.
> That says something about my writing. It says nothing about who wrote it.
> So let me be plain about it: Claude was not here. No LLM wrote this—not a sentence of it, nor was it outlined, drafted, reviewed, checked, etc. by one
Either the author is doing this on purpose, or this is some kind of art project. I don’t buy the “just a coincidence” story at all.
Right, I feel like I'm taking crazy pills reading all the comments taking this article seriously.
I think AI has broken HN. The readership is so head-in-the-sand about AI that they wouldn't recognize 2000 words of Claudeslop if it genuinely slapped them in the face
The alternative explanation (the top commenters actually read it and somehow didn't recognize it's obviously written with AI) is just too bleak and I choose to not believe that.
It's LLM-like in the details, but there are certain patterns of repetition and idiosyncrasies, also certain analogies I don't think an LLM would make (like the aliens thing), that make me believe that it's written by a human.
I think whether "taste" was ever important comes down to whether software is a commodity to be bought and sold, or something more strategic.
The company that needs a webapp because everyone else has one is never going to value "taste".
To adapt to this AI world, we need to either ship more lower-quality software, or start seeking out higher value problems. I don't see any other way around it.
> I do not miss the wall, and I will not pretend to.
The wall is the point, if you want to truly know the truth of something. I still believe there is no royal road. Taste isn’t the answer to why you would choose a bump allocator over a slab when designing a system. It is experience, wisdom, and the long hard road of trying yourself and failing. Taste comes after.
There is much joy in suffering when the result is knowing.
Too many people have tried to declare victory over software engineering prematurely and it shows.
If taste that's all that's left you've kinda really got nothing. Taste is an anti-moat, it's sort of a measure of how well you, as a literal receptor, agree with a pile of people out there on something. Like sure its something you can sort of develop but it costs you nothing to use, and if you're not creating something off-manifold you're just smearing stuff around that's already on it. Horrendous kind of take on "taste" that's taken hold of everyone. Kinda like if a music artist stopped making music to make spotify playlists. Sure they could maybe make _good_ playlists, but if they are sitting there crowing "taste is all thats left!" it's.. what is that? It just sounds like either a deluded giving up or just a very direct admission of giving up.
Taste seems something that should be more related to a craft but when doing engineering we don't talk about taste, we talk about theory, systems, practices, skill, intuition, experimentation, teaching and learning. So the debate on whether taste matters or not might come down to whether creating software can be called craft or engineering. In my opinion, maybe programming was more of a craft before LLMs but it looks like it is being squished between the engineering aspects of producing and maintaining software going ahead.
It reminds me of I think Nietzsche, isn't it a quote of him? "In the end all that's left to us is taste"?
I can't seem to find it so quickly, so maybe my memory is failing me.
He definitely had a lot to say about taste.
I won't try to posit whether this post was AI-generated or not and I do not think it matters. That focus is a defense mechanism for handling an uncomfortable idea.
The idea is this: the commodification of intelligence requires the codification of domain expertise to function reliably, and proper governance to scale.
Taste, in my opinion, is a fill-in for absent function and lack of utility. Systems function; they don't require taste. Taste is why PG put Sam Altman where he did — not a system of merit, not evidence. Taste is why DJT can do no wrong in the eyes of his base. It is unprovable and has no true measure, and that is exactly why it gets leaned on so heavily: it lets you claim superiority no matter what slop your system produces.
You say the box moved taste up a level. Out of the making and into decision layer. That relocation is a smell. Your unfalsifiable thing conveniently gets moved to the one place it still can't be checked, at precisely the moment the making became checkable. Just like: Iran/Strait of Hormus, AI curing diseases and all of the other bold promises made by people who claim to be "taste makers" but in the end its all the same. Bullshit! Because thats what taste is a euphemism for in this context. Convincing others that you are correct in some way with no actual proof to back it up.
The vector here should not be "taste"; it's tolerances: written down, measured, enforced. The judgment gets encoded into a spec, and the spec is what survives the person who had it. That is codification; not "taste".
But toil is not gone. We have still have code owners, lawyers, physical hardware, thousands of configurations that we can't test them all.
And notably: we will still be asked to write software we can maintain in the future.
Sure, there are short-lived non-critical software projects where LLMs will rule.
But for many big complex projects, you can't land code without reviews. You can argue its because taste is already being enforced in the name of maintainability.
> But for many big complex projects, you can't land code without reviews.
Why do you think orgs are pushing less reviews? Token-maxxing increased PRs but stupid meat brains are a constraint. So of course, remove that constraint and you’ll have your features faster!
Software engineering is and has always involved engineering situations where you can let your co-workers/third-parties run wild with limited review oversight and limited blast radius.
Example: a plugin architecture with a solid API that makes degenerate behavior limited to usages of said plugin.
If we define taste as the intuitive act of saying "no, again," then I fully disagree with this whole article.
Every AI-frustrated (but LLM-written, sigh) blog post about the loss of taste and craft and hard work in development sounds like we've given up the interaction with the machine. Like human work is sitting back in your chair and shitting on stuff.
That's obviously not true! That's not how any of this works!
It's hard and weird to develop with the LLMs because they just do stuff. Lots of it is good, some of it is OK, some of it's horrible. Unpacking what it's done is hard and weird because software isn't just lovely UX, it's also data structures that scale and performance and privacy and enterprise controls and SOC 2 and onboarding and accessibility.
If you want to build real software, all that stuff has to get done. Today you're working on the feature, tomorrow you're making it scale. It's long-term and iterative and complex and hard to pack into a prompt or a markdown spec.
The work is the work, done at and with the computer, and it's way more than just "taste."
« Software works the same way. »
No, I don't think so.
Modern software has been around for about 50 years, the Internet & OOP for about 35 years.
People have been cooking steaks since antiquity!
If your opinion or understanding of how a program works or how it should work is your main asset for keeping your job, then I think you're the one who's going to get cooked.
You’d have to be rather disillusioned to wake up in the post-COVID era, with technology that triples its algorithmic efficiency every year, and conclude that a dehumanizing industry will still need our expertise in the future.
The key words here are “dehumanizing industry”.
You can always bake your bread the old-fashioned way or cook your steak in « your own small restaurant », no one will stop you. But small restaurants, at least where I live, aren’t doing very well.
On a side note, since the author has included a detailed explanation that the text is not AI generated and this is her default writing style, I just find it very difficult to read this style of extra punchy writing. I don't know if it is just me or anyone else as well.
I assume it's common and understandable to dislike LLM generated content and therefore anything that resembles it regardless of origin. For those who happen to write in a similar way it's a pity.
I've noticed some people on the internet dismiss content just by its length too now. I assume this is because they're not used to the idea of a human producing long form content (as opposed to the short form rewarded by popular feeds). I wrote a post (other forum, other handle) where I included some markdown tables and detailed explanations in an attempt to be thorough and help others answer my question and got accused of having LLM generated it despite writing it myself. Maybe the usage of markdown tables or the length of the post was the issue. Anyway, it's not nice being dismissed as AI slop when no AI was involved. I also don't see how there could ever be a reliable way of identifying origin of digital content such as text on a forum.
> I also don't see how there could ever be a reliable way of identifying origin of digital content such as text on a forum.
Perfectly reliable, no.
But:
1. these models aren't even trying to hide their nature, if anything the opposite
2. a great many users of those models are (lazy or time constrained, doesn't matter which) or lacking in style of their own, and in each of these cases the result is they don't prompt the models to hide their default style; worse, the outputs keeps getting RLHF-ed so despite how much of an "ugh" factor it has to the reader, most people making these things, to the extent that they do review the content at all rather than blindly uploading, actually like the style
3. Models don't share state between users. This means the model won't act "bored" from having already written the same thing before, so when 5000 people all prompt for "blog post about X", then it will create 5000 blog posts with almost identical in writing style. In turn, this means that the style of the most-used model (I assume for English-language blog posts this is still ChatGPT) has become cliché faster and harder than almost anything else.
I really don't get why people here are so obsessed with "taste". I also find the premise of the article to be completely wrong. Taste, not only in software, is only the only thing that's left for humans, if you ignore the complex social environments in which our technological systems are embedded. AI puts the emphasis in the socio-technological system on the social. If you were never social to begin with, or if you deliberately ignore the social embeddedness of our systems b/c you are terminally valley pilled, you might come under the impression though that indeed, taste is the only thing that's left.
That's interesting. I would (and I think most?) would say the opposite. Good taste is hard to define because it is very contextual. There's popular taste but also there's artisanal/specialized knowledge taste. One potentially leads to mass adoption, the other leads to higher value for a few people.
LLM-based code generation requires disciplined operation to lead to the best outcomes. Shops who only ever had taste as a proxy for discipline were missing this before, and they're missing it now. The fact that it's missing is why so many organizations are tolerating ever-growing mountains of slop. (Not new, by the way; just dramatically accelerated.)
Discipline can be linked to system design standards and your review process, and it can be translated to measurable outcomes at scale and over time, regardless of how agents are involved: unplanned work, time to recovery, employee retention, etc. When outcomes like these become your priority, you see that you can't rely on taste alone.
I tend to like images in a split second, when they try to achieve a certain aesthetic at least. I can later inspect them more carefully and understand why they're good.
Reading text, it can either grab me or it doesn't. I've asked about that to an AI and there's some reason for that. Some text is more compatible with me, due to the rhythm or combination of chars/words used, how they actually look/feel, it's not just about the content. So I'm assuming some writers are going to be more compatible with me while others will make me dismiss them quickly or just skim.
I like to think I have the correct objective taste in music, but everybody thinks that about themselves.
> later inspect them more carefully and understand why they're good
Reductionism is hard to apply to intuitions, and we often fool ourselves.
I personally call this the obviously test: smart people correctly notice something and say it is obvious. It is a rather rarer skill to understand and explain why. I have particularly noticed it with coaching (which adds the next levels of difficulty in getting someone to grok what you've seen).
People will always want something to gatekeep their chosen field. Code went away, so now people need to be able to say if the machine wrote good code. In a year or two people won't be able to use that because the code will be measurably better than something a typical dev would write, and in five years it'll be better than anything anyone could write.
Then we'll need to accept that coding is better when it's done by computers, and we'll start telling everyone it's actually the prompts and context that steered the machine that really matter.
But this is basically reinventing the term "Aesthetics" which is a field that had several centuries of philosophy research about it. Not all of them will agree AI cannot have it.
Actually I think it is possible to reconcile Kant's third critique with the statement "LLM is possible to make aesthetic judgement" (with certain generous application of principle of charity, naturally).
AI does have something like Aesthetics and taste, but it's the average of all the inputs and so it's almost by definition the answer to 'what is the most average design possible based on all human prior data inputs?'
That's why AI prototypes of front end UIs tend to look the same without massaging them with your own design tokens, design language, and injecting your own taste.
It didn't require great character at all
our refusal disagreement and resistance
we had a shred of necessary courage
but fundamentally it was a matter of taste
Yes taste
in which there are fibers of soul the cartilage of conscience
Who knows if we had been better and more attractively tempted
sent rose-skinned women thin as a wafer
or fantastic creatures from the paintings of Hieronymus Bosch
but what kind of hell was there at this time
a wet pit the murderers' alley the barrack
called a palace of justice
a home-brewed Mephisto in a Lenin jacket
sent Aurora's grandchildren out into the field
boys with potato faces
very ugly girls with red hands
Verily their rhetoric was made of cheap sacking
(Marcus Tullius kept turning in his grave)
chains of tautologies a couple of concepts like flails
the dialectics of slaughterers no distinctions in reasoning
syntax deprived of the beauty of the subjunctive
So aesthetics can be helpful in life
one should not neglect the study of beauty
Before we declare our consent we must carefully examine
the shape of the architecture the rhythm of the drums and pipes
official colours the despicable ritual of funerals
Our eyes and ears refused obedience
the princes of our senses proudly chose exile
It did not require great character at all
we had a shred of necessary courage
but fundamentally it was a matter of taste
Yes taste
that commands us to get out to make a wry face draw out a sneer
even if for this the precious capital of the body the head
must fall
In software development, good taste translates to readability, succinctness and maintainability of the code. It also translates to intuitiveness, reliability, security, performance, scalability and utility of the software.
All this doesn't necessarily translate to sales though. Because, to evaluate those things requires attention, trust, time and effort.
This is a significant barrier because a lot of software will appear to meet all of these superficially.
After 1 day of usage, a piece of software may appear to be intuitive, reliable, secure, performant and useful... But then after some time (sometimes a whole week or longer) you run into a critical scenario and discover that it cannot be solved with that software... Or performance drops off sharply after you created the 1000th record in the software... Or a hacker takes months to find that one endpoint which allows full remote code execution.
Even in an optimistic scenario, 1 week is a long time to evaluate a piece of software. I've encountered software which took 6 months and large teams of people to realize that it wasn't suitable. That's how long it took to hit the critical limits. It's a very long evaluation loop.
So you cannot judge new software efficiently by just looking at it. Even industry consensus is problematic if the software is very new and complex... Plenty of trends have fallen off a cliff in the past. You need to understand who is behind the software... And even that's not so easy; social proof can be misleading when it comes to deep technical ideas. The people who are good at social networking aren't necessarily good with tech.
I think AI slop code is going to be a much bigger problem than people anticipate. And the irony of it is that the solutions already exist... What doesn't exist is the mechanism to identify those solutions. But even the mindset needs to be corrected first.
> I think AI slop code is going to be a much bigger problem than people anticipate.
I'm seeing the same thing. Of course, it's always possible that AI capabilities grow significantly from here and this worry is moot (and there will be other things to worry about, like whether I'll have a job). But seriously, after a while AI generated codebases really fall apart.
It's a bit of cope from me that this problem won't be solved in the future, but I'm also really disappointed with how many people really do not care about writing high-quality software that is easy to understand, performant, and resilient to bugs. Like, why did you decide to work on software in the first place?
I think we're going to see some pretty careless mistakes in the future that will result in a lot of pain...
I'm feeling quite confident that AI won't be able to produce quality code in the foreseeable future. Because the bottleneck is not technology, it's people and incentives.
Firstly, people don't agree on what good code is (developers who get paid by the hour tend to favor complexity so they think good code is complex code), secondly, AI companies have an incentive to produce more tokens; this works against good architecture since good architecture would cost fewer tokens to maintain. Thirdly, many decision makers tend to conflate complexity with intelligence, hence they are more likely to prefer complex solutions over optimal ones.
> The market timed you both with the same stopwatch and it did not see the difference. It cannot see the difference. Taste does not show up in the diff.
Doesn't that mean guaranteed failure in the market if you use taste as your guiding star? Therefore, there's really no way to enter the market anymore.
I don't know how you then ship software or monetise it...? The market will just clone it.
So... Don't share stuff publicly? Only share with closed groups? Where does this leave small-time software development...?
Someone can replicate any high fashion piece of clothing you make by hand. Maybe they could even make it better. Buyers still care about the origin, just in a pretty small market for it.
Maybe software moves to this after it's commoditized? Users might prefer an artisanal one vs ordering it off the shelf or using an agent themselves.
I've been circling this insight for a bit previously, leading me to a convo with AI about how we are now approaching more of a spec - no, prompt - no, context - no, [insert newest paradigm] driven design approach
The point being more that the development does weigh more heavily on having fun ideas to ask AI to build
I think there's probably also a place for people to make guides for AI-assisted programming or even personal essays on their views of "taste" or what they want code to look like?
I see pop up frequently objections to vibe coding (different from mere AI assistance in coding - vibe coding being creating things without looking at the code at all) that it doesn't account for architectural descisions, cybersecurity decisions, edge cases, and other such things. So to me it looks like a next step might be in identifying these things "vibe coding" doesn't do well, that people can learn more manually, so they can do at least AI assisted coding well.
> I do write like this sometimes. The short sentences, the reversals, the one-word lines—all of it. It’s just the way it is. A LLM writes that way too, because it was trained on the same essays I grew up reading
I'm extremely irritated that the post is being accused of being AI slop. I have a visceral reaction to AI generated writing, and this post rang 100% human to me.
Makes me actually quite angry that one can no longer use short, pithy sentences without being accused of using AI. If AI causes everyone to reject actual decent writing because someone can't distinguish between it and that from a human, we have other problems on our hand.
The content of the post resonated with me 100% as well.
Same, the post has an entirely different speech rhythm than AI.
In Walter Murch's "In the Blink of an Eye" he describes how, when cutting a movie, the reason cuts work at all is because human experience has cuts: we blink.
More specifically, we blink when a thought is complete, to finish one thought and begin another.
When a movie is well-edited, the audience will adapt to the rhythm of the movie, and you know you've placed a cut at the right moment when the audience will blink at the exact moment of the cut.
The same "rhythm of thought" applies to written essays as well, and once you get a feeling for it, you'll notice how AI generated text follows no such rhythm, no such structure.
AI generated text has the rhythm of TV salespeople combined with tiktok brainrot, overwhelming you while at the same time saying nothing at all.
It's an interesting article. To me it reads like someone who is not confident in their writing, but is very interested in getting their idea across, iterated with an LLM to produce it.
It set off all my LLM writing triggers and I noped out after about 3 sentences, because I'm designing/coding with AI all day and I'm 100% sick of its writing. I don't really care if it was written by an LLM or a person who writes in LLM style.
HOWEVER, the surprising fix for me was to ask a number of LLMs to summarize it. I omitted the "is this AI slop?" parts when I pasted it in. Like I said, I don't care about who wrote it, but I hated the writing style.
The AI summaries were quite readable.
They converged at a high level on the summary, but with a good bit of variance in nuance. I had a good follow up discussion with one of the LLMs, so I'm glad I tried it.
I agree with the idea that the taste, for lack of a better word, was the only thing that ever mattered. What I disagree with is the idea that slop that is “good enough” is indistinguishable from something that is crafted is “indistinguishable by the market”. I think it is indistinguishable in the short term, but very distinguishable on a long enough time span. The same way that hacked together slop has always been “good enough” to close a ticket and even get someone a promotion, but later costs the company thousands of engineering hours to rewrite once it becomes completely unreliable, slow and impossible to modify. We’re just speed running that now. It’s the same battle engineers with “taste” have been fighting with sales teams and product managers and lazy teammates before LLMs, just in fast forward.
That could be true for novel problems from Deep Tech industries and not everybody can solve those.
