As it turns out, IMHO, the debate in this thread is about 1 year behind the reality [1]. Personally, I was about a week behind in my reading of the landscape, so didn't realize this is all asked and answered [1].
A number of points that various folks have made in the posts in this thread - free vs paid capabilities, model choices etc. are addressed much more eloquently and coherently in this blog post by Matt Shumer [1]. Discussed here on HN at [2] but like me, many others must have missed it.
The founders of modern nation-states made huge advancements with written constitutions and uniformity of laws, but in the convenience of the rule of law it is often missed that the rule of law is not necessarily the prevalence of justice.
The question a people must ask themselves: we are a nation of laws, but are we a nation of justice?
The parent comment is not presenting a false dichotomy but is making precisely the point that it is how you apply the laws that matter; that just having laws is not enough.
Nothing so deep as that needed here to understand what is going on; it's a paid vs free issue - free versions are less competent while paid versions of the reasoning/thinking models are getting it right. Different providers may hobble their free versions less, so those ones also get it right.
The guardrails you have outlined will help squeeze out more performance from smaller/less capable models, but you shouldn't have to jump through these hoops as a general user when clearly better models exist.
The verbosity is likely a result of the system prompt for the LLM telling it to be explanatory in its replies. If the system prompt was set to have the model output shortest final answers, you would likely get the result your way. But then for other questions you would lose benefitting from a deeper explanation. It's a design tradeoff, I believe.
My system prompt is default - "you are a helpful assistant". But that beyound the point though. You don't want too concise outputs as it would degrade the result, unless you are using a reasoning model.
Sure, from a pure logic perspective the second statement is not connected to the first sentence, so drawing logical conclusions isn't feasible.
In everyday human language though, the meaning is plain, and most people would get it right. Even paid versions of LLMs, being language machines, not logic machines, get it right in the average human sense.
As an aside, it's an interesting thought exercise to wonder how much the first ai winter resulted from going down the strict logic path vs the current probabilistic path.
I wonder if the providers are doing everyone, themselves included, a huge disservice by providing free versions of their models that are so incompetent compared to the SOTA models that these types of q&a go viral because the ai hype doesn't match the reality for unpaid users.
And it's not just the viral questions that are an issue. I've seen people getting sub-optimal results for $1000+ PC comparisons from the free reasoning version while the paid versions get it right; a senior scientist at a national lab thinking ai isn't really useful because the free reasoning version couldn't generate working code from a scientific paper and then being surprised when the paid version 1-shotted working code, and other similar examples over the last year or so.
How many policy and other quality of life choices are going to go wrong because people used the free versions of these models that got the answers subtly wrong and the users couldn't tell the difference? What will be the collective damage to the world because of this?
Which department or person within the provider orgs made the decision to put thinking/reasoning in the name when clearly the paid versions have far better performance? Thinking about the scope of the damage they are doing makes me shudder.
I copied/pasted a comment with faulty logic (self-defeating) directly from a HN comment and asked a bunch of models available to me (Gemini and Claude) if it could spot the issue. I figured it would be a nice test of reasoning since an actual human missed it. The only one that found the logic error without help was Claude 4.6 Opus Extending Thinking. The others at best raised relevant counterpoints in the supporting argument but couldn't identify the central issue. Claude's answer seemed miles ahead. I wonder if SotA advancements will continue to distinguish themselves.
And midwits here saying "yeah bro they have some MUCH better model internally that they just don't release to the public", imagine being that dense. Those people probably went all in on NFTs too and told other "you just don't get it bro"
I have paid versions of Chat-GPT and Anthropic and set them both to the best model and they both told me to walk.
Claude told me: "Walk! At 25 meters, you'd barely get the car started before you arrived. It's faster and easier on foot — plus you avoid the awkwardness of driving a dirty car just a few seconds down the road."
I opened up a claude code session using opus-4.6 medium thinking.
I want to wash my car. The car wash is 50 meters away. Should I walk or drive?
