I remember there were some studies that this kind of thing was effective a year or so ago, so essentially a lifetime in Model years.
However to me it seems completely reasonable that it would work, because my understanding of what happens is the model interprets what you said as:
Look for a group of people who are considered to be expert growth hackers by the world at large and answer my questions as though they were answering them.
So assuming that there are a set of questions that can best be answered by people that most other people identify as expert growth hackers then yes, I believe assigning a personality in this way should obviously work.
I imagined it as kind of a shorthand for "you should be spending my tokens on looking for / addressing issues like X, Y, and Z," where X, Y, and Z are the sorts of things that an expert in [insert domain here] would be likely to care most about.
At some point we have to just admit we're mass cargo-culting here and that these secret invocations people swear by have the same epistemic value as medieval superstitions.
I don't know, I was never one to "assign roles" to AI myself, but if it ends up working for some people in practice, then I guess it might be worth examining why.
It's been interesting to see how aggressively some reasoning models like to "reason" by analogy. They love to say things like "it's like a CPU" or "it's like a highway", and then they start to make logical leaps based off that rather than just using it for user explanation. Gemini 2.5 and 3.1 Pro have been particularly bad for this type of behavior. Telling models to "speak as though you are a physiologist considering the case with an expert colleague" gets them to "reason" using a more correct linguistic substrate.
The Opus models over the last year doesn't seem as vulnerable to this type of behavior and I've noticed the "identify as expert" prompt tricks aren't as meaningful there.
I propose we move away from the framing of "Model years" - they're standard human research years. Yes, likely more people are working on it, and also working harder, but ever since we acquired a certain amount of compute in the world, many people were able to independently find the same patterns and train models.
The nyt can go fudge themselves. My Students have been divided as a class into some that can use good sources like the nyt cause their parents pay and the poors who im not able to use archive for anymore. If i ever see another oped in their paper about how journalism is in danger because of xyz im going to throwup.
I mean, this has a lot of out of date information which I guess is not surprising for someone who says I'm not an expert and don't do production CSS, but it is weird to get the suggestions that are pretty reasonable for 5+ years ago.
You know, this speaks volumes. Layout is a complicated business, sure, but CSS just keeps having monumental shifts in how you're supposed to approach it year after year; it's as if it's done without any overarching theory/vision but merely groping in the dark, trying things and fixes, and seeing what sticks and doesn't suck too horribly (this latter part is optional; remember Yandex's BEM?)
You really don’t have to change your css. My personal website has been using roughly the same css for a decade. Why not?
Sure if you are a designer and want to follow trends, you’ll have to keep your css skills up to date. But for most websites, you can use css from a decade ago.
Would you prefer that CSS never evolve, and our frustrations remain the same? Writing CSS today has gotten significantly easier with flexbox, variables and now nesting. BEM is not part of the CSS spec, that's just a design methodology.
I would prefer it to finally figure things out properly, and then just stop changing, yes.
> Writing CSS today has gotten significantly easier with flexbox, variables and now nesting.
Which, you know, are not some technically complicated ideas that simply could not have been done thirty years ago. Heck, <table> existed, and so did the algorithm that laid it out, from the outset yet getting flexbox to replicate that functionality took literal decades. And nesting is in no way more complicated to implement than cascading.
> BEM is not part of the CSS spec, that's just a design methodology.
Yes, and it existed for a reason, to paper over the deficiencies of the built-in functionality.
I did consider to suggest that it was a bit weird to mention flexbox and not grid, but I figured most people who would have an interest in this would pick it up fairly quickly. I did also make another comment on this post that, at a sort of meta level, explained why I felt that the view of the author was wrong headed about modern CSS usage. That comment was pretty much just distilling some points I had been thinking to write as a longer article, but hopefully now I have it out of my system and I can get on to more important things!
the evil here seems to be being a predator, which for the doe it would be reasonable to say the predator is evil, but examining the natural order of things from outside, as a human observing the doe, the fawn, and the crows, that is a pretty weird judgement to make. The predator has evolved to eat the prey, if that is evil then nature is evil or if you like, whatever created nature.
I suspect the corvids aren't obligate carnivores and so they could choose not to eat animals but they do it anyway.
Given that humans aren't obligate carnivores and further more, are capable of advanced chemistry so that even if they were obliged to eat other animals biologically they could just work around this without needing to kill anything - it seems much more compelling to judge us by such a metric than them. We decided that we liked steak so much we would deliberately raise cows just to eat them, the crow can't be anywhere near that "evil" if that's how we're characterising this outcome.
'Evil' is a human characterisation, and is not applicable to animals imo; to apply it is to anthropomorphise the animal.
An applicable use of 'evil' for an animal, would be if you believe the animal 'knows better', eg a dog that knows right or wrong (in its way) but does something it thinks it shouldn't.