If you are doing another app, the product experience and marketing will probably win. Many companies with better products lost, beaten to death example is Betamax
I think it's a pretty good direction to go down. I wonder if we can add a little more accounting to it. Cheap shoes solve an immediate problem but you pay more long term versus expensive well built ones. How much $ (in dev time + tokens) was spent in incidents and/or fixing code later. Credits and debits. Weigh money earned (somehow) against money spent.
The irony of OP championing for “taste” in the age of mass produced slop, only to then post something that was likely produced by repeatedly reprompting with “no, again”
I found this post a bit too verbose and incoherent. It does read like AI slop to me, but the bigger issue is that its core argument itself seems flawed.
The author seems to be advocating for choosing to "carry the burden" of one's taste (or aesthetics, judgement, standards, etc) while bemoaning that the market may not sufficiently reward for doing so.
IMHO, the core flaw is in mixing these two things 1) the market rewarding what it chooses to reward; and 2) individuals choosing to pursue some external/internal rewards.
I broadly agree with the premise that Gen AI has lowered the cost of software development and, as a result, the quality of the idea and its execution have become more salient differentiators than just the ability to bring an idea to reality. I also agree that pursuing one's own sense of taste/judgment or "objective function" is often intrinsically rewarding and worth it. But ultimately, an individual has to decide how much they want to (or need to) play for the external market rewards vs their own intrinsic rewards/motivations.
But I don't think it's coherent to argue that the market "ought to" reward one for their oh-so-special but mysterious, amorphous, and unmeasurable way of doing things, or that pursuing one's own taste/judgement is somehow automatically an exercise in greatness.
I think there's a coherent argument lurking in there, but the author hasn't refined it enough. Here's my attempt.
The industrial revolution seemed like the end of the artisan as an important man in his community, the death of good taste, and an apocalypse for professions like furniture makers.
And yet, do you remember the Aeron chair, darling of the dotcom era startup? The Eames chair, that wood veneer conversation piece which every real estate agent chooses when staging an apartment? In an era of mass production, taste moves up the market, but it never quite goes away.
Where does that leave our taste, dear reader? I am not sure. What I do know is that, just as the Aeron chair could not exist in a world without mass production, so too will there be tasteful software which cannot exist in a world without LLMs. I hope you're as excited as I am to build it.
I have no idea whether this was the actual intent, but it's where the disparate little facts seem like they might lead.
> In an era of mass production, taste moves up the market, but it never quite goes away.
I agree with the points in your version, and the part above is the “meat and potatoes” for me.
I too am unsure about how much did the author intend to say something similar.
The main area where I disagree with the original post is this apparent fusing of subjective/personal taste driven value and market rewards driven value.
To summarize/clarify my view using the Aeron and Eames chair examples from your version
1. I agree that someone had to pursue their taste/vision for those products even when they could’ve “gotten away” with producing something mass produced and mediocre. And that pursuing their taste/vision likely became more feasible once the technology allowed them to focus on the “taste” while taking care of the boilerplate.
2. But, I will also add that the market actually saw value in Aeron and Eames chairs and rewarded their creators. It would’ve been incongruent for those creators to assert that their chairs were superior merely because they chose to pursue their taste/vision.
I think the "taste" described here actually can be formulated as a measurable quantity, just one that can't be accurately measured in practice. According to James Shore's 2006 essay "Quality With a Name" [0], "good design" amounts to nothing more or less than minimising the expected engineering time needed to create a good enough product over the space of changes that are likely to be needed during its lifetime (I'm paraphrasing slightly). Realising this was a revelation to me. I would argue that OP's "taste" is Shore's "good design".
This conclusion may be unwelcome here, since it implies that AI can acquire "taste".
I don't think it's entirely unreasonable to conclude AI can acquire what I have described, and would continue to defend as taste. The real question, then, could be whether its acquired taste renders you obsolete? So what if you can be replaced, what then? A question that popped into my mind is whether taste, as immeasurable as a metric as it is, would be what keeps you relevant just enough to matter.
Unpleasant as it would be, I think it may well render us obsolete in the main. There will probably always be a handful of paid human programmers around, just as there will probably always be a handful of people paid to make buggy whips, but the history of capitalism shows that most buyers don't care enough about how their goods and services get made to pay the premium that would keep "handmade" items economically viable for sellers. So I don't yet see a compelling reason why ~all of us programmers (and all other white collar workers in their turn) won't be replaced by LLMs.
The only difference I see between what happened to master craftsmen during the Industrial Revolution and what is happening to programmers today is that I belong to the latter group, so for the first time, the coming employment crisis feels personal.
Unconscious intuition you can't explain in words, thus can't reason about, as the main guiding principle in the ever more cognitively demanding world, yeah, good luck.
If I ran a dev team, I’d spend a day a week on deliberately hand coding. Even if the clankers write most of the code, IMO we need deliberate practice to develop taste. There’s a side of taste that’s Rick Rubin, how the customer experiences something. There’s also what a classically trained cellist hears in a more tactile, layered sense in how another cellist plays a piece.
This was a great and resonant post. Don't worry about the green names.
I think there is almost a split between people who have read ZAOMM and people who have not.
I recently made a post in my circles about how "meaning" is what human quality is. If you look at classical art a lot of it could look objectively nonsense and have technical mistakes, but just because the machine can make a picture does not mean that it attributes any meaning or communicate any intention. The difference between human-made classical artwork and machine-made classical artwork is the meaning and symbology behind it
Before God and as an honest man I tell you that your son is the greatest composer known to me either in person or by name; he has taste, and, furthermore, the most profound knowledge of composition.
I'm really not convinced that the word "taste", as used in these type of AI discussions, refers to any meaningful concept. The AI that wrote this blog post (look at footnotes 2 and 3!), like many others, is great at defining what it is not and how people psychologically relate to it, but I've never seen a clear definition of what it is.
My best guess is that it's a pointer not to any actual concept but to latent semantic space between concepts. Either a very novel or very Buddhist phenomenon, depending on how you look at it.
Speaking of taste the author basically tells me not to read the article immediately as soon as I hit the page because the commenters on this site (HN) have made them feel so bad about it
I’m pretty sure it was not technology that killed the web it was the kind of toxic crap like that, that we’re seeing from communities and that’s nothing new
I was expecting them to state that they were doing a bit with the post-mortem, but the punchline never came. It is even more consistently AI coded than the full text. Give us a /s, give us a wink a nod, give us a markdown sygil of LOSS.png, anything that says "I am actually a human", because all that reads as is BEEP BOOP KILL ALL HUMANS.
Honestly, a lot of people are denying this, but the latest models ARE at the point where their systems design generally surpasses mine because they have a better grasp of how the domain of any problem is typically solved.
Ironically, this essay is an example of at least somewhat "questionable taste". It has provoked dialogue and thus is valuable, but I think it makes a number of cognitive errors in its argument.
"Taste" is just knowledge. Advanced knowledge. Knowledge that is not easily put into words.
This is another restatement that knowledge, in a human head, is the determining factor of production, over land, labor and capital. This a post-capitalist argument that's been made for over 30 years (Peter Drucker in his 1993 book is my favorite), and it's been at least somewhat true even back to the 70s, just increasingly so as AI takes off.
The author's sidebar of AI or even cloud computing "taking ownership of the means of production away from workers" is entirely misunderstanding how the modern economy operates and shows how badly we need a new Marx (and Keynes, Schumpeter, or Drucker) since his analysis doesn't seem to provide the same analytical value in late capitalism. Renting vs. owning the means of production is now entirely a function of knowledge of the entrepreneurs (and their team).
Capital is ubiquitous now and is desperate for return - it remains powerful but is not all-powerful. We've had years of near-low interest rates and even after COVID inflation, they're still historically low. It is far more accessible to entrepreneurs than it has ever been in history, between grants, loans, subsidies, private equity, angels, crowdfunding, donations, etc. It's practically a cliche that billions in capital can be wasted on an initiative because the people involved had no idea what they were doing. I don't think AI changes that.
We don't need to rent our tokens, we can run open models on our local rigs (which still requires capital). We don't need to rent our servers, we can buy, rack and stack them on our own. Which we choose is fentirely dependent on the knowledge we have. If we don't know how to run an AI rig, we must rent it. If we don't know how to run a server, we must rent it. etc.
There are several walls driven by taste
- Design
- Testing
- Production
- Security
- Scalability
In particular, the author seems confused on what "production" is. Production is still the wall. There was a joke on X the other day:
"I vibe coded my dream app! Try it out here: http://localhost:8080/coolapp"
Software running in production is not tokens. I'm sure properly prompted AI can deploy and run simple software in production but this requires a lot of knowledge to prompt to constrain it to run in a secure, scalable, adaptable way as complexity, scale, and capability grows.
It's also just patently false. Taste is not all that is left. Are software engineers really so full of hubris that they thought coding is all that there is to making a product?
> Are software engineers really so full of hubris that they thought coding is all that there is to making a product?
Frankly, this barely scratches the surface of the hubris of engineers. I personally think AI has been one of the best things to happen to software engineering. Writing boilerplate or my millionth auth implementation was never why I fell in love with the craft in the first place.
We have computers, computers can automate things. Boilerplate is a sign of a bad language (not a bad engineer). But then we add infinite automation features to reduce boilerplate and we get C++, and C++ programmers decry the excessive number of features and want to delete them all and go back to writing boilerplate by hand.
I once wrote a pile of C++ templates that would let you wrap a C++ function as a JavaScript function, automatically converting all the arguments and the return type. Then I left the company. How did my successors find it? No idea. Maybe they pulled out all the automation complexity and replaced it with manually operated boilerplate.
This pretty much, Emacs, Sublime, Vim, Jetbrain’s IDE,… have snippet tooling solutions that are very configurable. And most unix editors support piping so you can easily write a programs for that. Or just a good old code generator.
My goal when writing a software is always to reach a point where modifying a business rule is as easy as writing it in plain english. Which means abstracting the essential complexity (while being pragmatic).
This is true, but I was mostly referring to framework boilerplate (Spring, React, etc.) or plumbing boilerplate. If you've ever written code professionally, you know that most code that you write is just making your thing fit with someone else's thing.
If your consumer or your provider made bad engineering decisions, you have to absorb them. If your company mandates the use of a framework, you have to absorb that, and so on. So I think it's great that AI can write all the dumb shit I didn't come up with anyway.
> So let me be plain about it: Claude was not here. No LLM wrote this—not a sentence of it, nor was it outlined, drafted, reviewed, checked, etc. by one, and there is no prompt behind it either
Following form of the words I just read: these 100% animous filled HN threads about how everyone is bad and everything deserves mocker have to be some elaborate power game to undermine hope and human will, I genuinely do not understand how genuine human people could think it good to "comment" like this.
There's just such a ridiculous pile on of people with such enormous overwhelming negativity underneath this ocmment. This off-gassing is almost never engaged with the material. Now there's an excuse to write off any material we don't like instantly, to accuse it of fakeness. But I still do not see a realness enscribed upon people who so actively work to tear down & prevent thought & consideration.
To engage with the post: I do think the definition of fashion is interesting, and a real power dynamic to be aware of. But do think there's a lot of things that take off in the world because they are good and interesting and nice, especially in technology and products, that the world wants nice things, and will happily support you for doing that. And especially in tech, the technical decisions we start from and build underfoot steer us towards success in subtle compounding ways, that matter, and that taste is crucial, where fashion is not a factor.
Again, at a meta level: my distress & frustration at the top is less from seeing this comment, and far more from seeing everything down thread trying to out-do one another with how cynical and anti-meaning anti-trying they can be. But I think this post definitely started on an all too familiar note, blood in the HN waters that all too often gets us so sharky.
IMO giving "benefit of the doubt" to something beyond doubt is bad, straight up. If someone can't tell this article is AI then they need to go spend an hour a day reading known AI writing vs known human writing until they develop the skill; the voice of this author SCREAMS AI.
I'm a firm believer in not engaging with or promoting the content of misrepresented AI.
That's where you're wrong. It sounds close to AI, but it's very clearly not AI generated.
Let's look at
> So let me be plain about it: Claude was not here. No LLM wrote this—not a sentence of it, nor was it outlined, drafted, reviewed, checked, etc. by one, and there is no prompt behind it either
AI would've chosen rule of three (instead of 4 + etc), wouldn't have used the either at the end (doesn't fit into the rhythm), wouldn't have used "let me be plain" (more likely to use "let me be clear" or "let me be frank"), and wouldn't have chosen "Claude was not here".
If you've actually read a lot of essays and AI generated text (and sadly, I have), it's very obvious that this was neither generated nor rephrased by AI.
This is very much like the recent "kurzgesagt got banned by YouTube's AI detector" story, where the humans whose style was copied by AI now get accused of being AI
The first is an XKCD 605 (Extrapolating). The author makes bold predictions for the future as though they're inevitable, but gives zero evidence, because there is none, we're still at T=0. IBM is still in business. Wake me up when it's been taken out by someone vibe-designing a mainframe in kicad. Or even, let's see a dropbox clone get off the ground. Should be easy, right? (It's an HN tradition, after all.)
The second mistake is mistaking a large multiplier for infinity. The cost of code has not "collapsed to almost nothing" as the author claims. Talk to your CTO/CFO if you want the real story. They're probably freaking out over AI spend. But the point is, yes, AI has made things cheaper, but they didn't become monotonically zero.
AI had also raised the bar. You can't ship "good enough" anymore, it has to be great. And great still takes time. Time in QA, time obsessing over every workflow, time talking with customers and planning features.
Taste hasn't even entered the picture, and the analysis is already wrong.
In line with the infamous top comment during their launch, it does of course support FTP as a storage, and the other way around exposing your Dropbox as a FTP server, while supporting theme so you have a similar look and feel as well: https://www.filestash.app/img/screenshots/theme_dropbox.png + tons of stuff Dropbox never did
For those keeping track, this is now (at least) the third wholly or partially AI-generated article on the theme of "taste" as enduring source of human comparative advantage to hit the HN front page.
Previously:
Taste in the age of AI and LLMs - https://news.ycombinator.com/item?id=47677241
You Had No Taste Before AI - https://news.ycombinator.com/item?id=45288551
The author has updated the bottom of the post to clarify that it was not AI-generated. https://notashelf.dev/posts/taste-is-all-thats-left#post-mor...
I'm sceptical, where are all these essays he was supposed to have learned this very distinct style from? I certainly don't remember seeing them.
Pangram says "100% AI-generated". I trust Pangram more than I trust this author's word.
Saying "more" seems generous. I don't trust Pangram at all, and I also don't trust the author at all. Both can be trustless charletans simultaneously, and I feel this speaks volumes about the state of digital media today.
Pangram is a complete scam. There is no durable way to detect whether something was written with AI. I've had folks say they ran pangram on my writing and it was "100% AI". Except I don't use AI to write or edit anything I post publicly.
Paul Graham mentioned Pangram the other day.
I went to the site and pasted something into the text box. The next step was to sign up.
I don't trust Pangram.
So you would trust a known writers word over non deterministic software ? I read New Orleans will soon let AI handle 911 calls, Hope the AI believes the calls are from humans :)
Trusting snake oil is a wild choice.
Here's another AI detector I just built:
Holy shirt balls, they had AI in the 1800s! Fricken' Dickens was a bot, and Austen was a clanker.
Well, that is a shame, in my opinion. I trust humans significantly more, and I hope that never changes.
I get the market need for AI detection, but the fundamentals of it are so woo-woo that I despair
Are you serious? Pangram is notorios for false positives!
That's idiotic, frankly.
In this article, it's very obvious that a person put a lot of time into writing about their feelings; something that AI doesn't do.
Your statement is unfair and mean: The biggest flaw in the article is that the author is unable to be concise.
This article is not AI generated. Everyone is asserting it is but the author is well known as a critic of these tools in his communities. Why would he lie and use it to write an article with his genuine views, edit it to assuage the feelings of HN commenters, and then double down? Speedrunning losing his credibility, or, his writing style got on some people's nerves?
I showed some old Olga Tokarczuk essays to my friends to pick what was ai slop and what was not as experiment, pre- 2020 ones from Czuły Narrator to be exact. Guess what, they picked the ai one as human and her as most ai. Maybe because she uses a lot of big words and em dashes.
We are at the point, when there is too much cognitive energy needed to discern writing from human, human ai assisted and all ai. Should we know make mistakes on purpose to avoid being labelled as ai slop? :)
This is why I'm tired of the "this is AI slop" commentary. I'm well past trying to determine whether a human wrote something, and have fallen back to my old ways of critical reading to search for value in the thing I'm reading. And if I don't find it, it doesn't matter who wrote it. Everyone is so excited to be clever and call out others for using AI, when it's far more useful to cite and elevate good work, regardless of tooling.
AI community's obsession with the word taste has become somewhat distasteful for me.
Isn't aesthetic a better word for this?
Taste is the ability to perceieve an aesthetic.
On the topic of taste, I think of this quote in Susan Sontag's "Notes On 'Camp'" often:
“For taste governs every free — as opposed to rote — human response. Nothing is more decisive. There is taste in people, visual taste, taste in emotion — and there is taste in acts, taste in morality. Intelligence, as well, is really a kind of taste: taste in ideas. (One of the facts to be reckoned with is that taste tends to develop very unevenly. It's rare that the same person has good visual taste and good taste in people and taste in ideas.)
Taste has no system and no proofs. But there is something like a logic of taste: the consistent sensibility which underlies and gives rise to a certain taste. A sensibility is almost, but not quite, ineffable. Any sensibility which can be crammed into the mold of a system, or handled with the rough tools of proof, is no longer a sensibility at all. It has hardened into an idea..."
An important feature of taste is also that it is highly variable between people, and because there is no system or proof of good taste, we cannot really say that one person's taste is objectively better than another's. This is what causes one person to build a large software system in a classical OOP language while another builds almost the same system in a typed functional language, each following their own internal logic of what is good taste. In the end we learn something from it, and maybe someone who worked on both projects refine their taste and starts preferring something that lies in between the two paradigms and comes up with a new interesting programming language.
One thing I fear is going to happen when more and more people use LLM's for coding is that they will default to the "average taste" of the LLM, and we will stop getting new interesting ideas about how to build things.
> One thing I fear is going to happen when more and more people use LLM's for coding is that they will default to the "average taste" of the LLM, and we will stop getting new interesting ideas about how to build things.