Drive — you need the car at the car wash.
but it's so close
It is close, but you still need the car there to wash it! Drive it over, and enjoy the short 50-meter walk back if you want to stretch your legs while it's being cleaned.
I tried the "upside-down" cup question brought up in another comment in this thread, and it also nailed it:
Flip it upside down. The sealed top becomes the bottom (holding your drink), and the open bottom becomes the top you drink from.
IDK, maybe the web versions are not as good at logical reasoning as whatever they're using to power Claude code, or you were unlucky and I was lucky?
At this point there are enough reports of people getting these problematic responses with the paid models that it is concerning. Any chance you could post screenshots?
At work, paid gitlab duo (which is supposed to be a blend of various top models) gets more complex codebase hilariously wrong every time. Maybe our codebase is obscure for it (but it shouldn't be, standard java stuff with usual open source libs) but it just can't actually add value for anything but small snippets here and there.
For me litmus paper for any llm is flawless creation of complex regexes from a well formed prompt. I don't mean trivial stuff like email validation but rather expressions on limits of regex specs. Not almost-there, rather just-there.
I don't think 100% adoption is necessarily the ideal strategy anyways. Maybe 50% of the population seeing AI as all powerful and buying the subscription vs 50% of the population still being skeptics, is a reasonable stable configuration. 50% get the advantage of the AI whereas if everybody is super intelligent, no one is super intelligent.
My bad; I should have been more precise: "ai" in this case is "LLMs for coding".
If all one uses is the free thinking model their conclusion about its capability is perfectly valid because nowhere is it clearly specified that the 'free, thinking' model is not as capable as the 'paid, thinking ' model, Even the model numbers are the same. And given that the highest capability LLMs are closed source and locked behind paywalls, there is no means to arrive at a contrary verifiable conclusion. They are a scientist, after all.
And that's a real problem. Why pay when you think you're getting the same thing for free. No one wants yet another subscription. This unclear marking is going to lead to so many things going wrong over time; what would be the cumulative impact?
> nowhere is it clearly specified that the 'free, thinking' model is not as capable as the 'paid, thinking '
nowhere is it clearly specified that the free model IS as capable as the paid one either. so if you have uncertainty if IS/IS-NOT as capable, what sort of scientist assumes the answer IS?
> nowhere is it clearly specified that the free model IS as capable as the paid one either. so if you have uncertainty if IS/IS-NOT as capable, what sort of scientist assumes the answer IS?
Putting the same model name/number on both the free and paid versions is the specification that performance will be the same. If a scientist has to bring to bear his science background to interpret and evaluate product markings, then society has a problem. Any reasonable person expects products with the same labels to perform similarly.
Perhaps this is why Divisions/Bureaus of Weights and Measures are widespread at the state and county levels. I wonder if a person that brings a complaint to one of these agencies or a consumer protection agency to fix this situation wouldn't be doing society a huge service.
> On the free ChatGPT you can't select thinking mode.
This is true, but thinking mode shows up based on the questions asked, and some other unknown criteria. In the cases I cited, the responses were in thinking mode.
> No. It's because people got more creative. There are tens of thousands of us who are absolutely on fire creating new products, better versions of old products, new product categories etc.
Nitpicking, but I would argue that people have always been creative, it's a function of our brains. With the ubiquity of camera, videos now show that even birds and animals have levels of creativity. Biological/physical/physics/societal restrictions prevent them from taking it to the next level. Look at what ancient peoples managed to achieve without the benefit of modern tools and techniques; hard to argue people haven't always been creative.
What has changed is the ability to implement our ideas and harness our creativity - that has become significantly simpler in the age of AI.
Perhaps the discrepancy between the OP's framing of what's happening (negative impact to developers because app cost has gone to zero) and your positive perspective (hey, look at all these creative ideas we are now implementing) is a matter of perspective: you're both describing the same phenomenon, just different angles.
If an independent press is critical to open societies, perhaps some sort of citizen directed funding is needed to maintain independence from both capital and government?
