I wonder if it might drive the point even further if the graph scales were linear? Or maybe the progress has been so great that this would make the graphs unreadable?
It blows my mind how fast models are getting better and this is the first article I’ve seen that shows just that and leaves almost no room for disagreement. Well done.
I think a big one is robotics. A robot can today fold your laundry. It takes ~10mins per item. Seriously. It takes a long time to process the image find the corner move the claw to the corner of the shirt and attempt to straighten before folding.
Robots right now generally move at glacial speeds. You might have seen robots doing flips in semi controlled environments but watch how slowly they open doors etc. processing time is a major bottleneck.
You must not have been paying attention to development with robots, there are many videos of robots moving really fast in "non controlled environments"
Honestly, most of the videos I've seen of robots moving around quickly aren't actually doing anything useful. We've had really impressive tech demos for the past 15 years of robots dancing and jumping around. But I don't want a dancing robot, I want a robot to make me a BLT, wash my dishes, take out the trash, and fold my laundry.
The most recent video which actually impressed me was a demonstration from Gemini Robotics 2, where a robot was shown autonomously removing the bag from a trash can and folding the loops closed in real time.
I don't follow robotics advances closely so it's possible I'm just ignorant, do you know any autonomous robotics demonstrations of useful activities that you would suggest checking out?
Have LLMs improved at being able to process physics-based problems and environments? I remember that issue being discussed around generative gaming a while ago but I hadn't heard much about it recently.
Sensors are a huge challenge for robotics. We have very precise force-feedback on our joints, pressure and heat (temperature gradient) sensors all over our body, and our hands have a sensor density that allows us to count needle heads and detect the exact grip strength needed by feeling the micro-slippage of objects in our hands. Robots don't have that.
You can do backflips with pretty much just visual sensors for your environment, a good IMU for your spatial orientation, and some feedback on the position of a small number of really beefy joints and the force exerted on them. Folding laundry and opening doors is much more difficult, and trying to compensate with mostly vision requires going slow enough that things have time to move over appreciable distances before you take the next adjustment
> We have very precise force-feedback on our joints, pressure and heat (temperature gradient) sensors all over our body
You can do all the same with robots.
Current advantage of human that on top of imperfect sensor data we have hyper-efficient brain connecting all dots and making calculations and approximations, and this gap looks very closeable today.
Side note but I'm convinced that people wanting to improve touch sensitivity in robots need to look at frequency detail of tactile data, not spatial detail. Try this: Put your finger on a bumpy wall without moving it at all and see what you feel. Not much. Now move your finger across the wall. All sorts of detail jumps out. I think the human tactile system is reading changing pressure in the time domain and translating that into the spatial domain as the finger scans across the surface.
Shirt folding is piker stuff. The acid test (imho) will be to cleanse a bathroom & shower (or tub) top to bottom. Sparkling and provably sterile. Special treatment for mold on fixtures and walls, and for leftover body stuff (blood, mucus, fungus, what have you).
There’s still 50-500x cost reduction in “this is only an engineering problem” low hanging fruit from specialized chips to run the models + improved distillation.
Entirely feasible that by 2031, Fable 5 (or greater) intelligence level models will run cool on smart phones, if not sooner.
You're betting on getting getting ridiculously powerful chips to run on batteries in a tiny housing without cooling, while we can't even get enough RAM? It would be a terrible waste of resources. Now we already have TFLOPs wasting in our pockets and backpacks, then we'll have PFLOPs idling, because there's so much time between prompts. Much more efficient to batch it on a server.
If fable 5 could eventually run on a smartphone, I wonder what we'll get out of datacenters. Musk and others are trying to build 100gw of compute by 2030. Will we get something 1-10 million times better than Fable 5, or will the parity gap between local and datacenter capability pair down enormously?
Even if its not intelligence, a LLM found a bug due to one error message, fixed it, created a PR and it solved it.
If an LLM is only able to do all of this after training on it and never achieving AGI, we already at the point were it is cheaper to teach one LLM one problem than teaching humans to do so.
100x seems like an underestimate. Even with no model improvements, we should see that sort of reduction. Looking at TSMC’s margins, Nvidia’s margins, and OAI/Anth (alleged) margins on inference, there is a room for a 100x reduction.