I don't think it will have any impact at all on creativity in coding.
Creativity is rewarded in humans, just not by their employment where both training examples and monetary rewards can be found, rather the creativity is rewarded by challenges such as code golf, demoscene, and so on.
Go to a job interview and I have to know the shibboleths of Clean Code, SOLID, an arbitrary one of five frameworks, three different ways to handle asynchronous code, and four design patterns, and the magic word "agile". Invent something new at the workplace? If you're in a small team, or high enough in the hierarchy, fine. Big team? People will reject it with reference to those same shibboleths.
https://www.youtube.com/watch?v=1h0CHrQE794
there's a tension between innate preferences we’re born with, versus cultivated tastes that require effort
I wrestle with this also. A colleague pointed out that LLMs reduce language choice to an implementation detail… don’t have to know assembly to write python, and now you don’t have to know python to write a functional app. So why stick with python when rust might be better? Leave that to the LLM and let your engineering effort go toward designing the app instead of operating a programming language.
In which case yeah, projects are likely to converge on whatever pattern LLMs broadly fall into. But maybe that part doesn’t really matter. Not every project needs the engineer to care about that. If I’m not writing hardware level code, I don’t care what instructions my python compiles to, either.
Meanwhile nothing prevents someone from caring about these details if wanted or needed. Someone who knows and cares might absolutely do a better job than an LLM. Those people won’t necessarily go away.
But there -will- be a whole new category of software contributors who have ideas they don’t know how to express in a software language. Those ideas might well be very tasteful. So it could be argued that software development as a field will gain access to a whole new range of tastes.
I don’t know the answers but it all seems plausible
To be a bit more precise, SE is the practice to organize large codebase in a meaningful and human friendly way. AI will attract more hackers doing fun projects without careing about implementation details under a much higher level of abstraction, but they will not introduce more variation in SE.
Have you ever used enterprise software?
Seriously. Whatever slop is going to be an improvement for 99% of enterprise products. 99.9% of internal tools at large companies.
something something SAP something
Casey Muratori says they're both bad language choices as long as the software is slow.
>because there is no system or proof of good taste, we cannot really say that one person's taste is objectively better than another's
We can't say it in times of a general culture of confusion, like now. In times when there was an established cultural and artistic hierarchy, they could.
Was it objectively correct or was it just the king's taste?
literally: sumptuary laws restricting what you can wear, eat, or own
https://en.wikipedia.org/wiki/Sumptuary_law
not kings per se, maybe sultan, or the royal court, roman senate, etc.
The problem may not be that LLM taste is bad, but that it is incredibly easy to converge on
> One thing I fear is going to happen when more and more people use LLM's for coding is that they will default to the "average taste" of the LLM, and we will stop getting new interesting ideas about how to build things.
It is a possibility. But I also think we're underestimating the human need to create cool, unique things. Like yes most of the projects in this situation will have "average taste" but most of the projects out there today already do. Very few people have true taste and I don't think we're just going to see them disappear or be satisfied with LLMs feeding them "average taste." They will either likely stop using LLMs if that is the case or find ways around it to make it do things the way they want to, hence preserving their taste.
But I think it is a bit of an overblown fear that LLMs will just kill creativity and innovation in coding. Hackers will be hackin
>But I think it is a bit of an overblown fear that LLMs will just kill creativity and innovation in coding. Hackers will be hackin
Hacking and a taste for hacking is a muscle you develop. LLMs will make most people not train it at all. Same way a huge number of people are obese and out of shape. For LLMs to be bad news in this regard, it doesn't matter if some people will still find a way to do something creative. As long as far more people (than without LLMs) don't, that's already bad news.
It is exactly like that.
Yes, there are exercise nuts who are exercising like crazy, but the average population is getting weak and obese.
The same with mental muscles.
How many people can do mental calculations in their head nowadays?
What about remembering several phones numbers?
What about simply navigating with map but without GPS? (Or, heaven forbid, just by common sense and observation?)
Properly thinking about hard problems is exhausting. Majority will be glad to offload the hard thinking to an AI 'assistant'. The same way we offloaded computing to calculators, and memory and navigation to phones.
What remains is the easily manipulated lizard brain...
Heck, how many people in this room remember their own phone number?
I also don't think it will kill creativity in the space, although my phrasing might have suggested that, just that it might diminish it and make implementation styles more conformative.
On the other hand, we might trade that for creativity in other more high level designs where things like coding style is an unimportant detail for the experiment. For example, I have started playing around with microcontrollers because I can have the LLM generate the software while I focus on learning how to solder and think about how to modularize my hobby automation system. That's creativity and taste at a higher level which I didn't have access to before because I didn't want to invest my precious free time in learning the software ecosystem.
I feel two ways about it. I've written some algorithms and architectures for motion control that were good, after many years of chewing on the problem through iterations. I don't think I'll replicate that effort or those results, now that I use LLM's so much.
On the other hand, I did that in pursuit of a goal, and LLM's take me further, faster, toward that goal.
"Consistent sensibility" is a tasty little phrase.
I think taste is developed from a combination of sensual attunement (clarity of perception and focus) and experience (exposure to relevant, high quality 'stuff'), but the catalyst is joy, or at least some kind of pleasure.
This is an awesome thread. I agree - the catalyst is joy, in a broad sense of the word. One is lucky to find joy in something that can take you deep.
I'm a little confused. She follows describing people with good/bad taste immediately with the statement that there is no system for taste. If it is all relative, then how is she determining that someone has good or bad taste? Is there anything between moral relativism and mathematical proof? Just your own ineffable belief that you are right?
Even if something is relative, you can compare it to what other people believe. If most people who know person X think X has "good taste" (that is, that their tastes more or less agree, or at least that they aspire to having similar taste like how more people admire those who like classical music than listen to it themselves) then it is reasonable to say that person X has "good taste".
Someone with questions about this excerpt of Sontag on taste would do well to read the rest of the work, which does a fuller job of drawing distinctions between sensibilities like taste, and systematic ideas like good/bad, morality, relativism, mathematics, proofs, and wrong/right.
It looks like it's only 15 pages and I do like camp, so I think I may have a go at it
This seems to be circular.
Wouldn’t anyone need some reading “taste” in the first place to decide to read more.
And if they already have it, why would they need to read it?
If it’s all relative there can be no globally valid, or fixed, reference points by definition.
well yes, you can't help but have your own taste, and like the taste of others according to your taste of taste(s?).
And I think taste is subjective by definition, it doesn't compete with "objective taste" because that makes no sense.
> If it is all relative, then how is she determining that someone has good or bad taste? Is there anything between moral relativism and mathematical proof?
Social proof. Everyone has taste and everyone's taste in a given space is influenced and informed by the taste of others. It's mostly a consensus model. Newer members of a given community arise into said community and learn from it what "good" means, which can sound authoritarian but is really just... voluntary. If you like, for example, mens formal wear you will follow publications and noted figures in the mens formal wear space, you will learn what "good" means to them, and will emulate it out of a desire to fit in with them. As you get more knowledgeable in that space, you will develop your taste, enhanced by them, and ideally eventually contribute your own ideas to that mix. Some will be accepted, some will be rejected, again by group consensus.
It's a lot like language. If you want to speak a language you have to learn it first. If you go to France and simply blurt out French words at people, even with perfect annunciation, you will be incomprehensible to them. You learn structure, sentences, how to communicate ideas, how to inflect tone. How to tell the difference between telling someone to fuck off when they cut you off in traffic versus telling them to fuck off when they make a horrific pun. And, as people speak that language and use it daily, it will reliably change. The meanings of so many words in every language today are wildly different than they were even a hundred years ago.
i'd completely forgotten about notes on camp, had to read that in modernist criticism class. good stuff.
i see parallels to zen in the art of motorcycle maintenance as well, who's whole thesis if i can recall is something like, "quality cannot be put into words." Its a pre-rational thing that, try as you might, can't be pinned down.
This is an incredibly insightful comment, thank you. I had not read "Notes On 'Camp'" at the time of writing (and nor do much reading for the post in particular) but it seems very intriguing from what I can find on the internet.
Yeah once taste can be fully reduced to rules, it stops being taste and becomes something you can optimize for
I have mixed feelings... I stumbled upon, read a post-mortem that looks even more AI-generated than the article.
"That says something about my writing. It says nothing about who wrote it." - Two contradicting sentences, author's writing skills are part of the author as-a-whole
"The short sentences, the reversals, the one-word lines" and then "I will write the next one plainer.". - even more?
And... I wonder if this is like some agent wrote article, and, in the loop, added port-mortem, after visiting hacker-news comments and manipulate that this is human written. Or maybe author is just trolling us with AI-like prose...
> I'm sincerely sorry for publishing something that has allowed you to feel this way.
What fresh hell is this? If this is human then I wish it was AI
> something that has allowed you to feel this way
Sorry buddy, next time I won't ask for permission. I don't care what you allow.
> and this post will go down as my not-the-proudest, however, I take your criticism to heart—although not personally—and strive to improve.
I just do not understand why people - even when actively denying LLM usage - use the fucking em dash.
I don't care if it is the "right" thing to use there. The character has been associated with slop. Why do you keep using it without even just acknowledging that you did.
Just use any other character - like I am doing right now.
It must be AI indeed. Humans cannot be this dense.
___
Oh god I just saw the annotations on the side
> A fair pushback I got:
Dude. Why. Oh god. Why do you talk like claude. I want to believe you. I really do, but hell. Even claude could likely tell you that you sound like claude.
Oh my god.
> I just do not understand why people - even when actively denying LLM usage - use the fucking em dash.
Because they are literate? Unlike the "em-dash means AI" illiterate crowd?
I'm not going to change my writing because some idiots who never read a novel or heard about typography think only AI does that.
I just do not comprehend how that can be the hill people choose to die on.
Like.. I get it. It's wrong. It feels wrong to give the AI nonsense just an inch and all.
But in the grand scheme of things, it's just highly ineffective dogmatism that makes people unheard, miserable and just further increases struggle.
Which people are free to do, in the same way that I am free to purposefully drop a massive hammer on my foot just to make some kind of statement. Hurts me a lot, rest of the world stays entirely indifferent.
If you stop doing something right, because of fear, then the terrorists win.
I can see why you're saying that but
a) Using some other horizontal line is not in the class of "wrong" that would warrant this sort of reaction. You're debating the pixel count. Not the object. The aspect ratio of it.
b) If you're acting out of fear that's dumb, yes. If you however simply pick your battles for efficiency, that's not "the terrorists win". That's "you win" instead of "you loose due to attrition".
__
I've also consulted with the LLM regarding the "class of wrong" thing, because I thought it might be not quite clear.
So what I meant by that is that it's not giving up actual words or dumbing down your sentences. Just swapping out a character that is practically indistiguishable for the untrained eye.
Hell, whole classes of social engineering attacks rely on that very property.
I know I know. "Wehret den Anfängen". Slippery slope.
But those are insane as a default reaction to everything.
___
I get it. SV sucks. HN sucks. The world sucks. Can't let it also take this little thing. If it can take all the big things, then at least it won't take this tiny tidbit.
Emotionally, I get it.
But that's not how you make rational decisions. Nor is it how we would win.
Avoiding em-dashes is not going to save you from being blind-sided by slop.
Typography is not related to literacy. If you’re going to insist on em-dashes at least choose the right word.
I will continue to use em dashes as I always had, thank you very much.
I couldn't care less if you mistake me for an LLM. That's your problem for being overconfident in your misplaced judgement.
Th emdash is not something I use ever (before or after AI slop), but telling people not to use it because it’s associated with slop is not the way.
It’s like when people tried to tell everyone to stop using the ok symbol with their hands because a small group of hateful fucks pretended it meant something else. It only switches meaning if you let it, and stop using it yourself.
What happens when AI adopts something else? Do we give that up too?
With all of that said, this guy is definitely AI lol. He’s gotta be trolling us by doubling down.
I mean you would stop using the OK Hands thing when trying to communicate with a group that has for (possibly arbitrary) reasons decided that that symbol is bad, and thus anything you say is bad, right?
At least temporarily until you got synced up with that group?
There is a place for the strategy of just continuing to use something and push through the noise there, but the em dash is not in that place. Common perception of it has too much momentum you cannot overcome like that, which - at the time of writing - makes it a bad choice.
How it will look like in 12 months we shall see, but today, I would advise against using it (without acknowledgement) in texts that talk about AI.
No, that’s my whole point. I would not stop. If someone tries to attribute malice to something I do that has none, I might explain to them how misguided they are, but I’m not going to give in just because someone can’t be bothered to understand context.
I think it’s fine to use as one of many signals to identify AI (not the only) but, again, avoiding its use is not going to solve anything.
I'm being serious here: Looking for em-dashes is a terrible way to "scan" for slop. You need to look deeper than that and if you did, there's clearly a voice there. If you let yourself develop an "em-dash allergy" and think this helps you avoid LLM-slop, you're scanning for the wrong things and it's going to do the opposite for you. Actually stay on your toes and don't become this complacent.
sigh The em-dash is something we've read a lot from GOOD PUBLISHED WRITING so we as human authors tend to emulate what we read as good writing. It's a publishing artifact that's migrated. It's not a sign of the apocalypse. Neither is AI, but hey.
Em dashes are mine. I will use them whether they have been appropriated by LLMs or not. I used to use commas in their place. That resulted in a much more dense post layout and far less readable text.
FWIW I actually use three dashes (-) and my blog's build system transforms those into em dashes. This is not a defense, but nor is "this character is associated with slop" an argument. Yes LLMs use lots of words, phrases and patterns. Would you rather I stop using altogether? I am all for criticism, but please pick something I can actually change and improve without cutting a part of my style for no reason other than "Claude uses it too!"
> This is not a defense, but nor is "this character is associated with slop" an argument.
You're misunderstanding me. It's not an argument. I have no strong opinions about it.
I am telling you how people perceive it. You can (and should!) use that information however you like, but don't mistake this as some sort of opinion thing.
I've observed the world. I believe it operates like this. Feel free to discard if you have data that suggests it operates otherwise.
__
Don't mistake my delivery of that for personal outrage about you or that unicode character.
I just get very annoyed when I see systems not working optimally with failure modes that to me seem.. dumb.
If you allow me to quote the LLM:
> He isn't trying to police notashelf's writing. He doesn't care about notashelf's writing! He is just a guy who noticed a pothole in the road, pointed it out, and is now watching in pure fascination as people fiercely defend their God-given right to drive their cars straight into it.
Thanks, gemini.
Thank you for clarifying :)
Yeah, I am suprised this article ranks so high. It's either written by AI or it's poorly written. Very little to show on behalf of "taste".
"Taste" is one of these words that have been bandied about recently, in addition to "moat". I've never seen a productive conversation stem from either of those mentiosn.
People read the title not the article.
Had the same taste response, checked and Pangram also says AI. They have very low false positive rate.
yes this is 100% AI, the article and the 'post mortem'. People are so gullible. Anyone who has read a lot of AI can spot it instantly, but Pangram also says 99% of this is AI. So many signs, and no there were no human authors who wrote like this
"What it actually is"
"Here is the uncomfortable mechanism"
"Let me be plain about it"
"The floor"
"Everything downstream of the verdict"
I'll eat my hat if this isn't AI.
Very similar style to my AI 'utopia' https://aiking.dwyer.co.za/utopia
I don't even think all AI writing is bad, but lying about it definitely is.
Please, I couldn't care what Pangram (or any other "detector") says. If it feels like AI, does it even matter if LLM was co-author? It lands in the same "bad writing, I won't continue to read that" category.
That said I am recently getting paranoid about YT videos, I hear Claude traces in every 5th script. If people are going to spend substantial time recording and editing, is it worth "saving" time on most important part?
"it's not just ...., it's ..." is in EVERY youtube video I watch. There's got to be a reason AI uses it so much. Maybe it was already prevalent in bad scripts and we never noticed it before. Sort of like the "Webster dictionary defines ... as " cliche.
An antithesis is a completely valid stylistic device. It's just very easy to overuse or to use as a cliché when you're not using language as a tool for a specific purpose, but just want to sound smart or think "fancy stylistic devices = good writing".
Engagement farming has found its way to Hacker News.
I really don't know, their GH profile [1] and their projects, even the recent ones [2], seem pretty non-LLM. IMO the post-mortem is more a satirical response to the HN criticism, doubling down on the "AI slop" writing style.
[1] https://github.com/NotAShelf [2] https://github.com/NotAShelf/tuigreet
> IMO the post-mortem is more a satirical response to the HN criticism
They have edited it again to be slightly more emotional. It's why their "On Taste" closure doesn't actually pertain to Taste anymore. Of course, they wouldn't notice the mismatch - they didn't write it in the first place.
They aren't being satirical. They're pissed they got called them out. Were they not, they'd not bother with edits.
Ok but who's pissed off bere? A human or a let-loose agent? Or a human that wrote the post with an LLM although they swear they didn't?
It's neither satirical, nor by an agent on the loose/in a loop. I read the comment section with good intentions, assuming good faith---and in their defense there is plenty---but the amount of misattribution is counterproductive. Nothing about the post-mortem section is "pissed off". I wrote it in complete serenity, like I am writing this comment to explain. There are three groups of commenters:
1. Skeptic, but engaging in good faith - thank you, much appreciated
2. Skeptic, not truly engaging with the post - that's fine, can't win 'em all
3. Completely disengaged with anything in writing. They're under the impression that I am criticizing their taste, and felt the need to "strike back" - what?
The post-mortem is a response to all three groups. I've explained the situation, apologized for what I agree I did wrong, and clarified my boundary. It's not worth pointing time and time again it was not LLM authored, and I do not have anything to add.
Fine for me, I didn't ask any justification from you, but the post-mortem does/did sound as LLM-speech, and presuming it human I also presumed some kind of humor in it (I was probably projecting).
I'm a writer with what may be an unpopular opinion. I have a college degree in creative writing; I do technical writing and programming for a living, and I used to publish technical articles on Linux, Python, Elixir, and other topics—complete with em dashes—long before AI was a thing.
When confronting AI as a writer and deciding how I feel about it, I've landed on the conclusion that I've never accused mathematicians of churning out "calculator slop." A calculator is just a tool to make the complex work of mathematics easier. I now view AI as a calculator for writing, among other things. If the idea is good and the information is sound, I don't let the occasional awkward phrasing bother me. The truth is, most people can't write very well anyway, so if AI helps them communicate better—even if it is obviously generated by AI— more power to them.
