Taste Is All That's Left
18 points by pbsds
18 points by pbsds
As I read this, I was reminded of two quotes. The first is about the taste gap from Ira Glass:
Nobody tells this to people who are beginners, I wish someone told me. All of us who do creative work, we get into it because we have good taste. But there is this gap. For the first couple years you make stuff, it’s just not that good. It’s trying to be good, it has potential, but it’s not. But your taste, the thing that got you into the game, is still killer. And your taste is why your work disappoints you. A lot of people never get past this phase, they quit. Most people I know who do interesting, creative work went through years of this. We know our work doesn’t have this special thing that we want it to have. We all go through this. And if you are just starting out or you are still in this phase, you gotta know its normal and the most important thing you can do is do a lot of work. Put yourself on a deadline so that every week you will finish one story. It is only by going through a volume of work that you will close that gap, and your work will be as good as your ambitions. And I took longer to figure out how to do this than anyone I’ve ever met. It’s gonna take awhile. It’s normal to take awhile. You’ve just gotta fight your way through.
The second is from The Ze Frank Show:
For a very long time, taste and artistic training have been things that only a small number of people have been able to develop. Only a few people could afford to participate in the production of many types of media. Raw materials like pigments were expensive; same with tools like printing presses; even as late as 1963 it cost Charles Peignot over $600,000 to create and cut a single font family.
The small number of people who had access to these tools and resources created rules about what was good taste or bad taste. These designers started giving each other awards and the rules they followed became even more specific. All sorts of stuff about grids and sizes and color combinations — lots of stuff that the consumers of this media never consciously noticed. Over the last 20 years, however, the cost of tools related to the authorship of media has plummeted. For very little money, anyone can create and distribute things like newsletters, or videos, or bad-ass tunes about "ugly."
Suddenly consumers are learning the language of these authorship tools. The fact that tons of people know names of fonts like Helvetica is weird! And when people start learning something new, they perceive the world around them differently. If you start learning how to play the guitar, suddenly the guitar stands out in all the music you listen to. For example, throughout most of the history of movies, the audience didn't really understand what a craft editing was. Now, as more and more people have access to things like iMovie, they begin to understand the manipulative power of editing. Watching reality TV almost becomes like a game as you try to second-guess how the editor is trying to manipulate you.
As people start learning and experimenting with these languages authorship, they don't necessarily follow the rules of good taste. This scares the shit out of designers.
I'm not sure what I make of all this.
Between reading Zen and the Art of Motorcycle Maintenance, watching The Devil Wears Prada 2, and then reading this article, the following thought triggered:
After industrialization, the bulk of clothing is now produced without a human involved. This results in a lot of second-rate design, which yet derives one way or another from haute couture, where there is only room for real taste and original ideas. In haute couture, the best designers notoriously learned the hard way: got taught embroidery by their grandparents, apprenticed in rusty loom factories, inherited generations-old traditions of tailoring and transcended them, etc. They climbed a wall and experienced the true nature of their craft, which makes them stand out.
Once making good-enough software is truly as automated as making good-enough clothes, maybe software will operate the same as fashion? Maybe software developers with decades of manual programming under their belts will create the boutique software that defines all mediocre software out there?
I am not sure what I make of it all either, but I keep landing on the same spot. Thank you for the insightful comments :)
This feels like another thing that a proponent of AI thinks that the opposition must believe. They certainly completely lack in taste by definition, but this submarines the problem of instantaneous function or suitability, which it also lacks by definition.
Do excuse me for I cannot be sure how to parse this comment correctly. If I am reading it correctly, you seem to think the post is an AI-booster's strawman of the skeptic position, and that by framing the loss as taste I quietly bury the stronger objection and that the output often doesn't function or isn't suitable? If this is what your point is, then I would like to push back.
I am not an AI-booster, and I am not carrying water for the machines here. My goal is not to argue LLM output is reliable, or in stark contrast, unreliable but such is my observation: they are generally accepted as passable---whether I like it or not. Though, you are right. I granted "the output is good enough" and then argued that only taste is missing. That grant does enormous work for me. Guilty, I apologize.
In appreciation, I would like to amend the title with a small asterisk: taste is all that's left once the thing actually works. Please feel free to let me know if I misunderstood, and thank you for the fair criticism nevertheless.
argued
I think your pushback is superb because you think it is worthwhile to describe reality as it is, thanks ! Sorry, and thanks. It was more meant to illustrate my own myopia as a general message and shouldn't have perhaps so fully assumed an intent. I think this is the frontier of the hogwash, frankly, to make all arguments keep accepting further encroaching premises.
