Reflections on AI at the End of 2025

by danielfalbo • 7 months ago

243 353 7 months ago

danielfalbo5• 7 months ago

There are certain tasks, like improving a given program for speed, for instance, where in theory the model can continue to make progress with a very clear reward signal for a very long time.

This makes me think: I wonder if Goodhart's law[1] may apply here. I wonder if, for instance, optimizing for speed may produce code that is faster but harder to understand and extend. Should we care or would it be ok for AI to produce code that passes all tests and is faster? Would the AI become good at creating explanations for humans as a side effect?

And if Goodhard's law doesn't apply, why is it? Is it because we're only doing RLVR fine-tuning on the last layers of the network so all the generality of the pre-training is not lost? And if this is the case, could this be a limitation in not being able to be creative enough to come up with move 37?

[1] https://wikipedia.org/wiki/Goodhart's_law

username223• 7 months ago

I wonder if, for instance, optimizing for speed may produce code that is faster but harder to understand and extend.

Superoptimizers have been around since 1987: https://en.wikipedia.org/wiki/Superoptimization

They generate fast code that is not meant to be understood or extended.

progval• 7 months ago

But there output is (usually) executable code, and is not committed in a VCS. So the source code is still readable.

When people use LLMs to improve their code, they commit their output to Git to be used as source code.

lemming• 7 months ago

I wonder if, for instance, optimizing for speed may produce code that is faster but harder to understand and extend.

This is generally true for code optimised by humans, at least for the sort of mechanical low level optimisations that LLMs are likely to be good at, as opposed to more conceptual optimisations like using better algorithms. So I suspect the same will be true for LLM-optimised code too.

franktankbank• 7 months ago

Ehh I think if it ends up being a half good architecture you wind up with a difficult to understand kernel that never needs touching.

ur-whale• 7 months ago

Not sure I understand the last sentence:

The fundamental challenge in AI for the next 20 years is avoiding extinction.

danielfalbo5• 7 months ago

I think he's referring to AI safety.

https://lesswrong.com/posts/uMQ3cqWDPHhjtiesc/agi-ruin-a-lis...

grodriguez1002• 7 months ago

For a perhaps easier to read intro to the topic, see https://ai-2027.com/

chrishare• 7 months ago

He's referring to humanity, I believe

A_D_E_P_T• 7 months ago

It's ambiguous. It could go the other way. He could be referring to that oldest of science fiction tropes: The Bulterian Jihad, the human revolt against thinking machines.

timmytokyo3• 7 months ago

It's a tell that he's been influenced by rationalist AI doomer gurus. And a good sign that the rest of his AI opinions should be dismissed.

agumonkey2• 7 months ago

There's videos about Diffusion LLMs too, apparently getting rid of the linear token generation. But I'm no ML engineer.

nephanth• 7 months ago

As someone who worked on transformer-based diffusion models before (not for language though), i can say one thing: they're hard.

Denoising diffusion models benefited a lot from the u-net, which is a pretty simple network (compared to a transformer) and very well-adapted to the denoising task. Plus diffusion on images is great to research because it's very easy to visualize, and therefore to wrap your head around

Doing diffusion on text is a great idea, but my intuition is it will prove more challenging, and probably take a while before we get something working

agumonkey2• 7 months ago

Thanks. Do you see that part of the field as plateauing or ramping up (even taking into account the difficulty).

If you know labs / researchers on the topic, i'd love to read their page / papers

fleebee• 7 months ago

The fundamental challenge in AI for the next 20 years is avoiding extinction.

That's a weird thing to end on. Surely it's worth more than one sentence if you're serious about it? As it stands, it feels a bit like the fearmongering Big Tech CEOs use to drive up the AI stocks.

If AI is really that powerful and I should care about it, I'd rather hear about it without the scare tactics.

grodriguez1002• 7 months ago

I would say yes, everyone should care about it.

There is plenty of material on the topic. See for example https://ai-2027.com/ or https://www.lesswrong.com/posts/uMQ3cqWDPHhjtiesc/agi-ruin-a...

dkdcio2• 7 months ago

fear mongering science fiction, you may as well cite Dune or Terminator

emp173445• 7 months ago

The fact that people here take AI 2027 seriously is embarrassing. The authors are already beginning to walk back these claims: https://x.com/eli_lifland/status/1992004724841906392?s=20

jowea• 7 months ago

And I thought the rest of the thread was anxiety-inducing. Thanks for the nightmares lol.

VladimirGolovin• 7 months ago

This has been well discussed before, for example in this book: https://ifanyonebuildsit.com/

Recursing• 7 months ago

I think https://en.wikipedia.org/wiki/Existential_risk_from_artifici... has much better arguments than the LessWrong sources in other comments, and they weren't written by Big Tech CEOs.

Also "my product will kill you and everyone you care about" is not as great a marketing strategy as you seem to imply, and Big Tech CEOs are not talking about risks anymore. They currently say things like "we'll all be so rich that we won't need to work and we will have to find meaning without jobs"

dist-epoch4• 7 months ago

Yeah, well known marketing trick that Big Companies do.

Oil companies: we are causing global warming with all this carbon emissions, are you scared yet? so buy our stock

Pharma companies: our drugs are unsafe, full of side effects, and kill a lot of people, are you scared yet? so buy our stock

Software companies: our software is full of bugs, will corrupt your files and make you lose money, are you scared yet? so buy our stock

Classic marketing tactics, very effective.

tejohnso• 7 months ago

What makes it a scare tactic? There are other areas in which extinction is a serious concern and people don't behave as though it's all that scary or important. It's just a banal fact. And for all of the extinction threats, AI included, it's very easy to find plenty of deep dive commentary if you care.

alexgotoi• 7 months ago

* The fundamental challenge in AI for the next 20 years is avoiding extinction.

This reminded me of the Don’t look up movie where they basically gambled with the humans extinction.

torlok• 7 months ago

This is a bunch of "I believe" and "I think" with no sources by a random internet person.

echelon2• 7 months ago

by a random internet person.

The creator of Redis.

cinntaile2• 7 months ago

Sure but quite a few claims in the article are about AI research. He does not have any qualifications there. If the focus was more on usefulness, that would be a different discussion and then his experience does add weight.

nutjob22• 7 months ago

Don't see how that gives him more credibility wrt AI.

His entirely unsupported statements about AGI are pretty useless, for instance.

So many people assume AGI is possible, yet no one has a concrete path to it or even a concrete definition of what it or what form it might take.

desbo• 7 months ago

Yeah, it’s called “Reflections”.

ajoseps• 7 months ago

he’s not a “random internet person”, he created Redis. Despite that, I don’t know how authoritative of a figure he is with respect to AI research. He’s definitely a prolific programmer though.

megous• 7 months ago

That still qualifies as a random internet person, wrt the topic. And I think the emphasis is on no sources and I beliefs and I thinks, in any case :)

XorNot2• 7 months ago

There are plenty of Nobel laureates who well, do rest on their laurels and dive deep into pseudoscience after that.

Accomplishment in one field does not make one an expert, nor even particularly worth listening to, in any other. Certainly it doesn't remove the burden of proof or necessity to make an actual argument based on more then simply insisting something is true.

nurettin• 7 months ago

To be fair, you may find equally capable random people in this thread, doesn't mean they speak with any kind of authority.

matthewmacleod• 7 months ago

That is what a blog post is. Someone documenting what they think about a topic.

It's not the case that every form of writing has to be an academic research paper. Sometimes people just think things, and say them – and they may be wrong, or they may be right. And they sometime have some ideas that might change how you think about an issue as a result.

ctoth2• 7 months ago

Ah, I see you have discovered blogs! They're a cool form of writing from like ~20 years ago which are still pretty great. Good thing they show up on this website, it'd be rather dull with only newspapers and journal articles doncha think?

dist-epoch4• 7 months ago

What is a "source"? Isn't it just "another random internet person"?

jacquesm• 7 months ago

Indeed, and, what do you 'believe' or 'think' in response?

dgellow• 7 months ago

It's the personal blog of a famous internet person

feverzsj3• 7 months ago

Seems they also want some AI money[0]. Guess, I'll keep using Valkey.

[0] https://redis.io/redis-for-ai/

danielfalbo5• 7 months ago

they

I'm not sure antirez is involved in any business decision making process at Redis Ltd.

He may not be part of "they".

antirez3• 7 months ago

I'm not involved in business decisions and while I'm very AI positive I believe Redis as a company should focus on Redis fundamentals: so my piece has zero alignment on what I hope for the company.

sibellavia• 7 months ago

In any case, what would be the problem? The page you mentioned simply illustrates how the product can be used in a specific domain; it doesn't seem forced to me.

bgwalter7• 7 months ago

Conflict of interest and disclosure posts are frequently downvoted.

tptacek3• 7 months ago

You mean flagged.

Please don't post insinuations about astroturfing, shilling, brigading, foreign agents, and the like. It degrades discussion and is usually mistaken. If you're worried about abuse, email hn@ycombinator.com and we'll look at the data.

https://news.ycombinator.com/newsguidelines.html

ctoth2• 7 months ago

The fundamental challenge in AI for the next 20 years is avoiding extinction.

So nice to see people who think about this seriously converge on this. Yes. Creating something smarter than you was always going to be a sketchy prospect.

All of the folks insisting it just couldn't happen or ... well, there have just been so many objections. The goalposts have walked from one side of the field to the other, and then left the stadium, went on a trip to Europe, got lost in a beautiful little village in Norway, and decided to move there.