My points of disagreement are
- At least to me, AI writing (especially from Claude) is actively painful to read. It's worse than the average person who has flawed English (or just sucks at writing) because of how verbose and stylistically repetitive it is.
- AI writing is a strong probabilistic signal that what I'm reading is going to be crap in terms of the actual ideas and content too, even if it isn't the case 100% of the time. I just don't want to waste my time on that kind of thing.
When I use it for writing I usually have an idea or thesis to explore already. I typically write it out stream of consciousness. The core idea is mine.
I'm not using LLMs to think for me, instead just edit the writing so it has proper grammar and the run ons are reduced.
Even better for something that has data. You can pull PDFs with encoded tables and extract the data into consistent units to compare. You can have it write code to do it and cross check the outputs. You can do that in a minute. Would take hours without an LLM.
Calculating is not math. Calculators make mathematicians better mathematicians by allowing them to spend their time on actual mathematics.
Writing is thinking. LLM writing makes authors worsr authors and even dumber people overall.
Also a calculator gives the same answer as a human (except for human errors where the calculator isa strict improvement). LLMs absolutely do not give the same output as humans.
There's a bit of a difference to me between a calculator and writing.. Creativity. Words aren't just a vessel for ideas. If there were different calculation structures etc and space for people to develop their own styles and approaches to calculate things, then sure.
A person's writing style is developed through what they read. And if human's writing style becomes more homogeneous through AI then that's sort of a shame.
People that don't communicate well can't fix that with AI. Turning one line of confusing thoughts into a 30 lines text make it worse, not better.
Somebody at $formerwork wrote a "skill" that takes a poorly/briefly specified Jira ticket, and "fleshes it out", purportedly so that when an agent picks up the ticket it has more context.
Really at that point, with head in hands, rather than try to explain the foolishness, I decided I was better out.
"Zoom and enhance" used to be a joke/meme, apparently it's now "engineering".
And mathematicians who don't understand mathematical theory can't use calculators to write proofs.
Focus on the content and the idea; not how it was written.
One line of idea spread into 30 is the worse way to convene an idea. Only people terrible at writing think using AI for prose is good. Most of the times, they also hate reading it, but love asking people to read their AI slop.
Hello friend, do we anything subjective from a calculator?
[edit]:
Someone already pointed this out.
Except calculators provide verifiably good results. AI slop sucks up and wastes everyone's time with nothing verifiably worthwhile.
I don’t like the term taste, but the problem that I have is that LLMs don’t seem to work “good enough”. They seem to be able to solve the immediate problem, but stacking this on the scale of 3-4 devs over 6 months or so doesn’t seem to produce anything.
One thing that I’m particularly frustrated with is the writing quality of LLMs. Like this is the thing that they should be able to do, but I would say almost everything they write has almost no signal.
Over a mid sized AI generated codebase that means I’m reading like 500 words to figure out what a module is even doing.
I’ve started requiring that unless absolutely necessary, all comments and docstrings be one-liners in our code 10k+ codebase(s). LLM-generated code has a tendency to “over-justify”, which is reasonable when first reviewing the code as a human.
But then, once it’s passed the first human it should be considered human-to-human communication rather than LLM-to-human communication, and the long, winding, and often repeated justifications can be significantly condensed.
Also you can’t let it run loose with unit tests unless you want to be blasted in the face with hundreds of lines of absurd test fixture preparation.
I'll risk sounding antagonistic and ask you this: if LLMs are not "good enough", why are they still around?
Obviously "good enough" was a poor pick of words of words on my behalf---and I'll gladly own it---but they should be "good enough" for something if a significant amount of resources keep getting allocated to them. Yes on a very personal level you look at LLM-generated code and think to yourself "wow this is garbage" but what about the people, as pointed out many times in this thread, that simply do not care? Does that not count as "good enough"?
The market logic isn’t going to be defensible. The market is highly irrational and prone to be influenced by those with the most capital, not the, “invisible hand.” Cars are freaking everywhere not because they are the best solution of transportation but because people with vested interest in the sale of oil and cars made it so by buying politicians, starting political and marketing campaigns, and on for generations. It took decades for the car to become the de-facto mode of transportation in North America let alone the rest of the world. It’s not good for us. It’s highly irrational to prefer cars when freaking wildfires and climate change threaten our very existence. In a similar vein, LLMs as they are, are not around because they are necessary and superior.
"The market chose it" doesn't mean it was the best solution, just that it fit the incentives
Cars are actually the best solution to transportation once you stop ignoring inconvenient but real requirements like autonomy and isolation.
Cars are not a system.
You need roads, traffic lights, drivers and pedestrians that respect rules, not too many holy cows in the way, not too much CO2 in your quasi-self-sustaining spaceship, lungs and livers that can deal with the soot, a tolerance for high-velocity impacts risks, not too many other cars around, ...
> Cars are actually the best solution to transportation
"the" is the error here.
There's places and people for whom cars are so. There's other cases where public transport covers the people while business-owned vans cover deliveries of bulky items, which is essentially how I've lived the last 8 years, though for "bulky" I should note I've seen someone take an actual kitchen skink on public transport here in Berlin.
I've also found some islands here Berlin with no bridge connection to the mainland and hence no practical use for a car at those locations; that said, my main reference for "car free" domiciles would be Sark: https://en.wikipedia.org/wiki/Sark
The current kind of car is extremely good at doing what it does. It's just that most of the benefits are short term, and for the car user (and the oil salesman), while the costs are borne by everyone else, and some of the worst ones are long term.
(Some similarities to hard drugs here.)
> Yes on a very personal level you look at LLM-generated code and think to yourself "wow this is garbage" but what about the people, as pointed out many times in this thread, that simply do not care? Does that not count as "good enough"?
Aside from the "amateur" qualifier, the Booch quote below provides insight to those questions:
"Code" is shorthand for "encode", which can be defined as "encoding a solution to a problem", which implies knowing what problem to solve, and thus ultimately implying understanding of same.So "good enough" depends on the stakeholders' definition of that valuation.
IOW, if a person tasked with delivering a solution can instead deliver garbage and still get paid for it, then it is "good enough" for some definition of same. If the people approving a person's remuneration are not satisfied with this type of work product, then the answer as to whether it is "good enough" or not will be quickly answered.
0 - https://en.wikiquote.org/wiki/Grady_Booch#Quotes
The use of “amateur” is indeed a very poor choice. An amateur is someone who seeks to master a skill or perform a craft for the love of the art, not for money. It’s the professional who is looking for the shortcut, not the amateur.
The amateur is the one who hones a tool’s edge until it’s razor sharp, and cuts away all unnecessary detail until they’ve found the essence of what they’re trying to achieve. The work of an amateur is often highly impractical and uneconomical, but sublimely beautiful.
I think you are torturing this comparison by using different meanings of amateur and professional
A professional does mean "someone who does something for money" but it also often just means "someone who is very good at something"
Similarly, amateur does mean "someone who does something out of passion without wanting/needing to be paid" but also often just means "someone who is just starting out, is new at something, and is probably not good enough to be a (paid) professional doing it"
You're talking about the "passionate" amateur and comparing it to the "getting paid" professional. I think many people talking about amateurs and professionals are mostly talking about "amateurs (newbs)" and "professionals (very experienced)"
You just twisted the meanings of the words.
An amateur is not a master nor a disciple of a master (e.g. future master).
The vast majority of amateurs have no ambitions nor passion and simply give up early on, that's a rose tinted view.
Since they don't get paid to do it, they have to allocate their precious free time toward the skill, which means they don't have those 40 hour weeks where they spend all their time honing their skill to begin with.
Even someone who does a skill half heartedly without passion nor ambition is going to learn faster with those 40 hours at his disposal.
>The amateur is the one who hones a tool’s edge until it’s razor sharp, and cuts away all unnecessary detail until they’ve found the essence of what they’re trying to achieve.
That's not an amateur and you know that. By your logic Japanese master craftsmen doing this are amateurs.
They're here, because even "bad" code has uses. We can have quantity over quality. We can generate throw-away programs (where the lack of maintainability doesn't matter). We can solve some problems by brute force (keep trying until tests pass). For people who can't program, even bad code can do more than no code.
LLMs are also useful as a natural language interface. They're pretty useful for search (as RAG and autoresearch, not the glue-on-pizza Google search crap). They're good for filtering and classification against rules that can be specified in natural language instead of special syntax or by training dedicated models.
> They're here, because even "bad" code has uses. We can have quantity over quality. We can generate throw-away programs (where the lack of maintainability doesn't matter). We can solve some problems by brute force (keep trying until tests pass). For people who can't program, even bad code can do more than no code.
If this is an acceptable definition of software engineering productivity, where "quantity over quality" is prized, then hire me.
Because I guarantee I can produce dozens of PRs daily, each having tens of thousands of LoC deltas (pick any language you desire), none of which having research and/or understanding underpinning them, and all easily quantified as being "brute force."
I'm not happy about the drop in quality either, and I'm not advocating for just making software shittier.
However, look at history of DSLR cameras vs mobile phone cameras. Mobile phones were always worse than large format cameras, and we've got a massive quantity-over-quality explosion in photos taken. It didn't replace professional photography, but let people upload photos of their lunch to social media.
I think in programming we're getting frustrated because we're trying to naively use high-volume low-quality code generators in our previous low-volume high-quality workflows. We're still discovering what's the programming equivalent of bathroom selfies, photos of receipts and QR codes.
There's a quote I forgot that says: "There is a Quality to Quantity".
I think it was from a Chinese person, as they took manufacturing of cheap things, low quality stuff, but managed to scale it to mass quantities.
I agree with that, it may not be reliable, full featured, well designed, ergonomic, etc. But quantity enables things and thus is a quality in itself. It allows use-cases that benefit from accessible cheap software. Those use-cases were not viable prior, because cost/time was prohibitive.
Often attributed to Stalin in reference to the mass production focus of the Soviet war machine. https://quotees.co.uk/war/joseph-stalin-quantity-has-a-quali... When you treat scale as a design criteria, you get a different product
> Because I guarantee I can produce dozens of PRs daily, each having tens of thousands of LoC deltas (pick any language you desire), none of which having research and/or understanding underpinning them, and all easily quantified as being "brute force."
I can guarantee that you cannot produce such PRs in the same quantity/quality than a modern LLM. The speed at which today's LLMs output code is superhuman.
It's textbook historical revisionism. It's not like we didn't have mountains of bad code in the pre-LLM era. I'm old enough to remember endless complaint posts about having to wotk with legacy codebases full of horrible spaghetti code. Somebody took a long time to write it by hand.
Have you realized you are attacking a straw-man? Nobody said all “hand written” code is good. I _think_ the implicit argument is: “all vibed code lacks cohesion”.
> I'll risk sounding antagonistic and ask you this: if LLMs are not "good enough", why are they still around?
To quote the parent commenter:
> They seem to be able to solve the immediate problem,
but not long term
Breaking down long term problems into small shorter problems we can solve one after another until the long term problem is solved is the essence of engineering.
No, the "essence" of engineering is having a cohesive vision for a larger plan.
No quality large scale project can exist without it. And any seasoned engineer should understand by now that no SOTA LLM can produce quality engineering at scale.
Lol. That's not true at all. And not how projects are done on startups or big companies.
Execute fast or get nowhere, a series of small wins get you to live long term.
Long term projects that survived on short term wins are still around. But you're going to want to check how big the graveyard is of "project collapsed under tech debt from short term decisions" to see whether it's a good strategy.
Nah. You can succeed in every single individual thing but fail at the overall project because the parts don't align into anything that makes sense.
If you need examples, look at game dev. There's plenty of games that have good execution but aren't fun to play or they're a confusing mess because there wasn't good overall direction.
Videogames are a weird comparison because they are basically the ultimate culmination of the combination of both engineering and art. The two can crossover in weird ways but still have distinct aspects that do not really compromise the other.
Just because a videogame isn’t fun doesn’t mean there wasnt a cohesive engineering vision. God of War: Ragnarok is a good example of this. Excellent technical execution and (arguably) great direction but a terribly boring game.
And many games that are terribly engineered are also incredibly fun. Dark Souls is a good example - runs like ass, looks very rough even by the standards of the time when the game released, with terrible enemy AI in a combat focused game - one might even call it a confusing mess (and I laugh at the notion that it doesn’t have good direction)… despite all this it’s a classic and arguably the most influential game since its release.
Elden Ring carries a lot of legacy baggage from years of iteration on the same engine as ~~Dark~~ Demons Souls - and it’s widely considered From Software’s magnum opus and a masterpiece. Yet considering the scope of the game it is a great example of how small wins added up can solve large problems when executed with the skill and judgment gained from experience. So one might still consider it a success in terms of “engineering”.
Would From Software have been able to make the same game using Unreal Engine 5 built from scratch with dime a dozen “Unreal experts” brought in off the street? Not a chance. Want to know why western game development studios are suffering right now? Compare American layoff culture and race to the bottom economics to the retention rates of Japanese game studios like Capcom, Nintendo and other heavy hitters.
For many of the most marvellous things humans have ever built, if you take a peak behind the curtain you’ll still probably find some amount of duct tape and popsicles sticks somewhere keeping it all together.
Fun, however, is a matter of taste and you can’t engineer that.
But that means they're good enough for the short them. This is why I find it such an interesting question to wield, but also, it is why my wording was poor in the post.
Short term can, IME, be very short. I've seen people generate, say, a bash script with an LLM. It's generated: short term, the problem is "solved": we've generated a bash script.
… but does it work? Someone comes along, reviews it, "this is garbage, and does not do what it says it purports to do". Perhaps it even gave an output: the script computed … something, but it's just GIGO.
But that "check if this works" friction is the same friction that is what people try to avoid by generating it with an LLM in the first place. If you're too lazy to write the script, you're practically by definition too lazy to verify it.
If this is your working environment, it sounds like quite an unusual place.
I literally can't imagine generating a script with an LLM without testing it at all.
Bash is one of those situations where LLMs can do really well. No human on Earth can remember all of the commands and even fewer humans can remember all of the switches for all of the commands.
This kind of remembering, searching, and assembling is exactly what LLMs are good at - as long as you're not writing a gigantic build system with hundreds of moving parts, in which case you should probably be using something more streamlined anyway.
> I literally can't imagine generating a script with an LLM without testing it at all.
Then you're extremely unimaginative as well as unusually fastidious.
Certainly someone - several someones - are generating lots of scripts and not testing them, given the PRs I'm seeing.
You can have another agent write the tests and verify the former agent ? This is pretty basic stuff. Makes me question if people are actually trying to use AI
Now you have two problems lol, in that you don't know if the tests are any good or actually test the thing in question.
Sooner or later you run out of turtles to put on the stack.
That’s a solved problem, you just add another agent to check if the tests are any good, and one more to oversee the test-checker, and one more…
Because if you criticize the LLMs, you get headbutted into the Dante's Inferno that is the current job market
They're "good enough", but they're not good. If you're trying to run a business, programmer skill isn't worth the money you spend, LLMs are cheaper. If you care about the quality of what you produce, they're trash.
Programming is no longer skilled labor. Quality is a hobby.
If your business involves storing any kind of customer data or handling payments, you probably should still hire a programmer who at least reviews the results of your vibe coding. Getting hacked is bad for business.
AI is just fine at security audits these days. It's better at finding bugs than it is at writing code, overall. You need to prompt for it, but that's not skilled labor.
I'm also not sure getting hacked is so bad for business. Most businesses that got hacked seem to still be around just fine.
I'd agree with the OP you are responding, what he means is that this isn't just about taste. It's that the code, while it works, and the text, while it has the ideas in it, is still not to the level of a human author. It's not that it isn't tasteful, in the sense of having the flair of specific author X, but it is verbose, goes around the point to get to it, etc.
Taste seems to imply an art, but I don't think we're at the point of cleverness, of LLM innovating a new beautiful programming paradigm or pattern out of immaculate taste for code. We're just talking more basic, even bad devs have more terseness, more clear intent in their prose and code, etc.
It's "good enough" for many use-cases, but it could be better, even before expecting the most beautifully clear code from the best devs out there, it has not reached the level of the average one in readability and clarity of expression (both code and prose). That annoys the author, he has to intervene manually to trim all the fat, and therefore not "good enough".
The market desperately wants to replace wages and the idea of a machine doing the work you previously needed to do before is so enticing, and the initial results so promising, that the rest doesn't matter.
Look at how many AI lay offs there already were a year or two ago - back when almost everyone would agree they weren't good enough. If good enough didn't matter then does it matter how when AI is better?
And despite what you might be trying to prove, it's obvious I'm replying to an LLM (a green account at that), but it's not for you, I don't expect you to bother reading it.
It’s a good question and I don’t doubt they’re useful.
But I mean they’re not good enough to do the things I need them to. Not enough that any difference can be chalked up to “taste”.
Put it this way, if you hired a very good handyman to build you a cathedral, you wouldn’t stand looking at the smouldering wreckage of the construction site saying “oh well that’s just a matter of taste”.
One part of the answer is marketing it to people looking for an easy way out without having to do the work or thinking at all. There's value in that as shown in everything like things for the home that purport to make your daily life easier. LLMs market themselves in much the same way
Not disagreeing with you but the "resource allocation" which I presume to hint at the fallacy that capitalism is somehow great or optiomal at allocating resources is not really true. Just look at any big company such as meta and see how much inefficiency and bad choices there is.
Stock market as a measuring stick is completely irrational.
Facebook/Meta is a great example indeed when one remembers that they took part in a genocide, and still haven't faced consequences for it.
There's the hope that they'll make all employees redundant and that businesses that adopted AI quickly will be able to exist as a pure money-printing entity without the liabilities of products, customers, or employees.
Put another way, LLMs are just another part of enshittification.
> if LLMs are not "good enough", why are they still around
They are not good enough yet, the people using these tools every day, and with a high bar for quality and design 'taste' notice the rough edges.
They are still around because LLMs are plenty good enough for casual consumers, but where 'taste' really starts to matter in design or in writing, the rough edges show up much more obviously and users who have been working with these tools feel it immediately.
Conversely, if they are good enough, why do they degrade frome consuming their own output?
How come humans consuming their own output gets better over time, but llms and any other form of ai consuming their own output only get worse?