Between reading Zen and the Art of Motorcycle Maintenance, watching The Devil Wears Prada 2, and then reading this article, the following thought triggered:
After industrialization, the bulk of clothing is now produced without a human involved. This results in a lot of second-rate design, which yet derives one way or another from haute couture, where there is only room for real taste and original ideas. In haute couture, the best designers notoriously learned the hard way: got taught embroidery by their grandparents, apprenticed in rusty loom factories, inherited generations-old traditions of tailoring and transcended them, etc. They climbed a wall and experienced the true nature of their craft, which makes them stand out.
Once making good-enough software is truly as automated as making good-enough clothes, maybe software will operate the same as fashion? Maybe software developers with decades of manual programming under their belts will create the boutique software that defines all mediocre software out there?
I love this comment. One of the many insightful ones that I think are in good faith. In part because it is close to where the post tries to land; Morris and Ruskin watching the power looms and asking not "can we" but "should we." Fashion is the case where that question got answered and both answers came out true at once.
Coincidentally this reminds me on a philosophy class I have taken in college that I vaguely remember. Something that I will add from what I can remember is that the machines won the volume, the couture houses kept the taste, and the second defines the first without ever outselling it.
In short, I think the boutique-software future is plausible indeed. More importantly, I would like to live in it.
Author lacks taste in prose sufficient to recognize that this sounds distinctly like slop. Flagged as spam.
I was not sure whether or not to reply, but I decided to. I am sorry that I have allowed you to feel this way. Be assured that genuine effort goes into all of my posts, and that no LLMs were present in its writing/storyboarding/review/etc.
After reading this reply, I read the addendum, and then read a few of your other posts. I will say that many others very much do not sound like slop; when I compare them using the various AI detectors [1], posts from early 2025 say 10% or 0% and newer ones, including this one, say 100% AI or "AI paraphrased." That shift seems to have taken place, to my eye and to those of the detectors, between "The Federation Fallacy" (June 26, 2025) and "The Tradeoff Trap" (August 10, 2025).
I'm really torn: on the one hand, you seem very genuine, and I'd really like to believe that there were no LLMs involved. On the other hand, there are just so many of these tells present, from jarringly different sentences composed of three or four words in the form "<noun> <verb> [article] <noun>" to the little thinking-outside-of-thinking phrases like "I want to be careful here", "Here is the uncomfortable mechanism", and "[H]ere is the defense, such as it is" that emerge from the mechanism of LLMs' relentless forward march of token completion, that it's hard for me to believe.
If it's really true that the machines write like you, I'm sorry. That's awful. With that said, I do think that writing well these days means taking the time to make clear, through both careful editing and clear sentence construction, that what we put out doesn't sound like the machine -- and I wonder what happened about a year ago that made you stop doing that.
1: To me, the comparison is much more interesting than the raw number. I understand that simply reporting "GPTZero says AI" is mostly meaningless.
Thank you for taking the time for an in-depth comment, and the analysis of it all. This is by far one of the most insightful comments I have received, and I want to carefully respond to each one of your points.
when I compare them using the various AI detectors [1], posts from early 2025 say 10% or 0% and newer ones, including this one, say 100% AI or "AI paraphrased." That shift seems to have taken place, to my eye and to those of the detectors, between "The Federation Fallacy" (June 26, 2025) and "The Tradeoff Trap" (August 10, 2025).
AI detector or not, I have been feeling the same drop in quality and insight in my posts and I, well, can't pretend there hasn't been a shift. I'd be lying to you and to myself if I had said I didn't see it. Consequently, I have re-read my post after it hit Hacker News and a friend kindly informing me of it. I'll tell you this much: the only thing I had in my mind was "how the FUCK did I write like this". Hilariously, I was thinking "I have not written anything with real depth in a while" to myself earlier this week.
If it's really true that the machines write like you, I'm sorry. That's awful. With that said, I do think that writing well these days means taking the time to make clear, through both careful editing and clear sentence construction, that what we put out doesn't sound like the machine -- and I wonder what happened about a year ago that made you stop doing that.
I have spent a good while contemplating whether it was the machines that write like me, or if I write like the machines. I hesitate to say I think it is a combination of both. What I did do---without ever clocking it as a decision---was spend a year reading the things and working next to them daily, until their "cadence" was part of my mental archive.
Again, I appreciate your comment. I wrote the post with good, albeit selfish, intentions but it was never the goal to drop unfiltered LLM slop in the timeline and wait for clout. That goes against all the reasons I write. Admittedly I am embarrassed to have produced such a piece. Now that I think about it, I've read enough of the averaged voice that some of it has settled into mine, and the tells you can name are ones I reach for practically unprompted now---like the way you pick up an accent without deciding to. Yes I mentally filter out the typical tells like "This isn't X, this is Y" and the many like it, but clearly some have made it into my lexicon.
Like many comments such as yours, the care and the concern have been a wake-up call. I started working on a second post---partially inspired by this one---and if you ever get around to reading it then I leave it to you whether it reads as mine or as the machine's. With all of those said, I can't help but read my own writing with a certain level of paranoia after the fact. Much I should think about. Still, I leave you with my thanks.