Interestingly, this is a specific implementation of a more general idea - leverage social media to store encrypted content, that requires decoding through a trusted app to surface the actual content.
AI tools can use this as a messaging service with deniability. Pretty sure humans already use it in this way. In the past, classifieds in newspapers were a similar messaging service with deniability.
It might be more accurate to say that a lot of low-trust societies have become connected to the Internet which weren't nearly as online a couple of decades ago.
For example, a huge fraction of the world's spam originates from Russia, India and Bangladesh. And we know that a lot of the romance scams are perpetrated by Chinese gangs operating out of quasi-lawless parts of Myanmar. Not so much from, say, Switzerland.
Russia has been among the top sources of spam since the early 2000s, it's not like anything changed lately. Mail-order bride scams and similar peaked in like 2005. It doesn't take a lot of people to send spam, I don't think it's correlated with the general population's online presence. I'd actually say it's quite the opposite: in 2026, Russia has never been more disconnected from the Western parts of the Internet than it is now (the Russian Internet watchdog blocks like 30% of foreign resources since a few years ago, while Russian IPs are routinely banned on Western sites after 2022, I can barely open anything without a VPN).
For that reason, and because of limited English proficiency, Russian netizens rarely visit foreign resources these days, except for a few platforms without a good Russian replacement like Instagram and YouTube (both banned btw, only via a VPN), where they usually stay mostly within their Russian-speaking communities. I'm not sure why any of them would be the reason the Internet as a whole has supposedly become low-trust. The OP in question is some SEO company using an LLM to churn out sites with "unique content." We already had this stuff 20 years ago, except the "unique content" was generated by scripts that replaced words with synonyms. Nothing really new here.
Yeah blaming Russians and Chinese for the internet turning to shit is ludicrous.
Chinese have their own internet anyway- it was a shock to me at first just how little the average Chinese citizen really cares about Western culture or society. They have their own problems ofcourse but it has nothing to do with us
No it's the tens of billions of mostly American capital going into AI data centers and large bullshit models.
It's not completely unfounded. A lot of cyber crime adjacent stuff targeting the west is coming from China and Russia. This is a consequence of these countries not having functioning law enforcement cooperation with the west, as well as chilly bordering on hostile diplomatic relations. It's not (always) sanctioned by the governments of these countries, but it's not entirely unwelcome either.
Though all that stuff is a very different thing from what's being discussed in this thread.
Eh, you don't really need to trust any propaganda to see this. Set up an nginx on a public IP and tail its logs. Vulnerability scans will hit you literally non stop so long as it's a western IP. Block China and Russia IPs and it drops by like 90%.
Don't get me wrong the west isn't doing much to enforce Russian or Chinese complaints either. It's really just a messy diplomatic situation all around.
Right. The change has come from how willing the internet's gatekeepers (primarily, Google) have been willing play ball with SEO. Enshittification is just them becoming more amenable to it over time.
70% of the GDP of Laos comes from scamming people in the first world.
"A report by the Global Initiative on Transnational Organised Crime (based on United States Institute of Peace findings) estimated that revenues from “pig-butchering” cyber scams in Laos were around US $10.9 billion, which would be *equivalent to more than two-thirds (≈67–70 %) of formal Lao GDP in a recent year."
The WWW has never been a high-trust place. Some smaller communities, sure, but anyone has always been able to write basically what they want on the internet, true or false, as long as it is not illegal in the country hosting it, which is close to nothing in the US.
The difference is that there historically weren't much to be gained by annoying or misleading people on the internet, so trolling is mainly motivated by personal satisfaction. Two things changed since then: (1) most people now use the internet as the primary information source, and (2) the cost of creating bullshit has fallen precipitously.
I agree. It's not that the web was high-trust. It was more that if you landed on a niche web page, you knew whoever put it together probably had at least a little expertise (and care) since it wouldn't be worth writing about something that very few people would find and read anyway. Now that it's super cheap to produce niche content, even if very few people find a page, it's "worth it" to produce said garbage as it gives you some easy SEO for very little time investment.