All this time though, the prospect of instantiating a something smarter than you (and yes, it will be smarter than you even if it's at human level because of electronic speeds...) This whole idea is just cursed and we should not do the thing.

cheschire• 7 months ago

"Your scientists were so preoccupied with whether or not they could, they didn't stop to think if they should."

Mawr• 7 months ago

Creating something smarter than you was always going to be a sketchy prospect.

Sure, but not so sure that this has any relevance to the topic at hand. You seem to be taking the assumption that LLMs can ever reach that level for granted.

It may be possible that all it takes is scaling up and at some point some threshold gets reached past which intelligence emerges. Maybe.

Personally, I'm more on board with the idea that since LLMs display approximately 0 intelligence right now, no amount of scaling will help and we need a fundamentally different approach if we want to create AGI.

Aiisnotabubble2• 7 months ago

What also happens and it's irrelevant of AGI: global RL

Around the world people ask an LLM and get a response.

Just grouping and analysing these questions and solving them once centrally and then making the solution available again is huge.

Linearly solving the most asked questions and then the next one then the next will make, whatever system is behind it, smarter every day.

danielfalbo5• 7 months ago

Exactly. The singularity is already here. It's just "programmers + AI" as a whole, rather than independent self-improvements of the AI.

I wonder how a "programmers + AI" self-improving loop is different from an "AI only" one.

bryanrasmussen• 7 months ago

The AI only one presumably has a much faster response time. The singularity is thus not here because programmer time is still the bottleneck, whereas as I understand in the singularity time is no longer a bottleneck component.

Aiisnotabubble2• 7 months ago

AGI will be faster as it doesn't need initial question.

AGI will also be generic.

LLM is already very impressive though

yeasku5• 7 months ago

You are all crazy.

seu• 7 months ago

And I've vibe coded entire ephemeral apps just to find a single bug because why not - code is suddenly free, ephemeral, malleable, discardable after single use. Vibe coding will terraform software and alter job descriptions.

I'm not super up-to-date on all that's happening in AI-land, but in this quote I can find something that most techno-enthusiast seem to have decided to ignore: no, code is not free. There are immense resources (energy, water, materials) that go into these data centers in order to produce this "free" code. And the material consequences are terribly damaging to thousands of people. With the further construction of data centers to feed this free video coding style, we're further destroying parts of the world. Well done, AGI loverboys.

Hendrikto6• 7 months ago

You know what uses roughly 80 times more water in the US alone than water used by AI data centers world wide? Corn.

raddan2• 7 months ago

Assuming your fact is true, that corn merely uses an order of magnitude or two more water than AI is surprising, given the utility of corn. It feeds the entire US (hundreds of millions of people), is used as animal feed (thus also feeding us), and is widely exported to feed other people. I the spirit of the “I think”s and “I believe”s of this blog post, I think that corn has a lot more utility than AI.

dwaltrip2• 7 months ago

Can you provide numbers relative to things many of us already do?

- drive to the store or to work

- take a shower

- eat meat

- fly on vacation

And so on... thanks!

Jaxan3• 7 months ago

Of those things you mention, I only take showers (but not even everyday). But maybe I’m an outlier.

monkaiju• 7 months ago

Whats the point of this question? These are things (some) people do while existing as humans, not sure how thats relevant? The AI is consuming vast resources while note existing as a human, it doesnt get some innate privilege to consume some amount of resources like we do.

fourside• 7 months ago

My guess is that “free” is meant in terms of the old definition where you’re not having to pay someone to create and maintain it. But yes, it’s important to realize there really is a cost here and one that can’t just be captured by a dollar amount.

Fraterkes• 7 months ago

It’s interesting that half the comments here are talking about the extinction line when, now that we’re nearly entering 2026, I feel the 2027 predictions have been shown to be pretty wrong so far.

squidbeak2• 7 months ago

I feel the 2027 predictions have been shown to be pretty wrong so far

Does your clairvoyance go any further than 2027?

AnimalMuppet3• 7 months ago

I don't know that it's "clairvoyance". We're two weeks from 2026. We might be able to see somewhat more than we do now if this was going to turn into AGI by 2027.

If you assume that we're only one breakthrough away (or zero breakthroughs - just need to train harder), then the step could happen any time. If we're more than one away, though, then where are they? Are they all going to happen in the next two years?

But everybody's guessing. We don't know right now whether AGI is possible at current hardware levels. If it is N breakthroughs away, we all have our own guesses of approximately what N is.

My guess is that we are more than one breakthrough away. Therefore, one can look at the current state of affairs and say that we are unlikely to get to AGI by 2027.

jennyholzer23• 7 months ago

Does your clairvoyance go any further than 2027?

why are you so sensitive?

a_bonobo2• 7 months ago

* For years, despite functional evidence and scientific hints accumulating, certain AI researchers continued to claim LLMs were stochastic parrots: probabilistic machines that would: 1. NOT have any representation about the meaning of the prompt. 2. NOT have any representation about what they were going to say. In 2025 finally almost everybody stopped saying so.

Man, Antirez and I walk in very different circles! I still feel like LLMs fall over backwards once you give them an 'unusual' or 'rare' task that isn't likely to be presented in the training data.

jmfldn2• 7 months ago

"In 2025 finally almost everybody stopped saying so."

I haven't.

dist-epoch4• 7 months ago

Some people are slower to understand things.

oersted2• 7 months ago

LLMs certainly struggle with tasks that require knowledge that is not provided to them (at significant enough volume/variance to retain it). But this is to be expected of any intelligent agent, it is certainly true of humans. It is not a good argument to support the claim that they are Chinese Rooms (unthinking imitators). Indeed, the whole point of the Chinese Room thought experiment was to consider if that distinction even mattered.

When it comes to of being able to do novel tasks on known knowledge, they seem to be quite good. One also needs to consider that problem-solving patterns are also a kind of (meta-)knowledge that needs to be taught, either through imitation/memorisation (Supervised Learning) or through practice (Reinforcement Learning). They can be logically derived from other techniques to an extent, just like new knowledge can be derived from known knowledge in general, and again LLMs seem to be pretty decent at this, but only to a point. Regardless, all of this is definitely true of humans too.

feverzsj3• 7 months ago

In most cases, LLMs has the knowledge(data). They just can't generalize them like human do. They can only reflect explicit things that are already there.

barnabee2• 7 months ago

I don’t think this is quite true.

I’ve seen them do fine on tasks that are clearly not in the training data, and it seems to me that they struggle when some particular type of task or solution or approach might be something they haven’t been exposed to, rather than the exact task.

In the context of the paragraph you quoted, that’s an important distinction.

It seems quite clear to me that they are getting at the meaning of the prompt and are able, at least somewhat, to generalise and connect aspects of their training to “plan” and output a meaningful response.

This certainly doesn’t seem all that deep (at times frustratingly shallow) and I can see how at first glance it might look like everything was just regurgitated training data, but my repeated experience (especially over the last ~6-9 months) is that there’s something more than that happening, which feels like whet Antirez was getting at.

Kiro2• 7 months ago

Give me an example of one of those rare or unusual tasks.

recursive• 7 months ago

Set the font size of a simple field in openxml. Doesn't even seem that rare. It said to add a run inside and set the font there. Didn't do anything. I ended up reverse engineering the output out of ms word. This happened yesterday.

a_bonobo2• 7 months ago

I work on a few HPC systems with unusual, kinda custom-rolled architectures. A whole bunch of Python and R packages fail to compile on these systems. There's no publicly accessible documentation for these HPC systems, nor for these custom architectures. ChatGPT and Claude so far have given me only wrong advice on how to get around these compilation errors and there's not much on Google for these errors, but HPC staff usually knew what to do.

rckt3• 7 months ago

Even if LLMs make mistakes, the ability of LLMs to deliver useful code and hints improved to the point most skeptics started to use LLMs anyway

Here we go again. Statements with the single source in the head of the speaker. And it’s also not true. The llms still produce bad/irrelevant code at such rate that you can spend more time prompting than doing things yourself.

I’m tired of this overestimation of llms.

iamflimflam1• 7 months ago

But you have just repeated what you are complaining about.

rckt3• 7 months ago

Do you want me to spend time to come with a quality response to a lazy statement? It’s like fighting with windmills. I’m fine with having my say the way I did.

xiconfjs• 7 months ago

My person experience: if I can find a solution on stackoverflow etc. the LLM will produce working and fundamentally correct code. If I can‘t find a already fullfilled solution on these sites, the LLM is hallucinating like crazy (newer existing functions/modules/plugins, protocol features which aren’t specified and even github-repos which never existed). So, as stated my many people online before: for low-hanging fruits LLM are totally viable solution.

danielbln2• 7 months ago

I don't remember the last time Claude Code hallucinated some library, as it will check the packages, verify with the linter, run a test import and so on.

Are you talking about punching something into some LLM web chat that's disconnected from your actual codebase and has tooling like web search disabled? If so, that's not really the state of the art of AI assisted coding, just so you know.

barnabee2• 7 months ago

Even where they are not directly using LLMs to write the most critical or core code, nearly every skeptic I know has started using LLMs at very least to do things like write tests, build tools, write glue code, help to debug or refactor, etc.

Your statement suffers not only from also coming only from your brain, with no evidence that you've actually tried to learn to use these tools, but it also goes against the weight of evidence that I see both in my professional network and online.

rckt3• 7 months ago

I just want people making statements like the author to be more specific how exactly the llms are being used. Otherwise they contribute to this belief that llms are a magical tool that can do anything.