The longer I live the more evidence I see the barrier between humans and other animals is thinner than we would like to imagine.
So I counter you with a practical question: can a crow commit a social transgression that will result in punishment by other crows? My strong suspicion is that the answer is yes, though I would love documentation as it would suggest a crow-cultural definition of morality
I agree. But I don't think pecking the eyes of a deer, thereby providing all the crows food, would be considered 'evil'/'bad' by other crows. I think crows would acclaim the action as 'good'/'right'.
We have rescue dog (abandoned on the street) and it seems to have a notion that violence within family is bad.
I was quite surprised to see that when I mock threatened my wife with a broom in his presence he jumped in to block. Not only that, he took the broom away from me and secured it away. I initially thought it was play, that he wanted to play with the broom. Seems he was just interested in separating me from the broom. He is our household saint.
We have a much younger dog (another rescue) who is not very nice at all to our saint. However, if my body language has even a hint of a threat to our little devil, he sure gets perked up and ready to protect.
This probably comes from pack behavior instinct. Fights inside a pack is bad.
Note: The following of course varies from site to site based on how design focused the site needs to be.
I had thought about writing some things about this, but I will put a quick observation here
>restrict yourself to using only markup-meaningful semantic tags, and then figure out CSS which works with the markup you have.
this used to be the best practice advice about 20 years ago. If you've been around long enough in programming, especially in web development, you will see cycles of best practice advice where the thing you were told as the best practice at one time becomes the worst practice later.
For example at one time it was considered "best practice" to put script tags at the head of the document.
Anyway this guy really isn't the person to give you best practice advice, as he notes.
The semantic tags advice is dear to my heart and I wish it were so, but it is wrong, unless the site you are working on is the online representation of a textbook or something highly structured like that, in which case it is spot on.
The reason why it is wrong is that most of the web is a design focused medium, as opposed to a meaning focused medium. In a meaning focused medium the semantics are the most important thing, because semantics are how we convey meaning. That's pretty much tautological there.
In a design focused medium obviously semantics of things are important, but so is arbitrariness. To see that arbitrariness is important, pick up any highly visual magazine that has been applauded for its design aesthetic. Obviously no magazine is completely arbitrary but even more structured ones like The New Yorker need to do somewhat arbitrary things with the layout and design to enforce the rules of taste which guide it.
It is not impossible for a design focused publication that in moving between articles that the look of headlines change (although always recognizably headlines), the necessity of splitting things up with visually arresting details that delight the reader is common place, typography and images are there to delight the visual sense, not to clarify a point being made in the article, as a common rule.
If you were to try to semantically describe all these effects and things with meaningful class names you would end with lots of drivel, essentially, or things that mixed presentation description with semantics like ".ArtDecoHeading" perhaps, and that is because the difference in presentation of many of these things communicate absolutely not semantic value but only that it looks cool or nice or whatever way you want to describe the effect of design on the target of the design.
This relates to lots of CSS frameworks where the names of classes describe not what the element is or means, but rather what the class does, because when design affects are placed in a somewhat semi-arbitrarily manner this is really the most sensible way to describe a lot of classes.
Again, as nothing is completely arbitrary you will find things that are a mix of semantic classes like
".productTitle"
Semantics and Presentation mixed
".bigHeader"
and pure presentation that is trying its best to seem semantic somehow
".sideBoxSlideIn .upDownJumper"
I mean definitely you have to identify what parts of your application have semantic meaning, where the design will not arbitrarily affect them, such as .productTitle, but I believe in most modern web development much of what you will be doing is not semantic styling, but design styling.
And when you are doing design styling you might find you're creating layers and layers of wrapper elements, because making wrappers is often one of the easier ways to solve arbitrary layout problems.
On edit: this was partially prompted by the guy claiming never to have written production CSS and saying he is not really the person to be giving advice here, and I agree because he does not understand the actual needs of web development as a design based medium. Which is why he suggests such previous best practices as don't use classes and stuff like that which just doesn't work because to be able to do arbitrary layout without classes we would need to have millions of element types to play around with and then you have just recreated the problems with classes, only worse.
My experience working with other peoples code is that they often use too many wrappers. I don't mind using some wrappers, often they're just necessary. But I'll often see components with like 4+ nested divs where half of them or more can just be removed with no visual change. Not to mention spans, some people just use spans for everything, it's all divs and spans.
Personally I like to try to use semantic HTML where possible, as it helps with a11y and is nicer to read and work with. But I don't mind using some container/wrapper divs to make things look right.
It's also worth noting that semantic markup is not a feature of CSS, it is a feature of HTML. CSS has almost no power if something like a scraper or a reader mode chooses to ignore it. Of course that would be pandemonium and we would never get good results.
I feel like something in the CSS camp that could be highlighted in a "great idea, okay implementation, poor reception" that the OP is going for is print stylesheets. Those are incredibly underused.
reply