I appreciate your taking my comments in stride and I wish you all possible success in freeing your mind from the machines' yoke. Unflagged, and I have subscribed to your RSS feed. 💜
I got the same treatment from this community on a recent post. Don't apologize for your language. This is not constructive feedback. I hope this is not a new trend in this community. Some kind of LLM witch trial. "It floats on water like a duck, must be a witch!"
I'm sorry, but I just struggle accepting your claim. And to explain the reactions you're getting, it's like this: imagine that someone rings your doorbell 100 times and every time, you find flaming poo on your doorstep. You're not going to give it the benefit of the doubt when the doorbell rings for the 101th time, right?
Here's what immediately makes me very skeptical about your claim: (1) I can't find your pre-LLM writing that would be comparable in style or verbosity; (2) I see a wide number of consistent, high-level tells in the text (and I don't mean em-dashes); (3) your essay scores 100% AI on Pangram, which is a remarkably good tool (in contrast to the crappy detectors of the old); (4) your essay doesn't seem to be saying anything beyond what I'd expect an LLM to produce in response to a short prompt - the "thesis compression" test.
I'll admit that my judgment is colored by the fact that it's a peculiar affliction of AI-related op-eds (both pro- and con-) that they appear to be AI-generated at a much higher rate than any other form of blogging on the internet. Their authors often show up to assert that they haven't used AI tools, but in almost all instances, these claims are difficult to take seriously. I remember debating someone who had no online writing footprint whatsoever until 2025 and then started a blog that posted one long-form piece every single day at the same hour. Is there a one-in-a-million chance I was talking to a pseudonymous programming blog prodigy who works on a really precise schedule? Yes, but you gotta draw the line somewhere.
And it matters to me because if I'm to spend human time reading, reflecting on, and engaging with a post, I want to engage with content where another human actually put in the effort to articulate their own thoughts. I don't want to engage with an endless supply of machine-generated posts. There's just no point.
On the off chance you're actually being honest with us - it sucks, but I think you should direct your ire toward the flaming-poo pranksters who came before.
Everyone can make anything. Almost no one can tell you what is worth making.
Whilst I agree with the sentiment, it feels like quite a narrow view. I think it's possible to argue that in actual fact AI has taken very little from us. To paraphrase Graeber's Bullshit Jobs: AI mostly only amplifies pre-existing bullshitery. The signal is still there, just as loud and clear as ever, but the noise has changed, that's all.
It was always hard to find meaningful and fulfilling work, whether paid or personal. Maybe it felt like we had it a bit easier pre-AI, in the sense that yes, not everyone could make anything. Us geeks could do the monkey work and get well paid or receive open source kudos. But if that's all we ever contributed, then of course AI looks like a usurper.
Computers are hard for so many more reasons than the mere act of making things. Consider the humanity. I cannot express my gratitude for the warm smiles of colleagues on a dreary Monday morning, or the patient and considered comments from a stranger on a tricky Github Issue. Not only is that still left after AI, it hasn't even taken a dent.
Then there's the whole issue of copyright. Putting aside the legality of AI training, the morality of attributing our sources is still left after AI. I care about who makes my software. I value their time and wish for them to be recognised for their efforts. AI doesn't stop me having or expressing those feelings.
And of course the environmental impact of data centres. The deepening of the socio-political capitalistic violence of the concentration of wealth from AI investment. At the very least we still have the choice to be conscious of those things, and ideally we make a stand against them with our wallets and voices.
To say that taste is all that's left feels either superficial or like giving up. Keep your head up! If you value good, true work, if you care about other humans, if the craft inspires you, then it's all still up for grabs. In fact this is engineering: navigating unprecedented circumstances with patience, creativity and kindness. To state the hacker credo:
Do what you can, with what you have, where you are.
This is overall a pretty good blog, but to be honest I think the analogy towards 'if this is worth making at all' makes it seem like the origins of modern art movements did not start supported from a understanding that patronage of the arts (an argument of how this is decided could be made) was something that those that benefited from investment would do - such as Millas etc.
You could even argue this was important in the application of new art with El Lissitzsky in Eastern Europe as industrialisation would start to begin.
All this to say... is there really structures to support this? This ivory tower that is respected and looked towards for what is the next idea?
I don't see it just yet - I think often we refer that to genuinely amazing maintainers of our most crucial software that has been publicly scrutinised for decades, but I would not say that is new ideas as much as raw skill... to stretch the analogy, its like saying Dutch masters were crucial to the modern art movement, when you have identified more discrete artists had to define the modern art idea in a movement without much economic output for a very long time.
All this to say, I don't know if I see any incentive to create this apart from the moral argument from those that can't afford the level of patronage for those fighting the good fight, which is maybe the cynic/nihilist in me hoping this will evolve as a necessity in the future.