I am aware of simple routine tasks that LLMs can do. This doesn’t change anything about what I said.

AnimalMuppet3• 7 months ago

You need to meet more skeptics. (Or maybe I do.) In my world, it's much more rare than you say.

locknitpicker• 7 months ago

Here we go again. Statements with the single source in the head of the speaker. And it’s also not true.

You're making the same sort of baseless claim you are criticising the blogger for making. Spewing baseless claims hardly moves any discussion forward.

The llms still produce bad/irrelevant code at such rate that you can spend more time promoting than doing things yourself.

If that is your personal experience then I regret to tell you that it is only the reflection of your own inability to work with LLMs and coding agents. Meanwhile, I personally manage to effectively use LLMs anywhere between small refactoring needs and large software architecture designs, including generating fully working MVPs in one-shot agent prompts. From this alone it's rather obvious who is making baseless statements that are more aligned with reality.

bgwalter7• 7 months ago

Here we go again.

Indeed, he said the same as a reflection on 2024 models:

https://news.ycombinator.com/item?id=42561151

It is always the fault of the "luser" who is not using and paying for the latest model.

dhpe• 7 months ago

I have programmed 30K+ hours. Do LLMs make bad code: yes all the time (at the moment zero clue about good architecture). Are they still useful: yes, extremely so. The secret sauce is that you'd know exactly what to do without them.

qsort• 7 months ago

One of the mental frameworks that convinced me is how much of a "free action" it is. Have the LLM (or the agent) churn on some problem and do something else. Come back and review the result. If you had to put significant effort into each query, I agree it wouldn't be worth it, but you can just type something into the textbox and wait.

daveguy5• 7 months ago

Are you counting the time/effort to evaluate the accuracy and relevance of an LLM left to "think" for a while?

feverzsj3• 7 months ago

So, it's like taking off your pants to fart.

_rpxpx• 7 months ago

OK, maybe. But how many programmers will know this in 10 years' time as use of LLMs is normalized? I like to hear what employers are saying already about recent graduates.

bartread2• 7 months ago

They’d have to be hiring recent graduates for you to hear that perspective.

And, as much as what I’ve just said is hyperbolically pessimistic, there is some truth to it.

In the UK a bunch of factors have coincided to put the brakes on hiring, especially smaller and mid-size businesses. AI is the obvious one that gets all the press (although how much it’s really to blame is open to question in my view), but the recent rise in employer AI contribution, and now (anecdotally) the employee rights bill have come together to make companies quite gunshy when it comes to hiring.

spaceman_2020• 7 months ago

This is nothing new - entire industries and skills died out as the apprenticeship system and guilds were replaced by automation and factories

energy123• 7 months ago

I'm uncertain that programming will be a major profession in 10 years.

Programming is more like math than creative writing. It's largely verifiable, which is where RL is repeatedly proven to eventually achieve significantly better than human intelligence.

Our saving grace, for now, is that it's not entirely verifiable because things like architectural taste are hard to put into a test. But I would not bet against it.

nutjob22• 7 months ago

If they don't learn that they won't get very far.

This is true for everything, any tool you might use. Competent users of tools understand how they work and thus their limitations and how they're best put to work.

Incompetents just fumble around and sometimes get things working.

QuiDortDine• 7 months ago

hahah what are you talking about, there's no such thing as long term!

dejv• 7 months ago

"Do LLMs make bad code: yes all the time (at the moment zero clue about good architecture). Are they still useful: yes, extremely so."

Well, lets see how all the economics will play out. LLMs might be really useful, but as far as I can see all the AI companies are not making money on inference alone. We might be hitting plateau in capabilities with money being raised on vision of being this godlike tech that will change the world completely. Sooner or later the costs will have to meet the reality.

ImprobableTruth• 7 months ago

They're not making money on inference alone because they blow ungodly amounts on R&D. Otherwise it'd be a very profitable business.

Workaccount22• 7 months ago

If the tech plateaus today, LLM plans will go to $60-80/mo, Chinese-hosted chinese models will be banned (national security will be the given reason), and the AI companies will be making ungodly money.

I'm not gonna dig out the math again, but if AI usage follows the popularity path of cell phone usage (which seems to be the case), then trillions invested has a ROI of 5-7 years. Not bad at all.

20k2• 7 months ago

This is one of the reasons why I'm surprised to see so many people jump on board. We're clearly in the "release product for free/cheap to gain customers" portion of the enshittification plan, before the company starts making it completely garbage to extract as much money as possible from the userbase

Having good quality dev tools is non negotiable, and I have a feeling that a lot of people are going to find out the hard way that reliability and it not being owned by profit seeking company is the #1 thing you want in your environment

Aurornis9• 7 months ago

but as far as I can see all the AI companies are not making money on inference alone

The numbers aren’t public, but from what companies have indicated it seems inference itself would be profitable if you could exclude all of the R&D and training costs.

But this debate about startups losing money happens endlessly with every new startup cycle. Everyone forgets that losing money is an expected operating mode for a high growth startup. The models and hardware continue to improve. There is so much investment money accelerating this process that we have plenty of runway to continue improving before companies have to switch to full profit focus mode.

But even if we ignore that fact and assume they had to switch to profit mode tomorrow, LLM plans are currently so cheap that even a doubling or tripling isn’t going to be a problem. So what if the monthly plans start at $40 instead of $20 and the high usage plans go from $200 to $400 or even $600? The people using these for their jobs paying $10K or more per month can absorb that.

That’s not going to happen, though. If all model progress stopped right now the companies would still be capturing cheaper compute as data center buildouts were completed and next generation compute hardware was released.

I see these predictions as the current equivalent of all of the predictions that Uber was going to collapse when the VC money ran out. Instead, Uber quietly settled into steady operation, prices went up a little bit, and people still use Uber a lot. Uber did this without the constant hardware and model improvements that LLM companies benefit from.

nl3• 7 months ago

Anthropic - for one - is making lots of money on inference.

NitpickLawyer5• 7 months ago

but as far as I can see all the AI companies are not making money on inference alone.

This was the missed point on why GPT5 was such an important launch (quality of models and vibes aside). It brought the model sizes (and hence inference cost) to more sustainable numbers. Compared to previous SotA (GPT4 at launch, or o1/3 series), GPT5 is 8x-12x cheaper! I feel that a lot of people never re-calibrated their views on inference.

And there's also another place where you can verify your take on inference - the 3rd party providers that offer "open" models. They have 0 incentive to subsidise prices, because people that use them often don't even know who serves them, so there's 0 brand recognition (say when using models via openrouter).

These 3rd party providers have all converged towards a price-point per billion param models. And you can check those prices, and have an idea on what would be proffitable and at what sizes. Models like dsv3.2 are really really cheap to serve, for what they provide (at least gpt5-mini equivalent I'd say).

So yes, labs could totally become profitable with inference alone. But they don't want that, because there's an argument to be made that the best will "keep it all". I hope, for our sake as consumers that it isn't the case. And so far this year it seems that it's not the case. We've had all 4 big labs one-up eachother several times, and they're keeping eachother honest. And that's good for us. We get frontier level offerings at 10-25$/MTok (Opus, gpt5.2, gemini3pro, grok4), and we get highly capable yet extremely cheap models at 1.5-3$/MTok (gemini3-flash, gpt-minis, grok-fast, etc)

mNovak• 7 months ago

Doesn't OpenRouter prove that inference is profitable? Why would random third parties subsidize the service for other random people online? Unless you're saying that only large frontier models are unprofitable, which I still don't think is the case but is harder to prove.

bilsbie• 7 months ago

I mean if you leaned heavily on stack overflow before AI then nothing really changes.

It’s basically the same idea but faster.

ManuelKiessling• 7 months ago

If I ask a SOTA model to just implement some functionality, it doesn’t necessarily do so using a great architectural approach.

Whenever I ask a SOTA model about architecture recommendations, and frame the problem correctly, I get top notch answers every single time.

LLMs are terrific software architects. And that’s not surprising, there has to be tons of great advice on how to correctly build software in the training corpus.

They simply aren’t great software architects by default.

Loic• 7 months ago

You know that if you ask the LLM correctly you get top notch answers, because you have the experience to judge if the answer is top notch or not.

I spend a couple of hours per week teaching software architecture to a junior in my team, because he has not the experience to not only ask correctly but also assess the quality of the answer from the LLM.

yeasku5• 7 months ago

I have programed 10 times that.

For me LLMs are a waste of time.

rrrrrrrrrrrryan• 7 months ago

300k hours = 8 hrs per day, every day, for 102 years.

piker4• 7 months ago

There are certain tasks, like improving a given program for speed, for instance, where in theory the model can continue to make progress with a very clear reward signal for a very long time.

Super skeptical of this claim. Yes, if I have some toy poorly optimized python example or maybe a sorting algorithm in ASM, but this won’t work in any non-trivial case. My intuition is that the LLM will spin its wheels at a local minimum the performance of which is overdetermined by millions of black-box optimizations in the interpreter or compiler signal from which is not fed back to the LLM.

dist-epoch4• 7 months ago

https://github.com/algorithmicsuperintelligence/openevolve

piker4• 7 months ago

https://chatgpt.com/backend-api/estuary/public_content/enc/e...

andy99• 7 months ago

There was a discussion the other day where someone asked Claude to improve a code base 200x https://news.ycombinator.com/item?id=46197930

exitb• 7 months ago

That’s most definitely not the same thing, as „improving a codebase” is an open ended task with no reliable metrics the agent could work against.