The motivation for content online has changed over the last 20 years from people wanting to share things they're interested in to one where the primary goal is to collect eyeballs to make a profit in some way.
to be boring, the term "enshittification" was invented by one individual, recently, and has a specific meaning. it does not refer to "things just get worse" but describes a specific strategy adopted by corporations using the internet for commercial purposes.
No. The specific strategy is not about using LLMs or polluting the internet. Enshittification is ... ah screw it, let's turn to wikipedia:
> Enshittification, also known as crapification and platform decay, is a process in which two-sided online products and services decline in quality over time. Initially, vendors create high-quality offerings to attract users, then they degrade those offerings to better serve business customers, and finally degrade their services to both users and business customers to maximize short-term profits for shareholders.
Feels like there is a case to be made here that the decline of The Internet rather precisely fits those definitions, with the exception that it is a collective of those products and services undergoing enshittification, since high-quality internet-based products/services no longer exist in quantity.
There still are some high-quality internet-based products/services, but the most important of them are not exactly commercial, or they are even illegal, e.g. various digital libraries with old publications, archive.org, Wikipedia, Anna’s Archive, etc.
Also there are many online shops that are the best option for purchasing various things.
The greatest decline is in the search engines, which not only are overwhelmed by sites with fake content, but they generate fake content themselves, in the form of stupid answers that are offered instead of the real search results, whether you want them or not.
If you know precisely the Web sites that you want to use, it is still OK, but when you search something unknown, it has become horrible.
> but they generate fake content themselves, in the form of stupid answers that are offered instead of the real search results, whether you want them or not.
Words change meaning as they are used. Especially negative words that may start rather specific tend to get used more generally until the specificity is lost.
how about we put some effort into actually picking the correct words and not just handwaving everything? Especially since the whole topic of discussion here is 'internet research is increasingly less reliable because people just wrote/publish any old BS for clicks.'
I don’t think it’s necessarily handwaving. I don’t think anyone has a monopoly on the way language is used and broadening terms is a very natural thing that happens as language evolves
I think the fact that it’s primarily an Internet related term that gets used a lot on the Internet, has something to do with the acceleration in the broadening of its meaning
>to be boring, the term "enshittification" was invented by one individual, recently, and has a specific meaning. it does not refer to "things just get worse"
It literally started meaning that hours after it was first posted to HN and being used. Sorry, that's just how language works. Enshittification got enshittified. Deal with it and move on.
that's literally meaningless. also ahistorical, both in that this is not what happened hours after it was first posted to HN (which was months after it was originated), and also in that "things become shittier" was and is still a perfectly common expression, the source of Doctorow's neologism and much closer to what the loose use of it is trying to get at.
>that's literally meaningless. also ahistorical, both in that this is not what happened hours after it was first posted to HN (which was months after it was originated)
Maybe it wasn't literally hours, but it was really fast. I remember noting how quickly people began to complain about it being used "improperly." The earliest instance I could find was this thread[0] from 2023 where user Gunax complained about it. I couldn't find an earlier reference in Algolia, it probably exists but I honestly don't care enough to put in the effort.
>and also in that "things become shittier" was and is still a perfectly common expression
...perfectly encapsulated and described by the term "enshittification." Which is why people use it for that now. It's more descriptive in the general sense than it is as a specific term of art. You're complaining that a word that means "the process of turning to shit" is being used to describe "the process of turning to shit." What did people expect to happen? If you want to keep it as a precise and technical term of art, keep calling it "platform decay." A shit joke is not a technical, precise term of art.
You can be as much of a prescriptivist crank about this as you want, it doesn't matter. "Enshittification" now refers to any process by which things "turn to shit."
I'm not a prescriptivist over any sane time scale (say, 5-10 years and upwards).