This is just a problem of articulating what you want. Most people can't write for shit, I'm sorry. Because they maybe took one writing class in college. It's a skill in it's own. LLMs write better than anyone I've ever worked with and the reasons are obvious. If it rambles or is too "purple", that's on the person prompting it, because they can't write for shit to begin with.
> LLMs write better than anyone I've ever worked with
If you're talking about prose, this is obviously false. All the LLM prose I see is so mentally fatiguing to read that I usually just give up.
There are no lack of examples of poorly written LLM prose.
Then why is it that ALL content generated by LLMs, no matter who prompted it, is of such low quality?
I'd much rather see badly written text that someone invested time into instead of perfectly written and infinitely soulless AI slop
> LLMs write better than anyone I've ever worked with
And yet every time I see AI-written documentation, there's paragraphs worth of throat clearing and space-wasters like the word "genuinely" for things that could have been outlined in a couple of bullet points.
Since they can summarize, you'd really think they would be better at condensing their own output and cutting out the filler after. I wonder if you could use a specialized second pass for it or something?
In my recent experience it seems to be a contextual issue. Llms are constantly being dropped into new situations where they have to rediscover high level and cross cutting information about the system they're working on from contextual clues.
The internal representations of this state and its projection back out to human language wouldn't be as concise as that of a practitioner or team that develop their own verbiage and ontology over time molded to their system.
This verbosity might get better as we figure out better ways for agents to learn long term and use that knowledge to adapt to the users and projects over time.
There might also be some good harness improvements we could consider like forked output streams or multiple long lived filter subagents to ensure that output appropriate for thinking is separate from code output and separate from output given to the user driving the session.
Prompting is a skill.
I think there are really two types of use cases for LLMs here, and people that get value out of them.
First are those that care very little about the quality of the work, or the process behind it, and just want a 'finished' product no matter what. These are many of the folks vibe-coding everything without even looking at the output, and a depressingly number of those getting hacked or what not.
But hey, if you've got an idea for a complex app or website but no interest in actually building it, an LLM will get you... something vaguely like what you wanted.
Second are small projects and small sections of existing ones, most of which easily fit inside the LLM's context window. If you're making a fairly basic WordPress plugin or React component or one page website, an LLM will probably be able to handle it just fine. Heck, it might not even look all that different from what a human might have created code wise.
One of the big issues we have those is that plenty of people and companies are using these tools despite having requirements that aren't met by them in the slightest. If you're working on a Google/Microsoft/Meta scale product where performance and security and code quality are of the utmost importance, then an LLM isn't going to be a great fit. Unfortunately, that's exactly where this technology is being used, and the results are becoming more and more obvious.
I like judgement. Taste is a subset.
For me taste is judgement about things that don't matter. Like you have taste in clothes. But a judge doesn't have bad taste when he makes a wrong decision, he has bad judgement.
"When you don't create things, you become defined by your tastes rather than ability. Your tastes only narrow and exclude people. So create."
- _why the lucky stiff
We've been talking about taste for a year now as it relates to AI, so I'm going to counterpoint it: if competitors can reproduce your features, UX, and visual decisions within days, your taste is not really an advantage. How many apps implemented pull down to refresh once one person did it? How many people even remember (or care) whose taste spawned that UX pattern? [1]
AI is not inaugurating an age of taste. It is shortening the half-life of taste and commoditizing software. What happens when software features, UX, and visual fidelity aren't competitive advantages anymore? If you ask me, it looks like business did in the 90s. We'll probably see the return of department store software in the next few years.
1: for the record https://en.wikipedia.org/wiki/Pull-to-refresh
Also Microsoft couldn’t match Apple taste with effectively all resources in the world. And they really tried.
True taste is interesting because is more like a worldview, a modus operandi, a continuous process… those are harder to replicate.
Microsoft was superior in both in the windows 95-2000 era, and with Metro (on phone, on desktop it was half-assed). The problem was when they started to try matching Apple, instead of staying true to the core of their own aesthetics.
Both the Fisher-Price XP style which was bot zeitgeist, and an attempt to answer for the skeumorphic tragedy apple was pursuing, and the recent floats, spacings, central taskbar mimichking the dock, etc...
I'm not really convinced Metro was better than anything Apple at the time, even though it was refreshing and interesting. Was it actually easier to use? I didn't really find that nor see it with others.
I used Windows Phone, and it was great, minimal, intuitive, and prodcutive for most usecases I had (email, texting, calls, maps/navigation, banking, listening to music, light productivity, taxi). It did not fit the gimmicky ad ridden hellscape we live in now, I'm not sure how it got so far in Microsoft. Possibly from some old memories of living of selling products instead of creating addictions and selling data of the addicts.
Version 7 and 8 were great UX-wise, I have mixed memories about 10, mostly performance regressions bugged me, but was still faster than most non-flagship androids of the time... yet sometimes it felt sluggish to WP7 on potato hardware. The typography based UI was both fast and very readable and focused for well thought out apps. The standardized navigation buttons (on android button order may vary be vendor/model), and the clear gesture based navigation instead of hidden intearctions everywhere (long press, 3d press, app switching gestures etc), but using gestures for in-app navigaton were all easy to understand and grasp even by the elderly.
It was readable, well organized, nobullshit, but that is not what the masses crave. Metro was like how a "dumb smartphone" should be, in my opinion.
On desktop ot didn't work well. The live tiles were still a good idea, and if the taskbar supported them, it might have been a good idea (or not, was not tried, but could have be a bit like the old NeXT style refined, not stagnated and skeumorphized like on the mac).
That's because Microsoft understood taste as aesthetic - "we can put it in a pretty box", rather than ethic - fundamental to the process of making the product.
Most of their endless UI/UX fails are organisational in nature, no amount of design can create beauty from an org structure which generated stupid requirements.
And nobody there cared enough or had enough power to force the org into a shape compatible with Apple-level design outcomes.
Microsoft's taste was "good enough". They lost because of an app ecosystem, and because of Apple's brand power, the same way beautiful Chinese shoes will lose to Nike.
Nokia was able to match and surpass Apple in taste on all fronts with their N9 phone and MeeGo OS. So it is possible, but how to do it I couldn't tell, since all large organizations are saturated in corruption.
Funny you mention Apple because they are proof that taste can be king. Yeah they also kidnap their users into their ecosystem and make smart business decisions, but what really differentiates them is taste. And Microsoft is sort of the counterpoint. There's more than one way to succeed I guess.
Apple's taste during the 1990's almost killed the company.
It's not that the machines were ugly or badly designed (albeit not as good as the 80s heyday), but they were unable to compete on price/performance and OS tech stack, and that in turn was making them uncompetitive in application support, which would likely have killed them.
It's an interesting alt history - who would have won with no Apple around? I reckon we'd all own a lot more Sony and Nokia gear, it's possible the Cambrian explosion of weird featurephone form factors would have carried on a lot longer, perhaps the black rectangle form factor would never have won out.
Because let's admit it, later it was a taste of one guy who got another "chance" or the company got a "chance" to be lead by him. Now as he is gone the company still carries that taste legacy with better or worse results.
Or, used to be proof.
Apple the company that was rescued from bankruptcy 30 years ago due to a miracle, yeah taste gets nowhere without a big bank account.
Lets see if they will mostly sell iPhones to the end of their time now.
I think the point is without taste you wouldn't even know what things should be copied out of the countless competitors. How would you even know in the first place if a competitor did a particular thing tastefully without a good sense of taste yourself?
Also, there are a lot of areas where taste will absolutely reign supreme (and has been), especially anything related to media or entertainment (books, articles, videos, games, music, etc.)
I can see a possible future where taste is automated. If you can effortlessly make 1000s of iterations then you can test it on real people 1000s of times. The ones that do better you keep.
Another possible future is that taste is not automated but its not part of the designer's job (if designers still exist). Instead some AI will generate 1000 of iterations and people will be employed to evaluate 100s of software UI each day for taste. These will be taste farms if you like. These people full time job will be saying yes or no.
Other possible futures also exist.
That’s interesting there was creativity experiment results proving that could be true.
https://austinkleon.com/2020/12/10/quantity-leads-to-quality...
Loren Brichter [1]
Never thought of finding out who until you mentioned! What a small but fantastic contribution! Thanks Loren!!
[1] https://en.wikipedia.org/wiki/Pull-to-refresh
PS: Think about the next evolution of it, the auto refresh (from the great apps Instagram and LinkedIn).
Did I use the word evolution right?
Id argue that auto refresh is more a successor than an evolution!
I found this post yesterday on this exact same topic. As someone who is interested in the space of creating visual things and its design choices, it's a bit disheartening to see the current situation, and brings you to the point of "why even try doing anything anymore" rather than just follow the copywagon.
I feel this will be a similar effect to the current slowdown of hiring of juniors: you're killing the pipeline of future seniors that lead development.
https://xcancel.com/shadcn/status/2085163855408214356
Couldn’t disagree more. AI is a button that you press and magically get results. The ability to judge the quality of those results is “taste”.
Anyone can push the button, but not everyone can discern whether the result is a good one or not. That skill is either learned or to some extent inherent in one’s abilities.
My example I’ve been thinking about is: what will be the first new, big programming language in the age of LLMs? Theoretically, anyone could go out and make it right now. The skills and expertise required are well-encoded within all of the frontier models. The thing is, without taste, the language devolves to a useless state rather quickly. On smaller scales, I’ve seen this with API design and even object oriented design. I’ve also seen it in AI generated images and music: devoid of taste, and numbingly facile and generic.
Edit: not to toot my own horn, but “numbingly facile and generic” funny enough is probably a phrase that an LLM would never create without significant prompting to refine writing style.
That's not really what I think of when I think of taste in software development. I think of taste more in the sense of how to go about architecting something, what problems are worth solving, what technologies should be used, how to structure the code and documentation, how the APIs should function, what the scope of the thing should be, knowing what to do and what not to do, yada yada. That's all taste and LLMs suck at it (noting that taste is very subjective).
Agreed, considering "taste" to only mean UX or aesthetics is taking things too literally in my opinion.
"What happens when software features, UX, and visual fidelity aren't competitive advantages anymore?"
This is the question I've been thinking about. All can be copied very easily and in scale these days.
Copying means you have a valid target.
I'd argue that a well thought UX isn't easy to copy in the first place because it's not something that can easily be put into word.
It's the same for a lot of things, I have some clothes that I love because of their, design, quality and more ... something intangible that I have hard time describing and that can easily be explained with: It was done with taste.
Ironically given how much hate I have seen for them, this is what patents are for.
Invent something novel, get patent, you have exclusive economic rights to that for however long the patent lasts.
It was always easy to copy software (but yes, even easier now). The hard part is competing with the software that you're copying. Software does not exist in a vacuum. They usually have an already established user base, relationships with partner businesses, more domain knowledge due to having to actually think about the problems they are solving and so on.
You can start by copying but at some point you'll have to differentiate, and you might not be able to do so if you don't understand how your sloppy vibe-copied software works.
It's not about being the only person doing it, it's about choosing to include it. Many don't know what works and what doesn't. Being able to discern a set of features that are cohesive to a smooth experience is taste. Innovating and establishing oneself as the first person to deliver those experiences is a moat in itself.
Competitors can reproduce your discernment as well, it's the sum of your product and can be equally cloned. So how is this a moat?
Because reproduction is reactive. If an organisation only clones others' style then they'll always be behind the curve and will be widely recognised to be derivative and lacking their own 'taste'.
Look how long it takes Chinese brands to transition from being clones to establishing their own corporate image.
Dumbest idea ever. You are composing some text in a form, or filling in fields. You want to scroll up just a little to see something out of the viewport. Oops, your little swipe is misinterpreted as a refresh gesture: the input is gone.
Yeah, it's infuriating when poorly implemented.
I think it's impossible for it not to be poorly implemented because the gesture is inherently ambiguous (or becomes so when you overload it).
I think what "poorly implemented" actually means for touch screen gestures is to be lacking the customizability to reassign or disable infuriating behaviors.
Gesture configurability is nonnegotiable: when implementing gestures, have the configuration system in place first and make sure every single behavior is tied to a preference.
Imagine if artists were embracing AI the same way coders were being forced to. All that would be left there is taste, and your argument applies as readily there as well.
This article resonated tremendously with me, speaking as someone who has been coding since the 1980s and had to develop my own taste the hard way, through lots and lots of mistakes. Whenever I see a software demo built by someone mostly using agents to do the work, I have to question whether it's any good on the inside - or whether any real intuition or judgment went into its design. On the flip side, maybe none of that matters anymore - as long as it works, who cares how it was built.
For majority of people, including business executives it's the latter. No one cares about artisan code. Software is just a means to an end for them and not an art form.
People treat their computers like any other appliance. They don't care if it's riddled with bugs, they'll gladly restart their computer or refresh the page at infinitum to get it to work.
I think it's also why marketing and designers get more of a say in the software than engineers. UI polish matters more since it increases sales.
This is the hardest thing to come into terms with in this industry.
> they'll gladly restart their computer or refresh the page at infinitum
They won't gladly do it. They'll do it begrudgingly because there is nothing better, or they have no choice, and they need to get their work done.
Just because people use your AI-vomited garbage programs doesn't mean they like it.
They chose to pay for it, which is about the same as liking it, as far as business needs go.
I'm quite sad that I have to disagree with you and agree with GP: if quality siftware mattered to people, Microsoft would have gone out of business a long time ago.
If quality software mattered, javascript wouldn't be the most popular way to write desktop software.
If users cared at all about quality software, we wouldn't have 800MB chat apps.
Would this apply to software that runs oil refinery plants, or video games or an fighter jet flight stabilization systems?
Surely there is more to software than CRUD apps
Of course yes, and it's even easier in some cases. Since there's a language that defines the functionality (e.g: formal methods) that the agent can use/veriy against
Time to load is perhaps one of the key metrics software is measured by because time to load directly relates to user action.
Slower load times means user won't use the product, won't click the buy button, or will simply leave.
Marketing and designers do not get more say in software. They get more say on the "surface", which we all know is 10% of the software.
Yes but that 10% is all that matter to the people that matter to the business. Like a car, most people care about how it looks and how they interact with it... the only people who care what is under the hood, for the most part, are other engineers.
> .. why marketing and designers get more of a say in the software than engineers.
I hate it when visual designer do a product design or ux design job.
I hate it when marketing just imagine some fake use case in their head.
Market Research is real research. UX study is a real study.
now everybody just vibe their job
It resonated with me too, and even as a once fairly prolific coder that now uses agents to do all the coding, I feel like this is all relevant and does matter. The agent does whatever you ask and provides thin judgements based on what it thinks is your motivation and context, even feeding it the context it will often make bad generic choices. I think this is the part hard to optimize because it's not a verifiable reward, it's the feel of the software. I think Herdr is a great example, I've tried several dozen terminal multiplexers / session aggregators and they are all vibed and different, but this one felt so much better to use because of its choices and actually increased the number of parallel sessions I was able to keep going. My own projects I find I have to stress every detail, that is what makes the software good now is really diving into it, the experience feels like just as much of a craft as coding was and 'taste' captures the concept for me.
“as long as it works” is doing a lot of work here. Did the people building it say “no, again” enough times, and at the right times, in order to make something that _actually_ works instead of _merely_ works?
For what it is worth, I appreciate you sharing your experience. Looks like you have more programming experience than I have in life, so your comment matters to me. It also gets rather close to something I wanted to evaluate in the post: "whether any real intuition or judgment went into [its] design". I think that's a big question to dwell on these days.
Ironically, though, I had written a post that had people debating whether I put any intuition or judgement into. Would've been real ironic to instead add a post-scriptum that said "Yep, none of it matters. I just slopped something in 5 minutes and it easily went to the frontpage. Look how easy it is!" but unfortunately that is not why I write, not what I intended to achieve, and not something I am equipped to do. I care about my work, which is why I care about your comment as well.
I wonder, when you have a McDonald's McDouble, do you really care how it was made or do you just care that it's yummeroni and costs $2.99?
People who care about their health aren't ignorant about that. I think that can apply to software too: if you care about safety or privacy you might think twice before using vibe-coded software.
it is profitable to sell McDoubles, since they are popular and cheap, but McDonalds CEO doesnt eat its own burgers
https://www.reddit.com/r/popculturechat/comments/1rhug58/mcd...
so it is profitable to sell vibecoding tools, but its unlikely to be consumed by "serious organizations"
That is a significant leap in logic. A vibecoding tool is not a McDouble. Nor is Chris Kempczinski the same person as Dario Amodei.
The biggest difference is that there are only so many ways to eat a McDouble, but you can actually wield a vibe coding tool with skill. There's nothing stopping you from looking at the code and asking for structural reworks or refactors that please your code taste. In fact one of the best use cases for AI coding tools is vibe refactoring old apps where the code is straight up horrible.
> but you can actually wield a vibe coding tool with skill. There's nothing stopping you from looking at the code
It’s hard to give something like “vibe coding” a rigid definition, but I think we can all agree that it’s inversely correlated with "looking at the code". The more I examine the code while building, the less the activity is about vibes.
So, used this way, it's true that it's not a McDouble anymore, but it's also not so much of a vibecoding tool anymore either.
I think babysitting the AI would be a better term than vibe coding. At work they call it "AI assisted engineering" and "synthetic coworkers"
how about tools like Lovable ?
is Lovable used internally to create and improve Lovable product itself ?
A significant leap in logic is my middle name :) (I am not an AI)
Except AI corporations by and large all eat their own burgers (since any developer not leveraging agents effectively is woefully ignorant)
i have never had a "mcdonald's mcdouble", whatever that is. i cook my own burgers because i am a human being and i want to eat real food, not nauseating slop that should only be fed to livestock
it is profitable to sell McDoubles, since they are popular and cheap, but McDonalds CEO doesnt eat its own burgers
so it is profitable to sell vibecoding tools, but its unlikely to be consumed by "serious organizations"
The LLM has its own intuition and judgement which at this point is quite good. I’m not just throwing that around, I use LLMs 12-14 hours a day and for a long time reviewed atleast 75% of the code.
Perhaps you're the one making it work well?
I've spent a day fixing a feature my very junior colleague pushed, and which was merged by my should-have-known-better colleagues who have almost as much experience as me.
We have access to the same models. Why did her version suck and my features don't need fixing?
There's still something there, thankfully.