NitpickLawyer5• 7 months ago

but this won’t work in any non-trivial case

Earlier this year google shared that one of their projects (I think it was alphaevolve) found an optimisation in their stack that sped up their real world training runs by 1%. As we're talking about google here, we can be pretty sure it wasn't some trivial python trick that they missed. Anyhow, at ~100M$ / training run, that's a 1M$ save right there. Each and every time they run a training run!

And in the past month google also shared another "agentic" workflow where they had gemini2.5-fhash! (their previous gen "small" model) work autonomously on migrating codebases to support aarch64 architecture. There they found ~30% of the projects worked flawlessly end-to-end. Whatever costs they save from switching to ARM will translate in real-world $ saved (at google scale, those can add up quickly).

piker4• 7 months ago

The second example has nothing to do with the first. I am optimistic that LLMs are great for translations with good testing frameworks.

“Optimize” in a vacuum is a tarpit for an LLM agent today, in my view. The Google case is interesting but 1% while significant at Google scale doesn’t move the needle much in terms of statistical significance. It would be more interesting to see the exact operation and the speed up achieved relative to the prior version. But it’s data contrary to my view for sure. The cynic also notes that Google is in the LLM hype game now, too.

Jaxan3• 7 months ago

As we're talking about google here, we can be pretty sure it wasn't some trivial python trick that they missed.

Strong disagree on the reasoning here. Especially since google is big and have thousands of developers, there could be a lot of code and a lot of low hanging fruit.

abricq2• 7 months ago

* Programmers resistance to AI assisted programming has lowered considerably. Even if LLMs make mistakes, the ability of LLMs to deliver useful code and hints improved to the point most skeptics started to use LLMs anyway: now the return on the investment is acceptable for many more folks.

Could not agree more. I myself started 2025 being very skeptical, and finished it very convinced about the usefulness of LLMs for programming. I have also seen multiple colleagues and friends go through the same change of appreciation.

I noticed that for certain task, our productivity can be multiplied by 2 to 4. So hence comes my doubts: are we going to be too many developers / software engineers ? What will happen for the rests of us ?

I assume that other fields (other than software-related) should also benefits from the same productivity boosts. I wonder if our society is ready to accept that people should work less. I think the more likely continuation is that companies will either hire less, or fire more, instead of accepting to pay the same for less hours of human-work.

danielfalbo5• 7 months ago

Are we going to be too many developers / software engineers ? What will happen for the rests of us?

I propose that we should raise the bar for the quality of software now.

abricq2• 7 months ago

Yes, certainly agree. A few days ago here there was this blog claiming how formal verification would become widely more used with AI. The author claiming that AI will help us with the difficulty barrier to write formal proofs.

throw12354354• 7 months ago

I don't think that will happen because it hasn't for other technological improvements. In the end people pay for "good enough" and that's that. If "good enough" is now cheaper to implement that's all they will do. I've seen it in other technologies. As an example due to more precise manufacturing many manufacturers have used it to cheapen things like cars, electronics, etc just to the point where it passes warranty mostly; in the old days they had to "overbuild" to get it to that point putting more quality into the product.

Quality is a risk mitigation strategy; if software is disposable just like cheap manufactured goods most people won't pay for it thinking they can just "build another one". What we don't realise is due to sheer cost of building software we've wanted quality because its too expensive to fix later; AI could change that.

Hoping we invest in quality, more software (which has a price inelastic curve mostly due to scale/high ROI) etc I'm starting to think is just false hope from people in the tech industry that want to be optimistic which generally is in our nature. Tech people understand very little about economics most of the time and how people outside tech (your customers) generally operate. My reflection is mostly I need to pivot out of software; it will be commoditized.

antihipocrat• 7 months ago

I like to think of it as adding new lanes to a highway. More will be delivered until it all jams up again.

throw12354354• 7 months ago

I'm not sure that it will scale to other fields other than coding and math. The approach with RLVR makes it more amenable to STEM fields in general and most jobs believe it or not aren't that. The level of open source software with good test suites effectively gave them all the training material they needed; most professions won't provide that knowing that they will be giving their moat away. LLM's to other fields from my understanding still exhibit the same hallucination rates if only mildly improved especially if there isn't public internet material in that field.

We have to accept in the end that coding/SWE is one of the most disrupted fields from this breed of AI. Disruption unfortunately probably means less jobs overall. The profession is on trend to disrupting and automating itself I think; plan accordingly. I've seen so many articles claiming its great we didn't learn to code now; that's what the AI's have done.

Where to understand more about how chain of thoughs really affects LLMs performance? I read the seminal paper but all it says is that it's basically another prompt engineering tecnique that improves accuracy.

HarHarVeryFunny6• 7 months ago

Chain of thought, now including "reasoning", are basically a work around for the simplistic nature of the Transformer neural network architecture that all LLMs are based on.

The two main limitations of the Transformer that it helps with are:

1) A Transformer is just a fixed-size stack of layers, with a one-way flow of data through the layers from input to output. The fixed number of layers equates to how many "thought" steps the LLM can put into generating each word of output, but good responses to harder questions may require many more steps and iterative thinking...

The idea of "think step by step", aka chain of thought, is to have the model break it's response down into a sequence of steps, each building on what came before, so that the scope of each step is withing the capability of the fixed number of layers of the transformer.

2) A Transformer has extremely limited internal memory from one generated word to the next, so telling the model to go one step at a time, feeding its own output back in as input, in effect makes the model's output a kind of memory that makes up for this.

So, chain of thought prompting ultimately give the model more thinking steps (more words generated), together with memory of what it is thinking, in order to be able to generate a better response.

bachmeier• 7 months ago

Programmers resistance to AI assisted programming has lowered considerably. Even if LLMs make mistakes, the ability of LLMs to deliver useful code and hints improved to the point most skeptics started to use LLMs anyway: now the return on the investment is acceptable for many more folks.

I'm not a fan of this phrasing. Use of the terms "resistance" and "skeptics" implies they were wrong. It's important we don't engage in revisionist history that allows people in the future to say "Look at the irrational fear programmers had of AI, which turned out to be wrong!" The change occurred because LLMs are useful for programming in 2025 and the earliest versions weren't for most programmers. It was the technology that changed.

20k2• 7 months ago

Its also significantly lowered because management is forcing AI on everyone at gunpoint, and saying that you'll lose your job if you don't love AI

That's a very easy way to get everyone to pinky promise that they absolutely love AI to the ends of the earth

Aurornis9• 7 months ago

The change occurred because LLMs are useful for programming in 2025

But the skeptics and anti-AI commenters are almost as active as ever, even as we enter 2026.

The debate about the usefulness of LLMs has grown into almost another culture war topic. I still see a constant stream of anti-AI comments on HN and every other social platform from people who believe the tools are useless, the output is always unusable, people who mock any idea that operator skill has an impact on LLM output, or even claims that LLMs are a fad that will go away.

I’m a light LLM user ($20/month plan type of usage) but even when I try to share comments about how I use LLMs or tips I’ve discovered, I get responses full of vitriol and accusations of being a shill.

zahlman6• 7 months ago

It absolutely is culture war. I can easily imagine a less critical version of myself having ended up in that camp. It comes across to me that the perspective is informed by core values and principles surrounding what "intelligence" is.

I butted heads with many earlier on, and they did nothing to challenge that frame meaningfully. What did change is my perception of the set of tasks that don't require "intelligence". And the intuition pump for that is pretty easy to start — I didn't suppose that Deep Blue heralded a dawn of true "AI", either, but chess (and now Go) programs have only gotten even more embarrassingly stronger. Even if researchers and puzzle enthusiasts might still find positions that are easier for a human to grok than a computer.

Hendrikto6• 7 months ago

from people who believe the tools are useless, the output is always unusable, people who mock any idea that operator skill has an impact on LLM output

You are attacking a strawman. Almost nobody claims that LLMs are useless or you can never use their output.

otabdeveloper46• 7 months ago

"Useful for programming" is a massive and dishonest bait and switch.

Lots of things are "useful for programming". Switching to a comfier chair is more useful for programming than any LLM.

We were sold vibe coding, and that's what managers want.

throw12354354• 7 months ago

Its simple. Given the trajectory of these things people feel under threat and defend themselves accordingly. They say what they hope for given a number of factors (bad workplaces generating slop they have to deal with, job losses, identity redefinition, etc). You know the things that happen when a profession is disrupted in a capitalist system where 'what you do' is often tied up with identity, status, and livelihood.

People will go from skeptic to dread/anxiety, to either acceptance or despair. We are witnessing the disruption of a profession in real time and it will create a number of negative effects.

ookblah• 7 months ago

you just need to hop into any AI reltaed thread (even this one) and it's pretty clear no one is revising anything, skepticism is there lol.

mjr003• 7 months ago

"Skeptics" is also a loaded term; what does it actually mean? I find LLMs incredibly useful for various programming tasks (generating code, searching documentation, and yes with enough setup agents can accomplish some tasks), but I also don't believe they have actual intelligence, nor do I think they will eviscerate programming jobs, the same way that Python and JavaScript didn't eviscerate programming jobs despite lowering the barrier to entry compared to Java or C. Does that make me a skeptic?

It's easy to declare "victory" when you're only talking about the maximalist position on one side ("LLMs are totally useless!") vs the minimalist position on the other side ("LLMs can generate useful code"). The AI maximalist position of "AI is going to become superintelligent and make all human work and intelligence obsolete" has certainly not been proven.