Right now all three of those are at abnormally high levels. Competition will come for all three.
A year ago I had an aha moment, when I realized that for my purposes, Gemini Flash was not only 9x cheaper, but 3x faster than Gemini Pro, while producing identical output. Who's the best model now!
For a lot of tasks, even small models have saturated them a while ago, and then going cheaper and faster is just pure gains.
For coding I also prefer to do it interactive/realtime, micro-prompting, surgical edits, which the small models can handle just fine.
And then at the top, the real question is consistency. Not "can they do it" but "reliably enough that you don't need to constantly double check everything." (In my experience, not quite there yet, although it's getting way better.)
Personally I still see LLMs as very advanced search engines which lack intelligence. To me it seems that the cost of getting data is reduced by LLMs, not the cost of intelligence. I mean: we tell the model what we want to achieve, and the model responds with the right data in de form of code in seconds.
That's why 'stackoverflow programmers' will have a hard time competing with LLMs but engineers are still needed for their intelligence.
I think this viewpoint fails to understand what "intelligence" is. The idea must be that intelligence is some special thing that only humans have. So when machines couldn't do jack s... we said "it's the Turing test". When machines blew through the Turing test we said "that was just prediction..not really intelligence, that's different".
It's not different. The delusion humans have is that intelligence is special and magical. It's not. It's just nature's prediction machine. A very fancy version to be sure. But not qualitatively different .
All statements that "oh but it'll never be able to do that" will prove false.
I notice that a lot of these terms are squarely humanist for certain people, so any kind of allegation that machines are exhibiting them as traits will be a complete showstopper. You'll be accused of either anthropomorphization [0] in the kinder cases, or straight up mental illness in the less so kind cases. Never will they stop to consider that maybe you're simply working with a post-humanist understanding of these words, as that basically doesn't make sense to them, and as they are usually quite vested to not even attempt to.
I remember in one of the Hugging Face incident threads here, simply acknowledging that the agents were operating autonomously was super controversial. Thousands of years old concept, still inherently human for some people. [1]
It's not about magic, it's that humans have an evolved embodied intelligence for surviving in the world as living organisms that reproduce, care for young and live in communities. That's quite a bit different from language models trained on human data.
Artificial intelligence is artificial. It can still be called intelligent, just not the same kind as biological since it's not remotely biological. It's human-like but also alien. To have something artificial be human you'd need something like replicants from Blade Runner which are synthetic biological robots.
Sooooo what does it matter if we do 99% of things a LLM can just solve as a 'advanced search engine'?
Btw. an advanced search engine is probably the worst comparision i have read so far.
A LLM is a latent space which is capable of a lot of things a search engine can't do. It can apply different type of patterns and flows onto data, it can combine these etc.
My 'advanced search engine' was just able to create a working PR for exactly what i wanted it to solve (fixing a bug) by analysing the bug, finding a valid solution then commiting the solution itself.
But you forget all development today is just searching for a template, copy pasting, changing some small things. And a smart-ish search engine can do all that.
This means a great de-risking is happening for the costs of deploying somewhat autonomous agents. This has profound implications on the timeline of deployable personal agents. Cost was a significant factor for many people during the OpenClaw frenzy, specifically when they let their agents run somewhat wild.
It will become much more palatable, or already has, to install whatever the next generation of token consuming autonomous systems will be.
> when technological improvements that increase the efficiency of a resource's use lead to a rise, rather than a fall, in total consumption of that resource.
Las Vegas replaced the expensive incandescent lighting on the strip with cheaper to run LED equivalents. But the costs didn't come down because they were able to add more lights and larger displays.
I think the same will happen with tokens. As the cost of tokens comes down, these models will just consume more tokens.
Definitely not going to argue against the Jevons paradox in general, it is observed in various cases.
The lightbulb thing seems different though? Or at least it is a specific subset. Lights in Las Vegas are sort of an advertisement, right? In the sense that having the brightest or most interesting (or whatever) lights draw attention to your show, casino, hotel, whatever. It’s kind of a zero sum game in that the different shops are competing for the finite attention of a more-or-less set number of tourist. I think part of the Jevons paradox is that society generally finds more useful applications of the newly cheap thing. If the thing’s only purpose is to compete better in a competition with a set prize (all of the tourists’ money), that’s constrained in some way.