Generating the first version is almost free now. Understanding why that version is subtly wrong, debugging it in production, maintaining it six months later and deciding whether the abstraction was a mistake are all still expensive...
Yeah. The Fable oneshot is a myth, the software will still require constant refinement and maintenance. I've been daily driving my "vibecoded" projects and they're never really finished, there's always something to do.
Very nice article. I agree with it, call it taste, call it judgment, call it quality, I want things done the way I see as satisfactory, and it requires me to make LLMs do things differently several times. Like exploring a tree with a heuristic they don't have and need to rely on us humans to give.
Hard disagree on "Good enough", most things are mediocre at best. Despite the massive AI boom I can at best count on one hand "fresh" products made with AI that are meaningful in day-to-day usage. Not to mention most of them seem to be targeting AI-space itself instead of bridging some actual business use case. Just look at all the products on ShowHN, you can "tell" a vibed product in seconds.
Mediocre is good enough, that's what it means.
Mediocre literally means middling, which may or may not be good enough depending on requirements.
You are using countless products that are built or have become maintained by AI without realizing it apparently.
I do notice and realize, and I've gotten to the point where I had to fork and patch most of them to fix the shitty AI generated code.
It's so bad, and so everywhere, that I'm genuinely considering switching to a pre-AI stable distro and waiting 10 years for the AI hyped to be over before updating any piece of software ever again.
I'd believe it, my spotify breaks all the time now and I seem to remember them announcing none of their devs code anymore. not 10 seconds ago I had to force close it to get it to stop playing music on my phone.
Name them. Just 2! Any example would be enough.
Or well better than 0 and just a claim.
I think you agree with what is in the essence of the post, but has been worded poorly. I do agree that most things are mediocre at best. Hell, I'd argue that most of my comments on everyday observations are negative. Ranging from "this sucks :/" to "holy fuck". They are, however, "good enough" for some and despite how clever I must to notice how much something sucks, that does nothing to the fact that it will receive adoption. Turns out I'm not clever. Huh.
In hindsight, I have worded this rather poorly and let some phrases do a lot of heavily lifting for me in the post. Ultimately, I agree with you. Knowing, for this instance, is not half the battle.
How would you count them or even know if they were made with AI or not ? Such nonsense
If you are waiting for someone else to make a good product for you with AI, you are missing out. The best AI products have userbase of one.
The best AI products have a userbase of one. The average AI product has a userbase of zero.
I've run this thought experiment a number of times with different sections of my (relatively) diverse friend circle. I ask them a version of this question on repeat: "what if AI automates X aspect of your work?", and the final answer (either because of frustration with the loopy nature of the question or a genuine stopping point) arrives somewhere along the lines of "my judgement," "my knack," "my embodied nature," or "my quality of decision making."
I think the core still leads back to human agency and the ability to consider aspects that wouldn't fit into an LLMs limited context window or be able to be vectorized into a DB, including ultra-long-term consequences (especially those with great thinking abilities).
I still happen to think that AI/LLMs will never be able to "fully" replace humans, because the evolutionary process that led to our cognitive abilities and the way we train LLMs are vastly different thanks to different pressures, but maybe that's just me defending the last bastions of our collective humanity as a human; I don't know what else to root for.
Right now we're all training AI models on our taste. All of our chat logs are being uploaded, and the AI companies are training the models on them to get them right the first time.
It will soon no longer require taste. Or at the very least be 95% of the way there on the first run.
This was the basic theory behind Tesla eventually being able to stop killing Tesla drivers with FSD. It has not worked out.
Google also uses ML to train the Waymos, they just have better sensory capabilities. Lidar is much closer to how the human eye functions (inverted, I suppose) than cameras are anyway. But FSD is still really cool, you just have to make sure you don't take your hands off the wheel!
I think it's obvious that repeatedly noting that the absolute perfection isn't achieved yet is an absolutely useless way of tracking progress. No?
BTW, when was the last time Tesla driver died due to FSD fault?
> somewhere along the lines of "my judgement," "my knack," "my embodied nature," or "my quality of decision making."
Do your friends actually believe that AI can't automate these parts of their work? I think my answer would be that I'll just have to learn to do something else.
> I still happen to think that AI/LLMs will never be able to "fully" replace humans
Do you mean replace humans completely or just replace all the tasks humans don't want to do? I would hope that humans will still have a reason to live even if all the work is done by computers, and that the computers wouldn't intentionally wipe us all out.
Judgment was all there ever was to begin with. It's why it's what God himself does, although I'd be curious if there are deities that don't involve judgment in some form.
Anyway, I completely agree and the past several years have been an absurd journey into borderline psychosis with some people.
Taste ends up providing tangible value.
If you have taste in writing code, you spend less time on bugs and can implment features more quickly. This is still true with LLMs if you learn how to direct them instead of having them direct you.
Someone else without taste can re-implement what you have done with AI. But it will take them more effort, more time, and more money.
But that's always been the value in writing code. Any MBA can end up producing software to accomplish something by hiring other people. But their company might not get off the ground because of the expense of not having technical competency (or engineers with taste) at their core.
Taste was all that was left when product design peaked in like 2010-12 too and it was a great time.
But the advent of design systems and software like Sketch and Figma pretty much automated the job away. You didn’t need the raw Photoshop skills anymore - any CMU grad could be a top tier designer now.
I see the same thing basically happening here. Taste will matter for a while but eventually people are just choosing the most accepted/expected optimal choices and anything that falls outside the trend will be “not good” in the same way skeuomorphism in design is no longer good.
So we’re lucky we’re in a taste era - it’s a lot of fun. You (yes you!) can still change it, shape it, establish the status quo of tomorrow
Design is not moving pixels around in Photoshop or Figma. CMU grad could be a designer, but to be a great designer you do have to have... well, taste. Figma didn’t change that one bit.
I never quite fell for the scam of “designer who can’t photoshop”.
You can’t do product design without visual design - and many designers don’t. They’re picking from UI libraries like Tailwind or MUI, which is really something any developer could do, and many do. That’s why the job mostly died.
Even the most abstract of painters, fine artists, and so-on should have good fundamentals. They’ve gone through the discipline and developed taste, they don’t choose it from a library.
The CMU grad (using this as a stand-in now for any so-called educated hack, sorry CMU) is skipping all that and asserting personal authority without proving or living anything. And that shows IMO.
My probably unpopular opinion is that using AI can be very creative and you can get into flow states, and the quality of outputs vary by skill of the user.
> Taste was all that was left when product design peaked in like 2010-12 too and it was a great time.
I'm not sure I agree. I spent a good chunk of those years converting designer's PS sketches into CSS. Bootstrap and friends were a godsend, because even if standard design systems over-optimize for mobile, they are still far more functional than most designers's mediocre and incomplete views of user experience.
I have yet to see a good design falling out of people building flying castles in Figma, disconnected from the realities and basics of web development's languages HTML and CSS.
Taste is something you use to build the guardrails of AI over time and many paper cuts endured. Once you have a big enough set of these (call them rules) recorded, they become your development substrate. You narrow the paths the agents can take to arrive at a coherent system. If you don't do that, superficially it may look similar, but it's not and users will be able to tell.
"De gustibus non est disputandum" ("In matters of taste, there can be no disputes")
One thing I'm excited for is the expanded ability to modify and create my own interfaces and software so I don't have to be subjected to the bad taste of others. An example of this is something like Chrome reading mode that removes all the stupid decoration from what I'm trying to read.
My workflow has become a sort of "living process". When I encounter friction in some part of it, I focus my attention on it, it adapts, grows from from an annoyance to something that I delight in when using.
Some small examples: Google Chrome doesn't have default shortcuts for accessing bookmarks so I made an extension for myself that does this. I have a SketchyBar setup that I'm constantly updating. Recently added my CC/Codex plan usage to it so I'm no longer constantly running /status in the different CLIs. Etc, etc.
> Suppose you have taste. Suppose you paid the full price and you can feel the sag in the sentence and the wrongness in the function. > > Congratulations! You now ship at exactly the same speed as the person who cannot.
Yes, previously the person who didn't understand shipped faster. It was a long standing problem. The people who crapped, got the bonus, and ran were much faster than the patient engineer of quality.
By the way, what were the essays that formulated Claude Style? if the author was indeed formed by them, I'd like to read them - or at least enough of them to understand.
Also just a shout out to the author - you're getting a lot of flak here, but your article hit me at the right time in the right way. I trust you're telling the truth about writing it yourself, and thank you.
I keep seeing people trying to define judgement and taste and then they don’t actually like cite Kant’s Critique of Judgement where this is basically the entire concern.
Shouldn't this 236-year-old paper be updated in accordance with, say, neuroeconomics research on how provenance influences aesthetic taste?
Speaks more to the myopia of developers than anything else.
Most people in Comp Sci don’t read Kant unless they have a concentration in Artificial Intelligence, so I’m not blaming them.
I'm not only referring to that, I'm also referring to a reductionist view when I read "Taste Is All That's Left" - which is a statement that has been mindlessly parroted around as of late.
Yeah even if Dev Schumer is some absurd hype artist he was correct to note that it is judgement and taste, not taste alone, that is most key now. Although he attributed it to the models own subjectivity.
Why is it a reductionism view? A true reductionist will no doubt claim taste won't left as well, at least not as a non-private langauge.
Nor the work of sociologists like Bourdieu's Distinction.
Hume also has an essay about this. People don't read philosophy anymore. They think blog posts are just as good, or at least more "relevant", somehow.
I'm not disagreeing with you at all, but I have personally found most philosophy "classics" painfully hard to get into. That doesn't mean they shouldn't be respected, but the truth is that sections of society (like myself) do rely on the people specializing in philosophy parsing it into current language and metaphor.
Look I mean, many famous artists don't understand their work at all. You don't have to have read philosophy to create something meaningful or make an insightful comment. But if you are going to start using terms like Judgement or Taste, you really ought to know where they come from.
There's a lot here that I enjoy muddling on. And I think taste is very vital. Taste, ambition, seeing possibility: these resonate to me, long resonate, as what I've wanted to see.
And honestly: I've been tired of an age of engineers that are beat down, that only want simple, that finding the most blunt approach is the only way. It has felt tasteless. Anti-ambitious. It often is still very sensible and practical and what you should shoot for! But there are also people wandering around now trying much harder! And I am excited for those frontiers! I think it will challenge and shake the foundations that we've accepted as true, as There Is No Alternative, in exciting bold and fun ways.
I am however pretty unconvinced by the article. There's a lot that doesn't quite work for me, that isn't building the case I'd like, whose takes are off from my read.
> The output is good enough—that is the problem. Good enough is a solvent. It dissolves the reason to do better. For as long as making things was expensive, the expense did quiet work on our behalf. It rationed output.
Agreed about the rationing of output. I do think that the article though continues to show a magical thinking. That we have these things now and they autonomously do the thing. That the LLM's have solved it all.
The article itself goes on at length about how nebulous and abstract taste is:
> Taste is that. It is the compressed, wordless verdict you reach faster than you can justify. It is the “no, again” you say to yourself with total conviction and no available argument
But this contrasts so distinctly to me against "the output is good enough". Is it? That depends. That depends on your taste. The proximate first results come quick. But the technics underneath? Those themselves, in my view, rely enormously on engineering taste to support and advance. I think we see a very similar sort of magical delivery thinking, very clearly on display here:
> When the factories came, they could suddenly make everything—cheaply, uniformly, by the thousand.
As if there was some magical "good enough" transition where suddenly the aliens came and gave us this box that just does the thing. As if we discovered the right formulas and math and now: we had production. Again I think there's just an enormous amount of work and taste that is still actively required inside the factory box, that building the industrial processes is incredibly intense & difficult, even though we have reliable industry-line production and now robots doing the labor.
There's still so many gems, so much lovely material for thought. I love the provocations here, and I think there's a lot of great calibration.
> The friction was not an obstacle to developing taste. The friction was the curriculum.
Makes me think of yesterdays @apenwarr banger,
> Every slow prototype started out as a fast prototype, I think that’s how it goes https://bsky.app/profile/apenwarr.ca/post/3msemlo4rds2h
I agree with your criticism of "the good enough" and "done with taste" dichotomy.
I would like to comment on this, though:
> that only want simple, that finding the most blunt approach is the only way. It has felt tasteless. Anti-ambitious.
I am always aiming for simple. If it's simple, it's maintainable and can be easy to reason about. Doing that is hard work, though. I used to spend three iterations to achieve it: functions are simple, easy to reason about, properly named and composable.
It's the opposite of blunt, though, so it might be that you meant something else with the term "simple" :)
Perhaps it's... "simple to implement."
Usually it can be that the complex artifact is the simplest to implement. It's harder to make the simple thing
Classic, Rich Hicky, "simple made, easy" talk reference.
This post is awful! Annoyingly written and zero insight. Most posts about this taste concept are bad (and there have been so many), but this is below average.
You've read many of them? Then you'd be the best to review even more of them! I'm ready for a raw and truthful review of my very own edition of the "taste" blog post, posted mere ~6 months ago, on the same topic: https://emsh.cat/en/good-taste/
Ha! I have a heterodox view on the taste thing, which is that it’s mostly cope: the models apply something like taste in nearly everything they do, and these taste posts will disappear as that becomes more obvious. However, besides the substantive disagreement, I think your post is fine since it appears to be your actual point of view, written down by hand. It’s not particularly novel or insightful (most of these takes are basically “there are still some things the models can’t do” — no shit!) but it’s measured and reasonable. Fine.
Heh whoops. I thought my other comment didn’t post so I wrote this one instead.
This post sucks, and it was also written by Claude.
Everybody thinks they have good taste don’t they.
How convenient it is that we’re all so awesome according to the one metric that can’t be measured
yeah the skill du jour keeps shifting. Everyone wants to say they have something the llm doesn't. The more subjective and ill-defined that "something" is the better.
> Everyone wants to say they have something the llm doesn't.
LLM can't waste weeks debating stacks and tools and then recast the problem as a new framework that needs to exist.
*Check mate*
I used to think the average HN reader probably did have good taste, at least above average. After all HN is a place for builders, making things. Most people never make anything. Making things means deciding what to make and how it should be, and also what not to make instead, so I think charitably HN probably does have higher average taste than random people or most other web forums.
These artsy fartsy takes on a subject that would be better studied as a science are fun but are they useful? I am not saying there is no such thing as taste and neither will I attempt to define taste except to say "tasteless" is probably the opposite. "Judgement" seems more valuable as some have mentioned, maybe judgement and taste together identify something we want?
Still, I like better when we actually do try to name what we're talking about. We've probably all read the motorcycle guys take on quality and things like that right? I guess the theme some are going with is that quality can't be objectively quantified and to define it is futile but I've never liked that take. If we think LLM generated code is low quality then it certainly makes sense to start our attempts to define or at least describe what makes the quality needle move, because language is how we encode lots of information and information is what feeds LLMs.
People also like to rag on uncle Bobs takes, clean code, etc but I like how at least he tried. I've been a fan of his for awhile but I'm not big on objects so don't care about his pushes on that sort of thing. The best take that offered hints that I would call quality is "A Philosophy of Software Design" by John Ousterhout, there is plenty of stuff in this short little book that was very satisfying to me, like for instance where Bob wants short/small functions, Mr Ousterhout was much more flexible on this, like for instance maybe some functions can be bigger if the need arises, things like that.
This brilliantly articulates what is required for best practices in the AI era. I am putting this article on my required reading list for Software Engineering. On behalf of my students, thank you for writing this!
In an ideal world, something as nebulous, subjective, and unquantifiable as "taste" would never be the sole purpose of a technical practitioner.
That so many developers believe this, is surely a sign of the fields impending death.
> Everyone can make anything. Almost no one can tell you what is worth making. That was always the harder skill. It is now the only one left.
But people can and do tell you what is worth making. Taste is largely set by class and power. The “skill” is having the right context (wealth, education, network) and then pattern matching on it.
When resources are scarce, you are forced to use them in a meaningful way. You learn to do only what matters most: what is right, minimal, essential. When you need to choose wisely when resources are scarce -- this is how you learn the taste.
Now that creation is cheap, the scarcity has moved to adoption. Things built without taste are less likely to be adopted, to survive. At least that is what I'd like to believe.
Absolutely not. Scarcity forces you to do with what you have. Only in abundance, only when you can have everything, you have room and reason to develop taste.
It is quite the opposite in my experience.
Who had historically the most taste? The rich or the poor? And who had the worst taste? The people who grew up in poverty mindset that got rich. These are widely shared, common cultural observations.
You need a long term, persistent abundance to develop and even have a use for taste.
I don't see why would software be any different.
Is taste really the one thing LLMs can't take away or do we talk about taste because judging has become the cheapest move available? I think focusing too much on taste is how we lose focus on the two questions that actually matter: what do I actually care about, and why do I make things at all.
You cannot separate "taste" from the question of "why do I make things at all?" because they are inextricably bound. The reason for your making something is where taste is expressed.
I find this post disheartening. To my eye, it seems to be the author's first blog entry with obvious LLM contributions.
Given the recent "moving to a farm" ... 7 days later ... "leaving the farm" posts, it's hard not to read this as someone clearly struggling with the burnout which is all too common in our industry, and resultingly giving in to the false promise of a machine that can just handle the thinking for you.
From my interactions with him in some nix forums online, he is not in the industry. This is hobby work (not to devalue it, as his software is generally high quality).
They've updated the article to state that no LLM was used.
Yeah, I don't believe them.
It's only because you'd rather believe that you are being lied to than mistaken.
No, because the writing is objectively bad in ways that LLMs excel in producing.
Either they used an LLM or they have no taste and have used LLMs far too much elsewhere.
> No, because the writing is objectively bad in ways that LLMs excel in producing.
Author does not dispute that. Yet they claim to not use AI. So your preferred belief is that they are lying, rather than a thing that looks AI written might have been actually written by a human and you are wrong thinking that it can't be true in this case.
Yes that’s right, I and many other readers think the author is probably lying when they say they didn’t use AI.
The update itself has strong AI tells:
> I do write like this sometimes. The short sentences, the reversals, the one-word lines—all of it. It’s just the way it is.
> That says something about my writing. It says nothing about who wrote it.
> So let me be plain about it: Claude was not here. No LLM wrote this—not a sentence of it, nor was it outlined, drafted, reviewed, checked, etc. by one
Either the author is doing this on purpose, or this is some kind of art project. I don’t buy the “just a coincidence” story at all.