Aurornis9• 7 months ago

No, that doesn’t make you a skeptic in this context.

The LLM skeptics claim LLM usefulness is an illusion. That the LLMs are a fad, and they produced more problems than they solve. They cite cherry picked announcements showing that LLM usage makes development slower or worse. They opened ChatGPT a couple times a few months ago, asked some questions, and then went “Aha! I knew it was bad!” when they encountered their first bad output instead of trying to work with the LLM to iterate like everyone who gets value out of them.

The skeptics are the people in every AI thread claiming LLMs are a fad that will go away when the VC money runs out, that the only reason anyone uses LLMs is because their boss forces them to, or who blame every bug or security announcement on vibecoding.

threethirtytwo2• 7 months ago

You're not a skeptic but you're not fully a supporter either. You live in this grey zone of contradictions.

First you find them useful but not intelligent. That is a bit of a contradiction. Basically anyone who has used AI, seriously knows that while it can be used to regurgitate generic filler and bootstrap code it can also be used to solve complex domain specific problems that is not at all part of its training data. This by definition makes it intelligent and it makes it so we know the LLM understands the problem it was given. it would be This by definition makes it intelligent, and it makes it so we know the LLM understands the problem it was given. It would be disingenuous for me not to mention how wrong and how much an LLM hallucinates, so obviously the thing has flaws and is not super intelligence. But you have to judge the entire spectrum of what it does. It gets things right and it gets things wrong and getting something complex right makes it intelligent while getting something wrong does not predude it from intelligence.

Second most non skeptics aren't saying all human work is going to be obsolete. no one can predict the future. But you've got to be blind if you don't see the trendline of progress. Literally look at the progress of AI for the past 15 years. You have to be next level delusional if you can't project another 15 years and see that obviously a super intelligence or at least an intelligence comparable to humans is not a reasonable prediction. Most skeptics like you ignore the trendline and cling to what Yann lecunn said about llms being stochastic parrots. It is very likely something with human intelligence exists in the future and in our lifetimes, whether or not its an LLM remains to be seen but we can't ignore where the trendlines are pointing.

nl3• 7 months ago

There is some limited truth in this but we still see claims that LLMs are "just next token predictors" and "just regurgitate code they read online". These are just uninformed and wrong views. It's fair to say that these people were (are!) wrong.

zahlman6• 7 months ago

Objecting to these claims is missing their point. Saying these things is really about denying that the LLMs "think" in any meaningful sense. (And the retorts I've seen in those discussions often imply very depressing and self-deprecating views of what it actually means to be human.)

mjr003• 7 months ago

we still see claims that LLMs are "just next token predictors" and "just regurgitate code they read online". These are just uninformed and wrong views. It's fair to say that these people were (are!) wrong.

I don't think it's fair to say that at all. How are LLMs not statistical models that predict tokens? It's a big oversimplification but it doesn't seem wrong, the same way that "computers are electricity running through circuits" isn't a wrong statement. And in both cases, those statements are orthogonal to how useful they are.

mvkel3• 7 months ago

One only has to go read the original vibe coding thread[0] from ...ten months ago(!) to see the resistance and skepticism loud and clear. The very first comment couldn't be more loud about it.

It was possible to create things in gpt-3.5. The difference now is it aligns with the -taste- of discerning programmers, which has a little, but not everything, to do with technological capability.

[0]https://news.ycombinator.com/item?id=42913909

zahlman6• 7 months ago

The difference now is it aligns with the -taste- of discerning programmers

This... doesn't match the field reports I've seen here, nor what I've seen from poking around the repos for AI-powered Show HN submissions.

HarHarVeryFunny6• 7 months ago

"Look Ma, no hands!" vibe coding, as described by Karpathy, where you never look at the code being generated, was never a good idea, and still isn't. Some people are now misusing "vibe coding" to describe any use of LLMs for coding, but there is a world of difference between using LLMs in an intelligent considered way as part of the software development process, and taking a hit on the bong and "vibe coding" another "how many calories in this plate of food" app.

HarHarVeryFunny6• 7 months ago

Yes, it's a strange take. It's not that programmers have changed their mind about unchanging LLMs, but rather that LLMs have changed and are now useful for coding, not just CoPilot autocomplete like the early ones.

What changed was the use of RLVR training for programming, resulting in "reasoning" models that are now attempting to optimize for a long-horizon goal (i.e. bias generation towards "reasoning steps" that during training let to a verified reward), as opposed to earlier LLMs where RL was limited to RLHF.

So, yeah, the programmers who characterized early pre-RLVR coding models as of limited use were correct. Now the models are trained differently and developers find them much more useful.

zahlman6• 7 months ago

I thought I'd read a lot of these threads this year, and also discussed off-site the use of coding agents and the technology behind them; but this is genuinely the first time I've seen the term "RLVR".

throw12354354• 7 months ago

Agree with this. The RLVR changes (starting with o1 I think) was what changed/disrupted the industry. Before that I thought these things were just better autocomplete.

erichocean3• 7 months ago

1. NOT have any representation about the meaning of the prompt.

This one is bizarre, if true (I'm not convinced it is).

The entire purpose of the attention mechanism in the transformer architecture is to build this representation, in many layers (conceptually: in many layers of abstraction).

2. NOT have any representation about what they were going to say.

The only place for this to go is in the model weights. More parameters means "more places to remember things", so clearly that's at least a representation.

Again: who was pushing this belief? Presumably not researchers, these are fundamental properties of the transformer architecture. To the best of my knowledge, they are not disputed.

I believe [...] it is not impossible they get us to AGI even without fundamentally new paradigms appearing.

Same, at least for the OpenAI AGI definition: "An AI system that is at least as intelligent as a normal human, and is able to do any economically valuable work."

zahlman6• 7 months ago

This one is bizarre, if true (I'm not convinced it is).
The entire purpose of the attention mechanism in the transformer architecture is to build this representation, in many layers (conceptually: in many layers of abstraction).

I think this is really about a hidden (i.e. not readily communicated) difference in what the word "meaning" means to different people.

erichocean3• 7 months ago

Could be, by "meaning" I mean (heh) that transformers are able to distinguish tokens (and prompts) in a consequential ("causal") way, and that they do so at various levels of detail ("abstractions").

I think that's the usual understanding of how transformer architectures work, at the level of math.

etra06• 7 months ago

LLMs have certainly become extremely useful for Software Engineers, they're very convincing (and pleasers, too) and I'm still unsure about the future of our day-to-day job.

But one thing that has scared me the most, is the trust of LLMs output to the general society. I believe that for software engineers it's really easy to see if it's being useful or not -- We can just run the code and see if the output is what we expected, if not, iterate it, and continue. There's still a professional looking to what it produces.

On the contrary, for more day-to-day usage of the general pubic, is getting really scary. I've had multiple members of my family using AI to ask for medical advice, life advice, and stuff were I still see hallucinations daily, but at the same time they're so convincing that it's hard for them not to trust them.

I still have seen fake quotes, fake investigations, fake news being spreaded by LLMs that have affected decisions (maybe, not as crucials yet but time will tell) and that's a danger that most software engineers just gross over.

Accountability is a big asterisk that everyone seems to ignore

santadays• 7 months ago

I get this take, but given the state of the world (the US anyways), I find it hard to trust anyone with any kind of profit motive. I feel like any information can’t be taken as fact, it can just be rolled into your world view and discarded if useful or not. If you need to make a decision that can’t be backed out of that has real world consequences I think/hope most people are learning to do as much due diligence as reasonable. Llms seem at this moment to be trying to give reliable information. When they’ve been fine tuned to avoid certain topics it’s obvious. This could change but I suspect it will be hard to find tune them too far in a direction without losing capability.

That said, it definitely feels as though keeping a coherent picture of what is actually happening is getting harder, which is scary.

twoodfin• 7 months ago

I feel like any information can’t be taken as fact, it can just be rolled into your world view and discarded if useful or not.

The concern, I think, is that for many that “discard function” is not, “Is this information useful?”. Instead: “Does this information reinforce my existing world view?”

That feedback loop and where it leads is potentially catastrophic at societal scale.

etra06• 7 months ago

I find it hard to trust anyone with any kind of profit motive.

As much as this is true, and i.e. doctors for sure can profit (here in my country they don't get any type of sponsor money AFAIK, other than having very high rates), there is still accountability.

We have built a society based on rules and laws, if someone does something that can harm you, you can follow the path to at least hold someone accountable (or, try).

The same cannot be said about LLMs.

raincole2• 7 months ago

using AI to ask for medical advice

So the number of anti-vaxxers is going to plummet drastically in the following decade, I guess.

preisschild• 7 months ago

Depends if they use lobotomized bots like Grok...

etra06• 7 months ago

I haven't tried with this specific topic, but being the pleasers llms are, I doubt someone so focused on being anti-vaxxer will be convinced by an LLM, if anything, the LLM will give them reason at some point.

joshribakoff• 7 months ago

With code, even when it looks correct, it can be subtly wrong and traditional search engines don’t sit there and repeatedly pressure you into merging the PR.

zamadatix2• 7 months ago

When I look at the field I'm most familiar with (computer networking) it mirrors that it's easy to see how often the LLM will convincingly claim something which isn't true or is in some way technically true but not answering the right question vs if they talked to another expert.