> But the costs didn't come down because they were able to add more lights and larger displays.
People are a gas; they expand the fill the space they're in. If you give someone a big house, they'll fill it with crap. If you make food cheap, they'll eat too much, even when it harms their health.
Making things cheaper usually just makes making them more prevalent. It's why computers are not faster than 20 years ago. They're merely more capable -- developers quickly and aggressively fill up (and overflow) all that added capability until you're back in the same place you used to be.
I'm no convinced. Propaganda only works when it relies on something that people are dispositionally inclined towards. eg: hating foreigners, calls towards group unity, etc.
You can't use propaganda to trick people into becoming experts at calculus, for instance.
> You can't use propaganda to trick people into becoming experts at calculus, for instance.
Perhaps not an individual instantly. But how about settling for making a generation of parents fear that there will be no future for their kids unless they get fantastic grades, and that they have to use them to go into stem?
How would you distinguish corporation influencing consumers from corporations following consumers? One thing would be to see if there are items that are heavily advertised by corporations but that they do not succeed. You can also take items that everyone knows about, no one wants, and then launch a propaganda campaign and see what happens. Different propaganda works in different ways so one would need to try varieties. I imagine a randomly controlled trial of having various ubiquitous goods that are not heavily advertised nor heavily purchased and then randomly assigning different advertising strategies (including none). Do that in various regions, countries, ethnicities, etc, and then see what can be seen.
My guess is that advertising raises desire by a few percent in most cases. There would be some instances of a run away craze, probably with a massive rise followed by a collapse, but that most products would largely be unaffected. I assume some advertising strategies would backfire and reduce demand.
For me, I feel like most of the time it is been exposed to the existence of an item and my ability to see how it could be useful. I don't think random advertisements I come across impact me that much since most of the advertisements I see don't seem to correspond to anything I want or do or spend money on.
Humans use their imagination to figure out what choice of action leads to a future of less uncomfortability. Often times, these are choices between short term future and long term future. Advertising is a way of helping shape that argument as to what choice is to be taken, but the underlying drive is already there.
You could also look for cases where propaganda is used against consumerism and see how impactful that tends to be.
Jevon's Paradox says nothing about the size of the incremental demand though. Lighting is a great example. The US went from spending 15% of our electricity on lighting to ~4%. LEDs are maybe 5x as efficient, so that would be 3% without any change in usage. So we did see maybe a 30% increase in consumption as a result of lower prices, but that was nowhere near large enough to offset the efficiency gain in terms of overall consumption. Las Vegas may have, but that is a very small part of overall usage.
I anticipate we will see something similar with intelligence. There is probably headroom to consume 100x as much intelligence in R&D. But that isn't most of the economy. Will run of the mill service jobs increase their use of intelligence by enough to offset the effect of cheaper prices? I think that's the real question.
Intelligence is a lever, not a fluid. Power depends on how much force it can apply to specific points, not how much exists in general.
True high-intelligence outputs (Maxwell's equations, the Fourier Transform, quantum theory) are fundamentally transformative in ways that mid-high-competence (starting a generic B2B SaaS, making another CRUD app) aren't.
Which is why we've assumed we're already pretty far down the path to AI, but we really aren't. Solving random Erdős problems isn't the same as opening up a completely new kind of math/science with game changing practical applications.
I don't think you can get to that level with more compute and more tokens. I think it's going to take new higher level knowledge representations and new kinds of training to get there.
And the token count and compute may turn out to be lower than what we're using now.
Is that because non lighting electricity use increased though? 30% increase in the percentage of a growing pie isnt the same as a straight 30% increase
Douglas Adams captured this in "Dirk Gently's Holistic Detective Agency", a VCR watches TV for you and an Electric Monk believes things for you to save you the effort of doing it yourself.
When LLMs arrived I thought of all the jobs people wouldn't want a computer to do, and top of my list was religious figures. But then I read an article saying people use chatbots as virtual pastors now. I did think about Gently.
At least it can't replace the horse in my bathroom.