Right, I feel like I'm taking crazy pills reading all the comments taking this article seriously.
I think AI has broken HN. The readership is so head-in-the-sand about AI that they wouldn't recognize 2000 words of Claudeslop if it genuinely slapped them in the face
Most commenters only read the title. That's it.
The alternative explanation (the top commenters actually read it and somehow didn't recognize it's obviously written with AI) is just too bleak and I choose to not believe that.
It's LLM-like in the details, but there are certain patterns of repetition and idiosyncrasies, also certain analogies I don't think an LLM would make (like the aliens thing), that make me believe that it's written by a human.
The whole post has that immediate LLM "mouthfeel"
> To my eye, it seems to be the author's first blog entry with obvious LLM contributions.
Not to me. It's actually quite sharply written. I enjoyed the style a lot.
I think whether "taste" was ever important comes down to whether software is a commodity to be bought and sold, or something more strategic.
The company that needs a webapp because everyone else has one is never going to value "taste".
To adapt to this AI world, we need to either ship more lower-quality software, or start seeking out higher value problems. I don't see any other way around it.
> I do not miss the wall, and I will not pretend to.
The wall is the point, if you want to truly know the truth of something. I still believe there is no royal road. Taste isn’t the answer to why you would choose a bump allocator over a slab when designing a system. It is experience, wisdom, and the long hard road of trying yourself and failing. Taste comes after.
There is much joy in suffering when the result is knowing.
Too many people have tried to declare victory over software engineering prematurely and it shows.
No, the AI is writing code not architecture. And it's the same for all disciplines.
There will be no mass job layoffs for antying but relatively automatable stuff.
If taste that's all that's left you've kinda really got nothing. Taste is an anti-moat, it's sort of a measure of how well you, as a literal receptor, agree with a pile of people out there on something. Like sure its something you can sort of develop but it costs you nothing to use, and if you're not creating something off-manifold you're just smearing stuff around that's already on it. Horrendous kind of take on "taste" that's taken hold of everyone. Kinda like if a music artist stopped making music to make spotify playlists. Sure they could maybe make _good_ playlists, but if they are sitting there crowing "taste is all thats left!" it's.. what is that? It just sounds like either a deluded giving up or just a very direct admission of giving up.
Taste seems something that should be more related to a craft but when doing engineering we don't talk about taste, we talk about theory, systems, practices, skill, intuition, experimentation, teaching and learning. So the debate on whether taste matters or not might come down to whether creating software can be called craft or engineering. In my opinion, maybe programming was more of a craft before LLMs but it looks like it is being squished between the engineering aspects of producing and maintaining software going ahead.
It reminds me of I think Nietzsche, isn't it a quote of him? "In the end all that's left to us is taste"? I can't seem to find it so quickly, so maybe my memory is failing me. He definitely had a lot to say about taste.
I won't try to posit whether this post was AI-generated or not and I do not think it matters. That focus is a defense mechanism for handling an uncomfortable idea.
The idea is this: the commodification of intelligence requires the codification of domain expertise to function reliably, and proper governance to scale.
Taste, in my opinion, is a fill-in for absent function and lack of utility. Systems function; they don't require taste. Taste is why PG put Sam Altman where he did — not a system of merit, not evidence. Taste is why DJT can do no wrong in the eyes of his base. It is unprovable and has no true measure, and that is exactly why it gets leaned on so heavily: it lets you claim superiority no matter what slop your system produces.
You say the box moved taste up a level. Out of the making and into decision layer. That relocation is a smell. Your unfalsifiable thing conveniently gets moved to the one place it still can't be checked, at precisely the moment the making became checkable. Just like: Iran/Strait of Hormus, AI curing diseases and all of the other bold promises made by people who claim to be "taste makers" but in the end its all the same. Bullshit! Because thats what taste is a euphemism for in this context. Convincing others that you are correct in some way with no actual proof to back it up.
The vector here should not be "taste"; it's tolerances: written down, measured, enforced. The judgment gets encoded into a spec, and the spec is what survives the person who had it. That is codification; not "taste".
> Robert Pirsig spent an entire book circling a word he refused to define, because he had convinced himself that defining it would kill it.
Just reading the title and I was already thinking about that book. It's a great book though it gets progressively more difficult
> The friction was not an obstacle to developing taste. The friction was _the curriculum._
Prescient.
I really have no idea how future software engineers will know how to write software in a world wherein software is written for them wholesale.
Taste will always be there.
But toil is not gone. We have still have code owners, lawyers, physical hardware, thousands of configurations that we can't test them all.
And notably: we will still be asked to write software we can maintain in the future.
Sure, there are short-lived non-critical software projects where LLMs will rule.
But for many big complex projects, you can't land code without reviews. You can argue its because taste is already being enforced in the name of maintainability.
> But for many big complex projects, you can't land code without reviews.
Why do you think orgs are pushing less reviews? Token-maxxing increased PRs but stupid meat brains are a constraint. So of course, remove that constraint and you’ll have your features faster!
Software engineering is and has always involved engineering situations where you can let your co-workers/third-parties run wild with limited review oversight and limited blast radius.
Example: a plugin architecture with a solid API that makes degenerate behavior limited to usages of said plugin.
If we define taste as the intuitive act of saying "no, again," then I fully disagree with this whole article.
Every AI-frustrated (but LLM-written, sigh) blog post about the loss of taste and craft and hard work in development sounds like we've given up the interaction with the machine. Like human work is sitting back in your chair and shitting on stuff.
That's obviously not true! That's not how any of this works!
It's hard and weird to develop with the LLMs because they just do stuff. Lots of it is good, some of it is OK, some of it's horrible. Unpacking what it's done is hard and weird because software isn't just lovely UX, it's also data structures that scale and performance and privacy and enterprise controls and SOC 2 and onboarding and accessibility.
If you want to build real software, all that stuff has to get done. Today you're working on the feature, tomorrow you're making it scale. It's long-term and iterative and complex and hard to pack into a prompt or a markdown spec.
The work is the work, done at and with the computer, and it's way more than just "taste."
« Software works the same way. » No, I don't think so. Modern software has been around for about 50 years, the Internet & OOP for about 35 years. People have been cooking steaks since antiquity!
If your opinion or understanding of how a program works or how it should work is your main asset for keeping your job, then I think you're the one who's going to get cooked. You’d have to be rather disillusioned to wake up in the post-COVID era, with technology that triples its algorithmic efficiency every year, and conclude that a dehumanizing industry will still need our expertise in the future.
The key words here are “dehumanizing industry”. You can always bake your bread the old-fashioned way or cook your steak in « your own small restaurant », no one will stop you. But small restaurants, at least where I live, aren’t doing very well.
On a side note, since the author has included a detailed explanation that the text is not AI generated and this is her default writing style, I just find it very difficult to read this style of extra punchy writing. I don't know if it is just me or anyone else as well.
I assume it's common and understandable to dislike LLM generated content and therefore anything that resembles it regardless of origin. For those who happen to write in a similar way it's a pity.
I've noticed some people on the internet dismiss content just by its length too now. I assume this is because they're not used to the idea of a human producing long form content (as opposed to the short form rewarded by popular feeds). I wrote a post (other forum, other handle) where I included some markdown tables and detailed explanations in an attempt to be thorough and help others answer my question and got accused of having LLM generated it despite writing it myself. Maybe the usage of markdown tables or the length of the post was the issue. Anyway, it's not nice being dismissed as AI slop when no AI was involved. I also don't see how there could ever be a reliable way of identifying origin of digital content such as text on a forum.
> I also don't see how there could ever be a reliable way of identifying origin of digital content such as text on a forum.
Perfectly reliable, no.
But:
1. these models aren't even trying to hide their nature, if anything the opposite
2. a great many users of those models are (lazy or time constrained, doesn't matter which) or lacking in style of their own, and in each of these cases the result is they don't prompt the models to hide their default style; worse, the outputs keeps getting RLHF-ed so despite how much of an "ugh" factor it has to the reader, most people making these things, to the extent that they do review the content at all rather than blindly uploading, actually like the style
3. Models don't share state between users. This means the model won't act "bored" from having already written the same thing before, so when 5000 people all prompt for "blog post about X", then it will create 5000 blog posts with almost identical in writing style. In turn, this means that the style of the most-used model (I assume for English-language blog posts this is still ChatGPT) has become cliché faster and harder than almost anything else.
I really don't get why people here are so obsessed with "taste". I also find the premise of the article to be completely wrong. Taste, not only in software, is only the only thing that's left for humans, if you ignore the complex social environments in which our technological systems are embedded. AI puts the emphasis in the socio-technological system on the social. If you were never social to begin with, or if you deliberately ignore the social embeddedness of our systems b/c you are terminally valley pilled, you might come under the impression though that indeed, taste is the only thing that's left.
I'm worried about taste because to me it seems like something like taste would be the best thing for a ML model.
That's interesting. I would (and I think most?) would say the opposite. Good taste is hard to define because it is very contextual. There's popular taste but also there's artisanal/specialized knowledge taste. One potentially leads to mass adoption, the other leads to higher value for a few people.
LLM-based code generation requires disciplined operation to lead to the best outcomes. Shops who only ever had taste as a proxy for discipline were missing this before, and they're missing it now. The fact that it's missing is why so many organizations are tolerating ever-growing mountains of slop. (Not new, by the way; just dramatically accelerated.)
Discipline can be linked to system design standards and your review process, and it can be translated to measurable outcomes at scale and over time, regardless of how agents are involved: unplanned work, time to recovery, employee retention, etc. When outcomes like these become your priority, you see that you can't rely on taste alone.
Okay that's a clever way to put it
I tend to like images in a split second, when they try to achieve a certain aesthetic at least. I can later inspect them more carefully and understand why they're good.
Reading text, it can either grab me or it doesn't. I've asked about that to an AI and there's some reason for that. Some text is more compatible with me, due to the rhythm or combination of chars/words used, how they actually look/feel, it's not just about the content. So I'm assuming some writers are going to be more compatible with me while others will make me dismiss them quickly or just skim.
I like to think I have the correct objective taste in music, but everybody thinks that about themselves.
> later inspect them more carefully and understand why they're good
Reductionism is hard to apply to intuitions, and we often fool ourselves.
I personally call this the obviously test: smart people correctly notice something and say it is obvious. It is a rather rarer skill to understand and explain why. I have particularly noticed it with coaching (which adds the next levels of difficulty in getting someone to grok what you've seen).
People will always want something to gatekeep their chosen field. Code went away, so now people need to be able to say if the machine wrote good code. In a year or two people won't be able to use that because the code will be measurably better than something a typical dev would write, and in five years it'll be better than anything anyone could write.
Then we'll need to accept that coding is better when it's done by computers, and we'll start telling everyone it's actually the prompts and context that steered the machine that really matter.
In couple of years data compression will have progressed so much that you ll be able to store all the data in the world on a floppy disk!
But this is basically reinventing the term "Aesthetics" which is a field that had several centuries of philosophy research about it. Not all of them will agree AI cannot have it.
Actually I think it is possible to reconcile Kant's third critique with the statement "LLM is possible to make aesthetic judgement" (with certain generous application of principle of charity, naturally).
AI does have something like Aesthetics and taste, but it's the average of all the inputs and so it's almost by definition the answer to 'what is the most average design possible based on all human prior data inputs?'
That's why AI prototypes of front end UIs tend to look the same without massaging them with your own design tokens, design language, and injecting your own taste.
A poem by polish poet which say exactly this.
TASTE
For Professor Izydora Dambska
It didn't require great character at all our refusal disagreement and resistance we had a shred of necessary courage but fundamentally it was a matter of taste Yes taste in which there are fibers of soul the cartilage of conscience Who knows if we had been better and more attractively tempted sent rose-skinned women thin as a wafer or fantastic creatures from the paintings of Hieronymus Bosch but what kind of hell was there at this time a wet pit the murderers' alley the barrack called a palace of justice a home-brewed Mephisto in a Lenin jacket sent Aurora's grandchildren out into the field boys with potato faces very ugly girls with red hands
Verily their rhetoric was made of cheap sacking (Marcus Tullius kept turning in his grave) chains of tautologies a couple of concepts like flails the dialectics of slaughterers no distinctions in reasoning syntax deprived of the beauty of the subjunctive
So aesthetics can be helpful in life one should not neglect the study of beauty
Before we declare our consent we must carefully examine the shape of the architecture the rhythm of the drums and pipes official colours the despicable ritual of funerals
Our eyes and ears refused obedience the princes of our senses proudly chose exile
It did not require great character at all we had a shred of necessary courage but fundamentally it was a matter of taste Yes taste that commands us to get out to make a wry face draw out a sneer even if for this the precious capital of the body the head must fall
I don't agree with the premise that creating is no longer the hard thing. It's only been automated when your taste is bad.
In software development, good taste translates to readability, succinctness and maintainability of the code. It also translates to intuitiveness, reliability, security, performance, scalability and utility of the software.
All this doesn't necessarily translate to sales though. Because, to evaluate those things requires attention, trust, time and effort.
This is a significant barrier because a lot of software will appear to meet all of these superficially.
After 1 day of usage, a piece of software may appear to be intuitive, reliable, secure, performant and useful... But then after some time (sometimes a whole week or longer) you run into a critical scenario and discover that it cannot be solved with that software... Or performance drops off sharply after you created the 1000th record in the software... Or a hacker takes months to find that one endpoint which allows full remote code execution.
Even in an optimistic scenario, 1 week is a long time to evaluate a piece of software. I've encountered software which took 6 months and large teams of people to realize that it wasn't suitable. That's how long it took to hit the critical limits. It's a very long evaluation loop.
So you cannot judge new software efficiently by just looking at it. Even industry consensus is problematic if the software is very new and complex... Plenty of trends have fallen off a cliff in the past. You need to understand who is behind the software... And even that's not so easy; social proof can be misleading when it comes to deep technical ideas. The people who are good at social networking aren't necessarily good with tech.
I think AI slop code is going to be a much bigger problem than people anticipate. And the irony of it is that the solutions already exist... What doesn't exist is the mechanism to identify those solutions. But even the mindset needs to be corrected first.
> I think AI slop code is going to be a much bigger problem than people anticipate.
I'm seeing the same thing. Of course, it's always possible that AI capabilities grow significantly from here and this worry is moot (and there will be other things to worry about, like whether I'll have a job). But seriously, after a while AI generated codebases really fall apart.
It's a bit of cope from me that this problem won't be solved in the future, but I'm also really disappointed with how many people really do not care about writing high-quality software that is easy to understand, performant, and resilient to bugs. Like, why did you decide to work on software in the first place?
I think we're going to see some pretty careless mistakes in the future that will result in a lot of pain...
I'm feeling quite confident that AI won't be able to produce quality code in the foreseeable future. Because the bottleneck is not technology, it's people and incentives.
Firstly, people don't agree on what good code is (developers who get paid by the hour tend to favor complexity so they think good code is complex code), secondly, AI companies have an incentive to produce more tokens; this works against good architecture since good architecture would cost fewer tokens to maintain. Thirdly, many decision makers tend to conflate complexity with intelligence, hence they are more likely to prefer complex solutions over optimal ones.
"Taste" and "personalization" are this month's AI talking points.
> The market timed you both with the same stopwatch and it did not see the difference. It cannot see the difference. Taste does not show up in the diff.
Doesn't that mean guaranteed failure in the market if you use taste as your guiding star? Therefore, there's really no way to enter the market anymore.
I don't know how you then ship software or monetise it...? The market will just clone it.
So... Don't share stuff publicly? Only share with closed groups? Where does this leave small-time software development...?
> Doesn't that mean guaranteed failure in the market if you use taste as your guiding star?
If taste is maintainability, etc.. the no quite the opposite.
The startup that has no taste and vibecodes mindlessly might end up with a codebase they can't maintain or refactor. A product that can't advance.
But honestly, I don't feel like the toil is gone. Landing code in big projects remain hard.
Someone can replicate any high fashion piece of clothing you make by hand. Maybe they could even make it better. Buyers still care about the origin, just in a pretty small market for it.
Maybe software moves to this after it's commoditized? Users might prefer an artisanal one vs ordering it off the shelf or using an agent themselves.
Well, this has been true (to a lesser extent) for decades. See "Worse Is Better" and all the subsequent essays and writing following on from that
My taste agrees with the point made but thinks this could've been a shorter article
I've been circling this insight for a bit previously, leading me to a convo with AI about how we are now approaching more of a spec - no, prompt - no, context - no, [insert newest paradigm] driven design approach
The point being more that the development does weigh more heavily on having fun ideas to ask AI to build
I think there's probably also a place for people to make guides for AI-assisted programming or even personal essays on their views of "taste" or what they want code to look like?
I see pop up frequently objections to vibe coding (different from mere AI assistance in coding - vibe coding being creating things without looking at the code at all) that it doesn't account for architectural descisions, cybersecurity decisions, edge cases, and other such things. So to me it looks like a next step might be in identifying these things "vibe coding" doesn't do well, that people can learn more manually, so they can do at least AI assisted coding well.
I think you missed the point of this article.
Wow does it feel good to read human prose again. It just feels so - refreshing. Like cold water in a desert
Going to drop this https://commoncog.com/the-tacit-knowledge-series/. Tacit experience gives you taste in my opinion.
[flagged]
Whenever I see "is doing a lot of work here", I assume it's AI-generated.
> I do write like this sometimes. The short sentences, the reversals, the one-word lines—all of it. It’s just the way it is. A LLM writes that way too, because it was trained on the same essays I grew up reading
I'm extremely irritated that the post is being accused of being AI slop. I have a visceral reaction to AI generated writing, and this post rang 100% human to me.
Makes me actually quite angry that one can no longer use short, pithy sentences without being accused of using AI. If AI causes everyone to reject actual decent writing because someone can't distinguish between it and that from a human, we have other problems on our hand.
The content of the post resonated with me 100% as well.
Same, the post has an entirely different speech rhythm than AI.
In Walter Murch's "In the Blink of an Eye" he describes how, when cutting a movie, the reason cuts work at all is because human experience has cuts: we blink.
More specifically, we blink when a thought is complete, to finish one thought and begin another.
When a movie is well-edited, the audience will adapt to the rhythm of the movie, and you know you've placed a cut at the right moment when the audience will blink at the exact moment of the cut.