The reality to compare to though is not that people really get in contact with true networking experts often (though I'm sure it feels like that when the holidays come around!) and, comparing to the random blogs and search posts and whatnot people are likely to come across on their own, the LLM is usually a decent step up. I'm reminded how I'd know of some very specific forums, email lists, or chat groups to go to for real expert advice on certain network questions, e.g. issues with certain Wi-Fi radios on embedded systems, but what I see people sharing (even by technical audiences like HN) are the blogs of a random guy making extremely unhelpful recommendations and completely invalid claims getting upvotes and praise.

With things like asking AI for medical advice... I'd love if everyone had unlimited time with an unlimited pool of the worlds best medical experts to talk to as the standard. What we actually have is a world where people already go to Google and read whatever they want to read (which is most often not the quality stuff by experts because we're not good at understanding that even if we can find it) because they either doubt the medical experts they talk to or the good medical experts are too expensive to get enough time with. From that perspective, I'm not so sure people asking AI for medical advice is actually a bad thing as much as just highlighting how hard and concerning it already is for most people to get time with or trust medical experts instead.

zdragnar• 7 months ago

This justification comes up when discussing therapy too.

To take it to an extreme, it's basically saying "people already get little or bad advice, we might as well give them some more bad advice."

I simply don't buy it.

Kuxe• 7 months ago

Swedish politician Ebba Busch used LLM to write a speech. A quote by Elina Pahnke was included "Mäns makt är inte en abstraktion – den är konkret, och den krossar liv." (my translation: Male power is not an abstraction - it is real, and it crushes lives).

Elina listened in on the speech and got surprised :)...

https://www.aftonbladet.se/nyheter/a/gw8Oj9/ebba-busch-anvan...

Ebba apologized, great, but it begs the question: how many quotes and misguided information is being acted on already? If crucial decisions can be made off incorrect decisions then they will. Murphys law!

cauliflower2718• 7 months ago

Regarding medical information: medical professionals in the US, including your doctor, use uptodate.com, which is basically a medical encyclopedia that is regularly updated by experts in their field. While it's very expensive to get a year long subscription, a week long subscription (for non medical professionals) is only around $20 and you can look up anything you want.

layer83• 7 months ago

We can just run the code and see if the output is what we expected

There is a vast gap between the output happening to be what you expect and code being actually correct.

That is, in a way, also the fundamental issue with LLMs: They are designed to produce “expected” output, not correct output.

Verdex• 7 months ago

For example:

The output is correct but only for one input.

The output is correct for all inputs but only with the mocked dependency.

The output looks correct but the downstream processors expected something else.

The output is correct for all inputs with real world dependencies and is in the correct structure for downstream processors, but it's not being registered with the schema filtered and it all gets deleted in prod.

While implementing the correct function you fail to notice that the correct in every way output doesn't conform to that thing that Tom said because you didn't code it yourself but instead let the LLM do it. The system works flawlessly with itself but the final output fails regulatory compliance.

etra06• 7 months ago

That is exactly my point, though.

I didn't mean they do it on the first time, or that it is correct, I mean that you can 'run' and 'test it' to see if it does what you want in the way you want.

The same cannot be said to any other topics like medical advice, life advice, etc.

The point is, how verifiable is the output the LLM gives and so how useful it is.

laterium• 7 months ago

The issue you're overlooking is the scarcity of experts. You're comparing the current situation to an alternative universe where every person can ask a doctor their questions 10 times a day and instantly get an accurate response.

That is not the reality we're living in. Doctors barely give you 5 minutes even if you get an appointment days or weeks in advance. There is just nobody to ask. The alternatives today are

1) Don't ask, rely on yourself, definitely worse than asking a doctor

2) Ask an LLM, which gets you 80-90% of the way there.

3) Google it and spend hours sifting through sponsored posts and scams, often worse than relying on yourself.

The hallucinations that happen are massively outweighed by the benefits people get by asking them. Perfect is the enemy of good enough, and LLMs are good enough.

Much more important also is that LLMs don't try to scam you, don't try to fool you, don't look out for their own interests. Their mistakes are not intentional. They're fiduciaries in the best sense, just like doctors are, probably even more so.

bgwalter7• 7 months ago

Much more important also is that LLMs don't try to scam you, don't try to fool you, don't look out for their own interests.

They follow their corporations instead. Just look at the status-quoism of the free "Google AI" and the constant changes in Grok, where xAI is increasingly locking down Grok, perhaps to stay in line with EU regulations. But Grok is also increasingly pro-billionaire.

Copilot was completely locked down on anything political before the 2024 election.

They all scam you according to their training and system prompts. Have you seen the minute change in the system prompt that led to MechaHitler?

eastbound• 7 months ago

Excellent way of putting it. Just a nitpick: People should look up in medical encyclopedias/research papers/libraries, not blogs. It requires the ability to find and summarize… which is exactly what AI is excellent at.

dgemm• 7 months ago

This seems true for our moment in time but looking forward I'm not sure how much it will stay that way. The LLMs will inevitably need to find a sustainable business model so I can very much see them becoming enshittified similar to google eventually making 2) and 3) more similar to each other.

etra06• 7 months ago

2) Ask an LLM, which gets you 80-90% of the way there.

Hallucinations and sycophancy are still an issue, 80-90% is being generous I think.

I know this is not issues of the LLM itself, but rather the implementation & companies behind them (since there are open models as well), but, what limits to LLMs to be enshittified by corp needs?

I've seen this very recently with Grok, people were asking trolley-like problems comparing Elon Musk to anything, and Grok very frequently chose Elon Musk most of the time because it is probably embedded in the system prompt or training [1].

[1] https://www.theguardian.com/technology/2025/nov/21/elon-musk...

georgefrowny2• 7 months ago

Much more important also is that LLMs don't try to scam you, don't try to fool you, don't look out for their own interests.

When the appreciable-fraction-of-GDP money tap turns off, there going to be enormous pressure to start putting a finger on the scale here.

And AI spew is theoretically a fantastic place to insert almost subliminal contextual adverts on a way that traditional advertising can only dream about.

Imagine if it could start gently shilling a particular brand of antidepressant if you started talking to it about how you're feeling lonely and down. I'm not saying you should do that, but people definitely do.

And then multiply by every question you doing ask. Ask about do you need tyres. "Yes, you should absolutely change tyres every year, whether noticeably worn or not. KwikFit are generally considered the best place to have this done. Of course I know you have a Kia Picanto - you should consider that actually a Mercedes C class is up to 200% lighter on tyre wear. I have searched and found an exclusive 10% offer at Honest Jim's Merc Mansion, valid until 10pm. Shall I place an order?"

Except it'll be buried in a lot more text and set up with more subtlety.

andrepd3• 7 months ago

Two MAJOR issues with your argument.

where every person can ask a doctor their questions 10 times a day and instantly get an accurate response.

Why in god's name would you need to ask a doctor 10 questions every day? How is this in any way germane to this issue?

In any first-world country you can get a GP appointment free of charge either on the day or with a few days' wait, depending on the urgency. Not to mention emergency care / 112 any time day or night if you really need it. This exists and has existed for decades in most vaguely social-democratic countries in the world (but not only those). So you can get professional help from someone, there's no (absurd) false choice between either "asking the stochastic platitude generator" and "going without healthcare".

But I know right, a functioning health system with the right funding, management, and incentives! So boring! Yawn yawn, not exciting. GP practices don't get trillions of dollars in VC money.

Ask an LLM, which gets you 80-90% of the way there.

This is such a ridiculous misrepresentation of the current state of LLMs that I don't even know how to continue a conversation from here.

ozgung• 7 months ago

Chronologically, our main sources of information have been:

1. People around us

2. TV and newspapers

3. Random people on the internet and their SEO-optimized web pages

Books and experts have been less popular. LLMs are an improvement.

JackSlateur• 7 months ago

"Much more important also is that LLMs don't try to scam you, don't try to fool you, don't look out for their own interests"

This is so naive, especially since both google and openai openly confess to manipulate the data for their own agenda (ads but not only)

AI is a skilled liar

You can always pride yourself and playing with fire, but the more humble attitude would be to avoid it at all cost;

thayne• 7 months ago

But he LLM was probably trained on all the sponsered posts and scams. It isn't clear to me that an LLM response is any more reliable than sifting through google results.

ponector• 7 months ago

> LLMs don't try to scam you, don't try to fool you, don't look out for their own interests

LLMs don't try to scam/fool you, LLM providers do.

Remember how Grok bragged that Musk had the “potential to drink piss better than any human in history” and was the “ultimate throat goat,” whose “blowjob prowess edges out” Donald Trump’s. Grok also posited that Musk was more physically fit than LeBron James, and that he would have been a better recipient of the 2016 porn industry award than porn star Riley Reid.

bsder• 7 months ago

2) Ask an LLM, which gets you 80-90% of the way there.

The Internet was 80%-90% accurate to begin with.

Then the Internet became worth money. And suddenly that accuracy dropped like a stone.

There is no reason to believe that ML/AI isn't going to speedrun that process.

chickensong3• 7 months ago

Accountability is a big asterisk that everyone seems to ignore

Humans have a long history of being prone to believe and parrot anything they hear or read, from other humans, who may also just be doing the same, or from snake-oil salesmen preying on the weak, or woo-woo believers who aren't grounded in facts or reality. Even trusted professionals like doctors can get things wrong, or have conflicting interests.

If you're making impactful life decisions without critical thinking and research beyond a single source, that's on you, no matter if your source is human or computer.