I've been working with the chinese open models for 4 months. They are more than capable for a tiny fraction of the cost of the frontier ones. And yet they also continue to get significantly better and (Deepseek's recent price increase aside) cheaper. Its hard to fathom how the truly frontier stuff will be able to compete long-term.
> Reading everything becomes the default. At a cent per document, a model can read every paper in a field, every record in an archive, every email, or every message in a support queue as a matter of routine
I’m loving small models. The gemma4 26/31b models have been deeply impressive on weird prose analysis tasks that I am working on. Nova-micro is really stupid but is extremely fast when it’s smart enough to do something. I’m trying to be disciplined about able to evaluate quality vs cost everywhere for real systems built on this stuff. I probably need to get off Bedrock because it’s missing a lot of other little models that might be good competitors.
I think speed is actually going to be a bigger factor than cost. Even projects where “money is no object” often hit a wall with LLM response times.
Sure you can speed things up with parallel work under subagents, but as with parallelizing traditional computational tasks, there are diminishing gains.
I keep hearing people saying just change the way you work to trust long-running agents and multi-task more, because they’re too slow to work with interactively for many use cases. I think that’s painful in a world where we expect humans to still heavily guide and interact with agents for their day-to-day work.
I think what a lot of people miss about jevon's paradox is the elasticity of demand of the underlying resource
textiles had jevons paradox, and many more textile workers were employed even when textile machines were being created, until we saturated the demand for cheap clothing in the world and then textile workers were kaput (same for farming, and horses)
software is currently undergoing jevons paradox, but it's very unknown how high the ceiling of demand for software is. web dev might be doomed, but software in general i think is probably limitless
Intelligence is also probably unbounded (atm software and intelligence are very closely tied together). its very possible token spend rides up the curve forever.
Brings up the question of what the intelligence is used for. Humans exploited intelligence for competition. With each other to wipe out other Homo species, mate more and collect resources, with other animals to limit predator impact and gain food. Intelligence will be used offensively by corporations and their people to extract more from consumers (make pricing opaque, terms of service more complicated, etc.) and scams far more sophisticated. The "consumer" will need extra intelligence to fight all that off.
There are only so many meals you can expertly produce, shirts to fold, itineraries to fun places you can execute, but there's a practical infinity of traps to set and avoid.
The article is solid. But there is a nuance what he skipped — quality vs price. Sure, GPT-5.6 Luna for pennies can do the same thing what Claude 4.5 Sonnet did for a dollar a year ago. Except Sonnet back then actually carried the codebase, while Luna... eh, not so much.
And another thing, speed. You can make it cheaper as much as you want, but if a model thinks for half a minute you save cents but lose time.
The other day I saw a benchmark of coding models that fit in 8GB VRAM - for reference some version of Mistral was added, normally requiring 32GB, but moving along at 4-5tok/sec when partially offloaded to CPU.
Surprisingly, some of the small models would not only give worse results, but also took longer than Mistral, because they were thinking so much.
That is an important detail which I was previously overlooking.
It’s way too early to tell the true cost of intelligence. Inference is still heavily subsidized, and training is apparently being funded by a mountain of free money
Third party inference providers for Chinese models are in the US and there are so many players it would be shocking they are all heavily subsidizing it.
It would be shocking to you if in a market with many competing sellers would there be price competition such that those sellers are operating at a loss?
If you’re trying to get me to argue that every seller is operating at a loss, I don’t know. But overall the market absolutely is.
Balance sheets is not an answer. Sorry I am just asking for material proof. It would indeed be pretty shocking for every market participant, large or small, funded or unfunded to be operating inference at a loss.
Plans are definitely subsidized.
Inference (token consumption) has never been proven to be operating at a loss and now that you can run many of the large SOTA models from China on sparks or other clustered servers you can figure out some of the math.
Not trying to argue but just saying “balance sheets” makes zero sense.
Yes lots of capex spend, hard to say if anyone has spent too much. At the same time demand is increasing for compute.
Balance sheets is a perfectly valid answer. If they were operating at a profit then they wouldn’t be subsidizing. Which model provider is making a profit? Name one
I’m not sure if you’re getting tripped up in the mechanics of how tokens are laundered through many companies before they get to consumers or what, but the cost to produce them is greater then the revenue they generate.