The same "rhythm of thought" applies to written essays as well, and once you get a feeling for it, you'll notice how AI generated text follows no such rhythm, no such structure.
AI generated text has the rhythm of TV salespeople combined with tiktok brainrot, overwhelming you while at the same time saying nothing at all.
Is the post-mortem written to sound like AI slop, or is it just that? Offending either way :P
It's an interesting article. To me it reads like someone who is not confident in their writing, but is very interested in getting their idea across, iterated with an LLM to produce it.
It set off all my LLM writing triggers and I noped out after about 3 sentences, because I'm designing/coding with AI all day and I'm 100% sick of its writing. I don't really care if it was written by an LLM or a person who writes in LLM style.
HOWEVER, the surprising fix for me was to ask a number of LLMs to summarize it. I omitted the "is this AI slop?" parts when I pasted it in. Like I said, I don't care about who wrote it, but I hated the writing style.
The AI summaries were quite readable.
They converged at a high level on the summary, but with a good bit of variance in nuance. I had a good follow up discussion with one of the LLMs, so I'm glad I tried it.
You hate LLM writing, so you took human written words and ran them through an LLM to read it?
Hate to break it to you, but you don't hate LLM writing.
my idea is "skin is all that's left"
We have the Power of gods. Now we need the Wisdom of gods.
How is this at the top of HN...
I agree with the idea that the taste, for lack of a better word, was the only thing that ever mattered. What I disagree with is the idea that slop that is “good enough” is indistinguishable from something that is crafted is “indistinguishable by the market”. I think it is indistinguishable in the short term, but very distinguishable on a long enough time span. The same way that hacked together slop has always been “good enough” to close a ticket and even get someone a promotion, but later costs the company thousands of engineering hours to rewrite once it becomes completely unreliable, slow and impossible to modify. We’re just speed running that now. It’s the same battle engineers with “taste” have been fighting with sales teams and product managers and lazy teammates before LLMs, just in fast forward.
Humans will become the great filters.
Don’t need taste when you’ve got the market homie.
Great artists ship.
Taste good, taste hard to cultivate, taste easy to fake and hard to prove. There, TLDR.
We still have a metric for good taste, and it is $. Solve a problem that people care about? They give you $. Solve problem no one care about? No $.
It is not perfect, but I challenge you to find a better metric that is as effective and simple.
That could be true for novel problems from Deep Tech industries and not everybody can solve those.
If you are doing another app, the product experience and marketing will probably win. Many companies with better products lost, beaten to death example is Betamax
I think it's a pretty good direction to go down. I wonder if we can add a little more accounting to it. Cheap shoes solve an immediate problem but you pay more long term versus expensive well built ones. How much $ (in dev time + tokens) was spent in incidents and/or fixing code later. Credits and debits. Weigh money earned (somehow) against money spent.
What's special about the fleshy brain that a digital chip cannot replicate its judgement?
The Rick Rubin meme
Oof :)
The irony of OP championing for “taste” in the age of mass produced slop, only to then post something that was likely produced by repeatedly reprompting with “no, again”
I found this post a bit too verbose and incoherent. It does read like AI slop to me, but the bigger issue is that its core argument itself seems flawed.
The author seems to be advocating for choosing to "carry the burden" of one's taste (or aesthetics, judgement, standards, etc) while bemoaning that the market may not sufficiently reward for doing so.
IMHO, the core flaw is in mixing these two things 1) the market rewarding what it chooses to reward; and 2) individuals choosing to pursue some external/internal rewards.
I broadly agree with the premise that Gen AI has lowered the cost of software development and, as a result, the quality of the idea and its execution have become more salient differentiators than just the ability to bring an idea to reality. I also agree that pursuing one's own sense of taste/judgment or "objective function" is often intrinsically rewarding and worth it. But ultimately, an individual has to decide how much they want to (or need to) play for the external market rewards vs their own intrinsic rewards/motivations.
But I don't think it's coherent to argue that the market "ought to" reward one for their oh-so-special but mysterious, amorphous, and unmeasurable way of doing things, or that pursuing one's own taste/judgement is somehow automatically an exercise in greatness.
I think there's a coherent argument lurking in there, but the author hasn't refined it enough. Here's my attempt.
The industrial revolution seemed like the end of the artisan as an important man in his community, the death of good taste, and an apocalypse for professions like furniture makers.
And yet, do you remember the Aeron chair, darling of the dotcom era startup? The Eames chair, that wood veneer conversation piece which every real estate agent chooses when staging an apartment? In an era of mass production, taste moves up the market, but it never quite goes away.
Where does that leave our taste, dear reader? I am not sure. What I do know is that, just as the Aeron chair could not exist in a world without mass production, so too will there be tasteful software which cannot exist in a world without LLMs. I hope you're as excited as I am to build it.
I have no idea whether this was the actual intent, but it's where the disparate little facts seem like they might lead.
> In an era of mass production, taste moves up the market, but it never quite goes away.
I agree with the points in your version, and the part above is the “meat and potatoes” for me.
I too am unsure about how much did the author intend to say something similar.
The main area where I disagree with the original post is this apparent fusing of subjective/personal taste driven value and market rewards driven value.
To summarize/clarify my view using the Aeron and Eames chair examples from your version
1. I agree that someone had to pursue their taste/vision for those products even when they could’ve “gotten away” with producing something mass produced and mediocre. And that pursuing their taste/vision likely became more feasible once the technology allowed them to focus on the “taste” while taking care of the boilerplate.
2. But, I will also add that the market actually saw value in Aeron and Eames chairs and rewarded their creators. It would’ve been incongruent for those creators to assert that their chairs were superior merely because they chose to pursue their taste/vision.
I think the "taste" described here actually can be formulated as a measurable quantity, just one that can't be accurately measured in practice. According to James Shore's 2006 essay "Quality With a Name" [0], "good design" amounts to nothing more or less than minimising the expected engineering time needed to create a good enough product over the space of changes that are likely to be needed during its lifetime (I'm paraphrasing slightly). Realising this was a revelation to me. I would argue that OP's "taste" is Shore's "good design".
This conclusion may be unwelcome here, since it implies that AI can acquire "taste".
[0]: https://www.jamesshore.com/Articles/Quality-With-a-Name.html
I don't think it's entirely unreasonable to conclude AI can acquire what I have described, and would continue to defend as taste. The real question, then, could be whether its acquired taste renders you obsolete? So what if you can be replaced, what then? A question that popped into my mind is whether taste, as immeasurable as a metric as it is, would be what keeps you relevant just enough to matter.
Unpleasant as it would be, I think it may well render us obsolete in the main. There will probably always be a handful of paid human programmers around, just as there will probably always be a handful of people paid to make buggy whips, but the history of capitalism shows that most buyers don't care enough about how their goods and services get made to pay the premium that would keep "handmade" items economically viable for sellers. So I don't yet see a compelling reason why ~all of us programmers (and all other white collar workers in their turn) won't be replaced by LLMs.
The only difference I see between what happened to master craftsmen during the Industrial Revolution and what is happening to programmers today is that I belong to the latter group, so for the first time, the coming employment crisis feels personal.
Unconscious intuition you can't explain in words, thus can't reason about, as the main guiding principle in the ever more cognitively demanding world, yeah, good luck.
If I ran a dev team, I’d spend a day a week on deliberately hand coding. Even if the clankers write most of the code, IMO we need deliberate practice to develop taste. There’s a side of taste that’s Rick Rubin, how the customer experiences something. There’s also what a classically trained cellist hears in a more tactile, layered sense in how another cellist plays a piece.
This was a great and resonant post. Don't worry about the green names.
I think there is almost a split between people who have read ZAOMM and people who have not.
I recently made a post in my circles about how "meaning" is what human quality is. If you look at classical art a lot of it could look objectively nonsense and have technical mistakes, but just because the machine can make a picture does not mean that it attributes any meaning or communicate any intention. The difference between human-made classical artwork and machine-made classical artwork is the meaning and symbology behind it
Before God and as an honest man I tell you that your son is the greatest composer known to me either in person or by name; he has taste, and, furthermore, the most profound knowledge of composition.
Haydn to Leopold, 1785
https://en.wikipedia.org/wiki/Haydn_and_Mozart#The_%22Haydn%...
I'm really not convinced that the word "taste", as used in these type of AI discussions, refers to any meaningful concept. The AI that wrote this blog post (look at footnotes 2 and 3!), like many others, is great at defining what it is not and how people psychologically relate to it, but I've never seen a clear definition of what it is.
My best guess is that it's a pointer not to any actual concept but to latent semantic space between concepts. Either a very novel or very Buddhist phenomenon, depending on how you look at it.
If you need a full length blog post to explain it, it's not a good use of the word.
Speaking of taste the author basically tells me not to read the article immediately as soon as I hit the page because the commenters on this site (HN) have made them feel so bad about it
I’m pretty sure it was not technology that killed the web it was the kind of toxic crap like that, that we’re seeing from communities and that’s nothing new
It’s because the writing reads like AI slop, and the author’s note about how it’s not AI slop ALSO reads like AI slop
i type code by hand
so emacs is also left :)
My non-dominant hand is also left.
Oof, even the post-mortem promising none of it is AI slop is obvious AI slop.
I was expecting them to state that they were doing a bit with the post-mortem, but the punchline never came. It is even more consistently AI coded than the full text. Give us a /s, give us a wink a nod, give us a markdown sygil of LOSS.png, anything that says "I am actually a human", because all that reads as is BEEP BOOP KILL ALL HUMANS.
AI slop that's also simply wrong. Flagged.
Pangram classifies 100% of this text as AI-generated.
I disagree. It has none of the rhythm of AI generated text. Even humans that copy AI patterns adopt the AI rhythm at least on smaller scale.
But this follows human structure and rhythm through and through, whether on the level of sentences, paragraphs, sections, or the whole text.
I'd say it's handwritten.
The author claims no AI was used.
He's lying.
Honestly, a lot of people are denying this, but the latest models ARE at the point where their systems design generally surpasses mine because they have a better grasp of how the domain of any problem is typically solved.
Have you ever considered getting really good at one domain in particular?
Ironically, this essay is an example of at least somewhat "questionable taste". It has provoked dialogue and thus is valuable, but I think it makes a number of cognitive errors in its argument.
"Taste" is just knowledge. Advanced knowledge. Knowledge that is not easily put into words.
This is another restatement that knowledge, in a human head, is the determining factor of production, over land, labor and capital. This a post-capitalist argument that's been made for over 30 years (Peter Drucker in his 1993 book is my favorite), and it's been at least somewhat true even back to the 70s, just increasingly so as AI takes off.
The author's sidebar of AI or even cloud computing "taking ownership of the means of production away from workers" is entirely misunderstanding how the modern economy operates and shows how badly we need a new Marx (and Keynes, Schumpeter, or Drucker) since his analysis doesn't seem to provide the same analytical value in late capitalism. Renting vs. owning the means of production is now entirely a function of knowledge of the entrepreneurs (and their team).
Capital is ubiquitous now and is desperate for return - it remains powerful but is not all-powerful. We've had years of near-low interest rates and even after COVID inflation, they're still historically low. It is far more accessible to entrepreneurs than it has ever been in history, between grants, loans, subsidies, private equity, angels, crowdfunding, donations, etc. It's practically a cliche that billions in capital can be wasted on an initiative because the people involved had no idea what they were doing. I don't think AI changes that.
We don't need to rent our tokens, we can run open models on our local rigs (which still requires capital). We don't need to rent our servers, we can buy, rack and stack them on our own. Which we choose is fentirely dependent on the knowledge we have. If we don't know how to run an AI rig, we must rent it. If we don't know how to run a server, we must rent it. etc.
There are several walls driven by taste - Design - Testing - Production - Security - Scalability
In particular, the author seems confused on what "production" is. Production is still the wall. There was a joke on X the other day:
"I vibe coded my dream app! Try it out here: http://localhost:8080/coolapp"
Software running in production is not tokens. I'm sure properly prompted AI can deploy and run simple software in production but this requires a lot of knowledge to prompt to constrain it to run in a secure, scalable, adaptable way as complexity, scale, and capability grows.
[flagged]
[flagged]
It's also just patently false. Taste is not all that is left. Are software engineers really so full of hubris that they thought coding is all that there is to making a product?
> Are software engineers really so full of hubris that they thought coding is all that there is to making a product?
Frankly, this barely scratches the surface of the hubris of engineers. I personally think AI has been one of the best things to happen to software engineering. Writing boilerplate or my millionth auth implementation was never why I fell in love with the craft in the first place.
We have computers, computers can automate things. Boilerplate is a sign of a bad language (not a bad engineer). But then we add infinite automation features to reduce boilerplate and we get C++, and C++ programmers decry the excessive number of features and want to delete them all and go back to writing boilerplate by hand.
I once wrote a pile of C++ templates that would let you wrap a C++ function as a JavaScript function, automatically converting all the arguments and the return type. Then I left the company. How did my successors find it? No idea. Maybe they pulled out all the automation complexity and replaced it with manually operated boilerplate.
> Boilerplate is a sign of a bad language (not a bad engineer).
It's also a sign of refusal to learn an editor or IDE deeply. Most of the ones I know have mechanisms (or plugins) for generating the boilerplate.
Although back to the language, usually the biggest boilerplate killer are macros and so many developers seem to really have a problem with them.
This pretty much, Emacs, Sublime, Vim, Jetbrain’s IDE,… have snippet tooling solutions that are very configurable. And most unix editors support piping so you can easily write a programs for that. Or just a good old code generator.
My goal when writing a software is always to reach a point where modifying a business rule is as easy as writing it in plain english. Which means abstracting the essential complexity (while being pragmatic).
> Boilerplate is a sign of a bad language
This is true, but I was mostly referring to framework boilerplate (Spring, React, etc.) or plumbing boilerplate. If you've ever written code professionally, you know that most code that you write is just making your thing fit with someone else's thing.
If your consumer or your provider made bad engineering decisions, you have to absorb them. If your company mandates the use of a framework, you have to absorb that, and so on. So I think it's great that AI can write all the dumb shit I didn't come up with anyway.
Spewing out tons of boilerplate has always been the hallmark of a bad engineer.
What is left? The second-best architecture for any given technical problem is, in the majority of cases, still good enough.
When all else breaks, absurd() is your friend.
So i guess proven secure architecture will come and stay.
I think I am over opinion articles on AI, as a genre, in general.
Are you saying they are... tasteless?
> So let me be plain about it: Claude was not here. No LLM wrote this—not a sentence of it, nor was it outlined, drafted, reviewed, checked, etc. by one, and there is no prompt behind it either
Following form of the words I just read: these 100% animous filled HN threads about how everyone is bad and everything deserves mocker have to be some elaborate power game to undermine hope and human will, I genuinely do not understand how genuine human people could think it good to "comment" like this.
There's just such a ridiculous pile on of people with such enormous overwhelming negativity underneath this ocmment. This off-gassing is almost never engaged with the material. Now there's an excuse to write off any material we don't like instantly, to accuse it of fakeness. But I still do not see a realness enscribed upon people who so actively work to tear down & prevent thought & consideration.
To engage with the post: I do think the definition of fashion is interesting, and a real power dynamic to be aware of. But do think there's a lot of things that take off in the world because they are good and interesting and nice, especially in technology and products, that the world wants nice things, and will happily support you for doing that. And especially in tech, the technical decisions we start from and build underfoot steer us towards success in subtle compounding ways, that matter, and that taste is crucial, where fashion is not a factor.
Again, at a meta level: my distress & frustration at the top is less from seeing this comment, and far more from seeing everything down thread trying to out-do one another with how cynical and anti-meaning anti-trying they can be. But I think this post definitely started on an all too familiar note, blood in the HN waters that all too often gets us so sharky.
IMO giving "benefit of the doubt" to something beyond doubt is bad, straight up. If someone can't tell this article is AI then they need to go spend an hour a day reading known AI writing vs known human writing until they develop the skill; the voice of this author SCREAMS AI.
I'm a firm believer in not engaging with or promoting the content of misrepresented AI.
That's where you're wrong. It sounds close to AI, but it's very clearly not AI generated.
Let's look at
> So let me be plain about it: Claude was not here. No LLM wrote this—not a sentence of it, nor was it outlined, drafted, reviewed, checked, etc. by one, and there is no prompt behind it either
AI would've chosen rule of three (instead of 4 + etc), wouldn't have used the either at the end (doesn't fit into the rhythm), wouldn't have used "let me be plain" (more likely to use "let me be clear" or "let me be frank"), and wouldn't have chosen "Claude was not here".
If you've actually read a lot of essays and AI generated text (and sadly, I have), it's very obvious that this was neither generated nor rephrased by AI.
This is very much like the recent "kurzgesagt got banned by YouTube's AI detector" story, where the humans whose style was copied by AI now get accused of being AI
This post makes two mistakes.
The first is an XKCD 605 (Extrapolating). The author makes bold predictions for the future as though they're inevitable, but gives zero evidence, because there is none, we're still at T=0. IBM is still in business. Wake me up when it's been taken out by someone vibe-designing a mainframe in kicad. Or even, let's see a dropbox clone get off the ground. Should be easy, right? (It's an HN tradition, after all.)
The second mistake is mistaking a large multiplier for infinity. The cost of code has not "collapsed to almost nothing" as the author claims. Talk to your CTO/CFO if you want the real story. They're probably freaking out over AI spend. But the point is, yes, AI has made things cheaper, but they didn't become monotonically zero.
AI had also raised the bar. You can't ship "good enough" anymore, it has to be great. And great still takes time. Time in QA, time obsessing over every workflow, time talking with customers and planning features.
Taste hasn't even entered the picture, and the analysis is already wrong.
> dropbox clone get off the ground
Thats exactly what Im working on:
- web client: https://github.com/mickael-kerjean/filestash
- desktop sync: https://github.com/mickael-kerjean/fdrive
In line with the infamous top comment during their launch, it does of course support FTP as a storage, and the other way around exposing your Dropbox as a FTP server, while supporting theme so you have a similar look and feel as well: https://www.filestash.app/img/screenshots/theme_dropbox.png + tons of stuff Dropbox never did
Hey looks good. If you productize it and get subscribers that'll be awesome.
would you let taste make withdrawls from your bank account
Very well written, this gave me goosebumps. Full of flavor.
No! It’s garbage!