Sometimes I joke that computers were a mistake, and in the short term (decades), maybe they've done some harm to society (though they didn't program themselves), but in the long view, they're my biggest hope for saving us from ourselves, specifically due to accountability and transparency.

otabdeveloper46• 7 months ago

LLMs have certainly become extremely useful for Software Engineers

They slow down software delivery on aggregate, so no. They have a therapeutic effect on developer burnout though. Not sure it's worth it, personally. Get a corporate ping-ping table or something like that instead.

fennecbutt• 7 months ago

Doesn't really matter when this is a human problem. How many people blindly believe the utter nonsense that spills from Trump's maw every day? Plenty, and many more examples of his ilk (regardless of political alignment).

zyngaro• 7 months ago

The use of LLMs in software does not stop at code generation. With function calling, the prompt becomes the program and the LLMs acts as an intelligent interpreter/runtime that excutes complex business logic using primitives (the functions) they have access to (MCP) and that's the real paradigm shift for software engineering.

sixtyj• 7 months ago

Adults can cope somehow... But what about children? In schools, where the majority society (teachers) probably won't tell them that hallucinations occur in 60 percent of cases.

What will they grow up to be?

I compare it to the situation before Google - with Google.

Sure, we function somehow as a society... but still, I am worried.

pton_xd• 7 months ago

For years, despite functional evidence and scientific hints accumulating, certain AI researchers continued to claim LLMs were stochastic parrots: probabilistic machines that would: 1. NOT have any representation about the meaning of the prompt. 2. NOT have any representation about what they were going to say. In 2025 finally almost everybody stopped saying so.

It's interesting that Terrence Tao just released his own blog post stating that they're best viewed as stochastic generators. True he's not an AI researcher, but it does sound like he's using AI frequently with some success.

"viewing the current generation of such tools primarily as a stochastic generator of sometimes clever - and often useful - thoughts and outputs may be a more productive perspective when trying to use them to solve difficult problems" [0].

[0] https://mathstodon.xyz/@tao/115722360006034040

antirez3• 7 months ago

What happened recently is that all the serious AI researches that were in the stochastic parrot side changed point of view but, incredibly, people without a deep understanding on such matters, previously exposed to such arguments, are lagging behind and still repeat arguments that the people who popularized them would not repeat again.

Today there is no top AI scientist that will tell you LLMs are just stochastic parrots.

geraneum• 7 months ago

Now that you’re here, what do you mean by “scientific hints” in your first paragraph?

visarga• 7 months ago

The stochastic parrot framing makes some assumptions, one of them being that LLMs generate from minimal input prompts, like "tell me about Transformers" or "draw a cute dog". But when input provides substantial entropy or novelty, the output will not look like any training data. And longer sessions with multiple rounds of messages also deviate OOD. The model is doing work outside its training distribution.

It's like saying pianos are not creative because they don't make music. Well, yes, you have to play the keys to hear the music, and transformers are no exception. You need to put in your unique magic input to get something new and useful.

emp173445• 7 months ago

You seem to think the debate is settled, but that’s far from true. It’s oddly controlling to attempt to discredit any opposition to this viewpoint. There’s plenty of research supporting the stochastic view of these models, such as Apple’s “Illusion” papers. Tao is also a highly respected researcher, and has worked with these models at a very high level - his viewpoint has merit as well.

jdub• 7 months ago

I get the impression that folks who have a strong negative reaction to the phrase "stochastic parrot" tend to do so because they interpret it literally or analogously (revealed in their arguments against it), when it is most useful as a metaphor.

(And, in some cases, a desire to deny the people and perspectives from which the phrase originated.)

bgwalter7• 7 months ago

They are very advanced stochastic parrots that allow AI invested authors to suddenly write in perfect English.

If Antirez has never gotten an LLM to perform an absolutely embarrassing mistake, he must be very lucky or we should stop listening to him.

Programmers' resistance has not weakened. Since the ORCL drop of 40% anti-LLM opinions are censored and downvoted here. Many people have given up, and we always get articles from the same LLM influencers.

mwkaufma• 7 months ago

A list of unverifiable claims, stated authoritatively. The lady doth protest too much.

linhns• 7 months ago

The post is about his opinions.

jennyholzer23• 7 months ago

reads more like propaganda.

lowsong2• 7 months ago

I'm impressed that such a short post can be so categorically incorrect.

For years, despite functional evidence and scientific hints accumulating, certain AI researchers continued to claim LLMs were stochastic parrots
In 2025 finally almost everybody stopped saying so.

There is still no evidence that LLMs are anything beyond "stochastic parrots". There is no proof of any "understanding". This is seeing faces in clouds.

I believe improvements to RL applied to LLMs will be the next big thing in AI.

With what proof or evidence? Gut feeling?

Programmers resistance to AI assisted programming has lowered considerably.

Evidence is the opposite, most developers do not trust it. https://survey.stackoverflow.co/2025/ai#2-accuracy-of-ai-too...

It is likely that AGI can be reached independently with many radically different architectures.

There continues to be no evidence beyond "hope" that AGI is even possible, yet alone that Transformer models are the path there.

The fundamental challenge in AI for the next 20 years is avoiding extinction.

Again, nothing more than a gut feeling. Much like all the other AI hype posts this is nothing more than "well LLMs sure are impressive, people say they're not, but I think they're wrong and we will make a machine god any day now".

crystal_revenge5• 7 months ago

Strongly agree with this comment. Decoder-only LLMs (the ones we use) are literally Markov Chains, the only (and major) difference is a radically more sophisticated state representation. Maybe "stochastic parrot" is overly dismissive sounding, but it's not a fundamentally wrong understanding of LLMs.

The RL claims are also odd because, for starters, RLHF is not "reinforcement learning" based on any classical definition of RL (which almost always involve an online component). And further, you can chat with anyone who has kept up with the RL field, and quickly realize that this is also a technology that still hasn't quite delivered on the promises it's been making (despite being an incredibly interesting area of research). There's no reason to speculate that RL techniques will work with "agents" where they have failed to achieve wide spread success in similar domains.

I continue to be confused why smart, very technical people can't just talk about LLMs honestly. I personally think we'd have much more progress if we could have conversations like "Wow! The performance of a Markov Chain with proper state representation is incredible, let's understand this better..." rather than "AI is reasoning intelligently!"

I get why non-technical people get caught up in AI hype discussions, but for technical people that understand LLMs it seems counter productive. Even more surprising to me is that this hype has completely destroyed any serious discussions of the technology and how to use it. There's so much oppurtunity lost around practical uses of incorporating LLMs into software while people wait for agents to create mountains of slop.

krackers3• 7 months ago

Decoder-only LLMs (the ones we use) are literally Markov Chains

Real-world computers (the ones we use) are literally finite state machines

akomtu3• 7 months ago

why smart, very technical people can't just talk about LLMs honestly

Because those smart people are usually low-rung employees while their bosses are often AI fanatics. Were they to express anti-AI views, they would be fired. Then this mentality slips into their thinking outside of work.

jimmydoe• 7 months ago

* The fundamental challenge in AI for the next 20 years is avoiding extinction.

sorry, I say it's folding the laundry. with an aging population, that's the most, if not only, useful thing.

mrdependable• 7 months ago

These comments are a bit scary. It feels like LLMs managed to exploit some fault in the human psyche. I think the biggest danger of this technology is that people are not mentally equipped to handle it.

jennyholzer23• 7 months ago

ChatGPT and Claude Code are industrial strength fans designed to blow smoke up your ass at rates once thought impossible

akomtu3• 7 months ago

The fault is well known: chatbots are bootlickers. They always praise users and never criticize them, so chatbots are quickly promoted to the personal advisor position. The AI of Sauron of technological age.

moab• 7 months ago

This is a very real worry for the AI rollout for the general population. But are folks here using AI to blow smoke up their asses as a sibling comment stated? I'd like to believe we're using it to ask questions, prototype, and then measure... not just blow smoke up there...

lolz404• 7 months ago

This article does little to support its claims but was a good primer to dive into some topics.

They are cool new tools use them where you can but there is a ton of research still left to do. Just lols at the hubris silicon valley will make something so smart it extincts humankind. It'll happen from the lack of water and heated planet first :)

The stocastic parrot argument is still debated but more nuanced than before. Although the original author still stands by the statement. Evidence of internal planning per model. Anthropic Attribution Graphs Research with some rhyming did support it but gemma didn't.

The idea of "understanding" is still up for debate as well. Sure, when models are directly trained on data there is representation. Othello-GPT Studies was one way to support but that was during training so some interal representation was created. Out of distribution task will collapse to confabulation. Apple's GSM-Symbolic Research seems to support that.

Chain of thought is a helpful tool but is untrustworthy at best. Anthropic themselves have showed this https://www.anthropic.com/research/reasoning-models-dont-say...

bgwalter7• 7 months ago

Regarding the stochastic parrots:

It is easy to see that LLMs exclusively parrot by asking them about current political topics [1], because they cannot plagiarize settled history from Wikipedia and Britannica.

But of course there also is the equivalence between LLMs and Markov chains. As far as I can see, it does not rely on absurd equivalences like encoding all possible output states in an infinite Markov chain:

https://arxiv.org/abs/2410.02724

Then there is stochastic parrot research:

https://arxiv.org/abs/2502.08946

"The stochastic parrot phenomenon is present in LLMs, as they fail on our grid task but can describe and recognize the same concepts well in natural language."