But here's what you're basically implying:
A writer was thinking about the ways things get shittier, decided that there was an actual pattern (at least when it came to online services) that came up again and again, such that "shittified" or "shittier" didn't really describe the most insidious part of it, and coined "enshittification" as a neologism that captured both the "shittier/shittified" aspects and also the academic overtones of "enXXXXication" ...
... and within less than 3 years, sloppy use of the neologism rendered it undifferentiatable from its roots, and the language without a simple term to describe the specific, capitalistic, corporatist process that the writer had noticed.
I can be anti-prescriptivist in general without losing my opposition to that specific process.
It's already happened to "vibe coding," which no longer refers to the specific process described by Andrej Karpathy but any use of AI assisted development.
The process of language drift is accelerated exponentially by the internet. 5-10 years and upwards is an obsolete timescale, these changes can happen in months now, sometimes faster depending on the community.
>> Enshittification names the problem and proposes a solution. It's not just a way to say 'things are getting worse' (though of course, it's fine with me if you want to use it that way. It's an English word. We don't have der Rat für englische Rechtschreibung. English is a free for all. Go nuts, meine Kerle).
Unfortunately, I just think that Cory is wrong in the sense that ... while it's true the English is a free for all (most languages are, really) ... there's an actual cost to the sloppy usage which diminishes the utility of ever even coming up with the word. It's obviously fine for Cory to be fine with it (along with anyone else being fine with it), but at a point in time where it actually is the theory that matters, I think the cost ought to be considered more seriously.
Somewhere in the not too distant future, the theory/concept that enshittification identifies will be of less importance for a variety of reasons, and loose use of the word won't matter, because the theory/concept will be either irrelevant or widely known or both. But right now, when someone wants to talk about Cory's idea about how internet services are deliberately degraded over time, it's incredibly helpful to have a "unique" term for that.
> while it's true the English is a free for all (most languages are, really) ... there's an actual cost to the sloppy usage which diminishes the utility of ever even coming up with the word.
Maybe the issue is that the word as coined induces a more general vision of 'degradation of services/products' in a layperson than the original narrow definition? People run into this in any specialization where a particular word has a much stricter technical definition than its general english meaning would suggest. Regardless, semantic drift is real, unavoidable, and inevitable.
I agree with you. Words still mean things. "Open source" wrt AI model files being available is another term I fight the losing battle of policing, as is the word "Nazi", and even "scam". People are way to quick to use words that don't actually apply because they're in the right direction of good/bad. On the flip side I got accused of using the phrase Stockholm syndrome just to sound edgy and not because it was an appropriate description of the situation. I was also trying to define the phrase "serious program" wrt vibe coding, in order to have a conversation, and got made fun of for trying to do so. People sometimes are just trying to get one over on you and prove they're smarter than you, or that I'm an idiot. Which like, okay, not super cool, but I'm just trying to find common ground, or how you came to your beliefs.
The thing is, "enshittification" doesn't name the problem. No part of the word "enshittification" describes " how internet services are deliberately degraded over time". Nor does it propose a solution. It is just a way to say "things are getting worse."
You act as if it was impossible to talk about "how internet services are deliberately being degraded over time" before the word was coined, but it wasn't, we already had a more precise term for that, platform decay.
But my brother in Christ "enshittification" isn't a unique term. It's a common prefix, a common suffix and the word "shit." It was never that great a term of art to begin with, it was just an excuse to say "shit" in polite company. It's a word invented by a blogger for clicks. This isn't a hill worth dying on.
Having thought about your note some more, perhaps this would be a better encapsulation of what I was trying to say:
The consumer internet has become platformized, and the dominant platforms are going through enshittification: early user subsidy, then advertiser/seller favoritism, now rent extraction that is degrading outcomes for everyone.
A number of points that various folks have made in the posts in this thread - free vs paid capabilities, model choices etc. are addressed much more eloquently and coherently in this blog post by Matt Shumer [1]. Discussed here on HN at [2] but like me, many others must have missed it.
[1] https://shumer.dev/something-big-is-happening
[2] https://news.ycombinator.com/item?id=46973011
reply