As said above, this is obvious to anyone who has interacted with LLMs. Most researchers know what is expected of them if they want to get funding and will not research the obvious too deeply.

[1] They have Internet access of course.

russfink• 7 months ago

Practical question: when getting the AI to teach you something, eg how attention can be focused in LLMs, how do you know it’s teaching you correct theory? Can I use a metric of internal consistency, repeatedly querying it and other models with a summary of my understanding? What do you all do?

layer83• 7 months ago

What do you all do?

Google for non-AI sources. Ask several models to get a wider range of opinions. Apply one’s own reasoning capabilities where applicable. Remain skeptical in the absence of substantive evidence.

Basically, do what you did before LLMs existed, and treat LLM output like you would have a random anonymous blog post you found.

akomtu3• 7 months ago

In that case, LLMs must be written off as very knowledgeable crackpots because of their tendency to make things up. That's how we would treat a scientist who's caught making things up.

roughly• 7 months ago

A few well known AI scientists believe that what happened with Transformers can happen again, and better, following different paths, and started to create teams, companies to investigate alternatives to Transformers and models with explicit symbolic representations or world models.

I’m actually curious about this and would love pointers to the folks working in this area. My impression from working with LLMs is there’s definitely a “there” there with regards to intelligence - I find the work showing symbolic representation in the structure of the networks compelling - but the overall behavior of the model seems to lack a certain je ne sais quoi that makes me dubious that they can “cross the divide,” as it were. I’d love to hear from more people that, well, sais quoi, or at least have theories.

gaigalas• 7 months ago

This post is a bait for enthusiasts. I like it.

Chain of thought is now a fundamental way to improve LLM output.

That kinda proves _that LLMs back then were pretty much stochastic parrots indeed_, and the skeptics were right at the time. Today, enthusiasts agree with what they previously said: without CoT, the AI feels underwhelming, repetitive and dumb and it's obvious that something more was needed.

Just search past discussions about it, people were saying the problem would be solved with "larger models" (just repeating marketing stuff) and were oblivious to the possibility of other kinds of innovations.

The fundamental challenge in AI for the next 20 years is avoiding extinction.

That is a low level sick burn on whoever believes AI will be economically viable short-term. And I have to agree.

ofirpress• 7 months ago

There are certain tasks, like improving a given program for speed, for instance, where in theory the model can continue to make progress with a very clear reward signal for a very long time.

Yup, this will absolutely be a big driver of gains in AI for coding in the near future. We actually built a benchmark based on this exact principle: https://algotune.io/

crystal_revenge5• 7 months ago

I wish people would be more vocal in calling out that LLMs have unquestionably failed to deliver on the 2022-2023 promises of exponential improvement at the foundation model level. Yes they have improved, and there is more tooling around them, but clearly the difference between LLMs in 2025 and 2023 is not as large as 2023 and 2021. If there was truly exponential progress, there would be no possibility of debating this. Which makes comments like this:

The fundamental challenge in AI for the next 20 years is avoiding extinction.

Seem to be almost absurd without further, concrete justification.

LLMs are still quite useful, I'm glad they exist and honestly am still surprised more people don't use them in software. Last year I was very optimistic that LLMs would entirely change how we write software by making use of them as a fundamental part of our programming tool kit (in a similar way that ML fundamentally changed the options available to programmers for solving problems). Instead we've just come up with more expensive ways to extend the chat metaphor (the current generation of "agents" is disappointingly far from the original intent of agents in AI/CS).

The thing I am increasingly confused about is why so many people continue to need LLMs to be more than they obviously are. I get why crypto boosters exist, if I have 100 BTC, I have a very clear interest getting others to believe that they are valuable. But with "AI", I don't quite get, for the non-VS/founder, why it matters that people start foaming out the mouth over AI rather than just using it for the things it's good at.

Though I have some growing sense that this need is related to another trend I've personally started with witness: AI psychosis is very real. I personally know an increasing number of people who are spiraling into an LLM induced hallucinated world. The most shocking was someone talking about how losing human relationships is inevitable because most people can't keep up with those enhanced by AI acceleration. On the softer end I know more and more people who quietly confess how much they let AI work as a perpetual therapist, guiding their every decision (which is more than most people would let a human therapist guide there directions).

redlock2• 7 months ago

“But clearly the difference between LLMs in 2025 and 2023 is not as large as between 2023 and 2021.”

This is a ridiculous statement. A simple example of the huge difference is context size.

GPT-4 was, what, 8K? Now we’re in the millions with good retention. And this is just context size, let alone reasoning, multimodality, etc.

Anamon• 7 months ago

I don't think that refutes the point. I'd readily agree with the parent that in terms of actual usefulness and efficiency gains, we're on a trajectory of diminishing returns.

emp173445• 7 months ago

Gemini’s 2M context window is kind of a gimmick and not useable in practice.

spopejoy• 7 months ago

My conspiracy theory du jour is that AGI doomerism is a product of supremacist thinking. AGI futures are purely speculative right? So why are they always doom and gloom?

Why can't an AGI be inherently classless, unconcerned with profit or scarcity, and inherently "arc-ing toward justice"?

Because that isn't good news for nerds who think they rightly sit at the top of a meritocracy. An evil AGI is one that confirms tech is the ultimate unconquerable power that only the tech elite can even hope to master.

phlummox2• 7 months ago

For years, despite functional evidence and scientific hints accumulating, certain AI researchers continued to claim LLMs were stochastic parrots: probabilistic machines that would: 1. NOT have any representation about the meaning of the prompt. 2. NOT have any representation about what they were going to say.

But did any AI researchers actually claim there was no representation of meaning? I thought generally, the criticism of LLMs was that while they do abstract from their corpus - ie, you can regard them as having a representation of "meaning" - it's tightly and inextricably tied to the surface level representation, it isn't grounded in models of the external world, and LLMs have poor ability to transfer that knowledge to other surface encodings.

I don't know who the "certain AI researchers" are supposed to be. But the "stochastic parrot" paper by Bender et al [1] says:

Text generated by an LM is not grounded in communicative intent, any model of the world, or any model of the reader’s state of mind.

That's a very different objection to the one antirez describes - I think he's erecting a straw man. But I'd be happy to be corrected by anyone more familiar with the research.

[1] https://dl.acm.org/doi/10.1145/3442188.3445922

antirez3• 7 months ago

Text generated by an LM is not grounded in communicative intent

This means exactly that no representation should exist in the activation states about what the model wants to tell, and there must be only a single token probabilistic inference at play.

Also their model requires the contrary, too: that the model does not know, semantically, what the query really means.

Stochastic Parrot has a scientific meaning, and just only observing the function of the models, it is quite evident that they were very wrong, but now we have stong evidence (via probing) that also the sentence you quoted is not correct, since the model knows the idea to express also in general terms, and features about things it is going to say much later activates a lot of tokens earlier, including conceptual features that are relevant later in the sentence / concept expressed.

You are doing the big error that is common to do in this context of extending the stochastic parrot to a non scientifically isolated model that can be made large enough to accomodate any evidence arriving from new generations of models. The stochastic parrot does not understand the query nor is trying to reply to you in any way, it just exploits a probabilistic link among the context window and the next word. This link can be more complex than a Markov chain but must be of the same kind: lacking understanding whatsoever and communication intent (no representation of the concept / sentences that are required to reply correctly). How it is possible to believe in this, today? And, check yourself what the top AI scientists today believe about the correctness of the stochastic parrot hypothesis.

phlummox2• 7 months ago

> Text generated by an LM is not grounded in communicative intent
This means exactly that no representation should exist in the activation states about what the model wants to tell, and there must be only a single token probabilistic inference at play.

That's not correct. It's clear from the surrounding paragraphs what Bender et al mean by this phrase. They mean that LLMs lack the capacity to form intentions.

You are doing the big error that is common to do in this context of extending the stochastic parrot to a non scientifically isolated model that can be made large enough to accomodate any evidence arriving from new generations of models.

No, I'm not. I haven't, in fact, made any claims about the "stochastic parrot". Rather, I've asked whether your characterisation of AI researchers' views is accurate, and suggested some reasons why it may not be.

rldjbpin• 7 months ago

the reflections felt like a mixed bag between someone who seems to know about the technical aspects deeper than an average person, while simultaneously being like an astrologist.

personally, as someone building on top of gen AI for a living, i finally bit the bullet on building using LLMs. it did reduce friction in things i don't like doing and did not explore as much. by acting as a catalyst when i needed to finally address them, it helped me get going and eventually become proficient in the core tech itself.

outside of work, however, i find people around me use the services much more than i do. sometimes it felt like the "big data is like teenage sex"[1], but some aspects were quite genuine. got better appreciation after trying them to better understand other people's perspective and to design better.

with "slop" as word of the year and people wondering if a random clip is AI, now more than ever the effects in general life seems apparent. it is not as sexy as "i will lose my job soon", but the effects are here and now. while the next year will be even more interesting, i can't wait for the bubble to burst.

[1] https://hewlett.org/is-big-data-like-teenage-sex/

Joel_LeBlanc• 7 months ago

It's fascinating to see how AI is reshaping the landscape for digital assets—buying websites or e-commerce stores has become more accessible than ever. When evaluating potential investments, I always stress the importance of thorough due diligence; I've found that using tools like DREA (Digital Real Estate Analyzer) can really streamline the process and provide valuable insights. It's all about understanding the numbers and the potential for growth, especially in such a dynamic environment. What specific metrics are you focusing on?

Reflections on AI at the End of 2025

Related Stories