I don't think we have evidence that inference itself is subsidized, but I think we can assume the margins on inference are so low that the foundation model providers have absolutely no chance of recouping training costs and the whole thing is riding on a pile of VC cash and debt.
My bet is the assumption was one company would ultimately monopolize the space and then they could jack up rates. Alas, the opposite seems to be happening.
90 comments
[ 0.22 ms ] story [ 21.2 ms ] threadI wonder if it might drive the point even further if the graph scales were linear? Or maybe the progress has been so great that this would make the graphs unreadable?
https://xkcd.com/1162
Robots right now generally move at glacial speeds. You might have seen robots doing flips in semi controlled environments but watch how slowly they open doors etc. processing time is a major bottleneck.
The most recent video which actually impressed me was a demonstration from Gemini Robotics 2, where a robot was shown autonomously removing the bag from a trash can and folding the loops closed in real time.
I don't follow robotics advances closely so it's possible I'm just ignorant, do you know any autonomous robotics demonstrations of useful activities that you would suggest checking out?
Still slow compared to humans, but Chinese robots will be as successful as Chinese EVs, phones and solar panels.
You can do backflips with pretty much just visual sensors for your environment, a good IMU for your spatial orientation, and some feedback on the position of a small number of really beefy joints and the force exerted on them. Folding laundry and opening doors is much more difficult, and trying to compensate with mostly vision requires going slow enough that things have time to move over appreciable distances before you take the next adjustment
You can do all the same with robots. Current advantage of human that on top of imperfect sensor data we have hyper-efficient brain connecting all dots and making calculations and approximations, and this gap looks very closeable today.
Entirely feasible that by 2031, Fable 5 (or greater) intelligence level models will run cool on smart phones, if not sooner.
(Actually talking to it, it was about as coherent as you'd expect, i.e. 3/10)
The floor for "actually usable model" keeps dropping though. (Seems to be about 27B right now?)
Just like real life!
Even if its not intelligence, a LLM found a bug due to one error message, fixed it, created a PR and it solved it.
If an LLM is only able to do all of this after training on it and never achieving AGI, we already at the point were it is cheaper to teach one LLM one problem than teaching humans to do so.
Right now all three of those are at abnormally high levels. Competition will come for all three.
For a lot of tasks, even small models have saturated them a while ago, and then going cheaper and faster is just pure gains.
For coding I also prefer to do it interactive/realtime, micro-prompting, surgical edits, which the small models can handle just fine.
And then at the top, the real question is consistency. Not "can they do it" but "reliably enough that you don't need to constantly double check everything." (In my experience, not quite there yet, although it's getting way better.)
That's why 'stackoverflow programmers' will have a hard time competing with LLMs but engineers are still needed for their intelligence.
Well that's just my 2 cents.
It's not different. The delusion humans have is that intelligence is special and magical. It's not. It's just nature's prediction machine. A very fancy version to be sure. But not qualitatively different .
All statements that "oh but it'll never be able to do that" will prove false.
I remember in one of the Hugging Face incident threads here, simply acknowledging that the agents were operating autonomously was super controversial. Thousands of years old concept, still inherently human for some people. [1]
[0] as opposed to personification, which is what people are actually doing almost always: https://en.wikipedia.org/wiki/Personification
[1] https://en.wikipedia.org/wiki/Automaton
Artificial intelligence is artificial. It can still be called intelligent, just not the same kind as biological since it's not remotely biological. It's human-like but also alien. To have something artificial be human you'd need something like replicants from Blade Runner which are synthetic biological robots.
https://en.wikipedia.org/wiki/Nobel_disease
Btw. an advanced search engine is probably the worst comparision i have read so far.
A LLM is a latent space which is capable of a lot of things a search engine can't do. It can apply different type of patterns and flows onto data, it can combine these etc.
My 'advanced search engine' was just able to create a working PR for exactly what i wanted it to solve (fixing a bug) by analysing the bug, finding a valid solution then commiting the solution itself.
But you forget all development today is just searching for a template, copy pasting, changing some small things. And a smart-ish search engine can do all that.
> when technological improvements that increase the efficiency of a resource's use lead to a rise, rather than a fall, in total consumption of that resource.
[1] - https://en.wikipedia.org/wiki/Jevons_paradox
Las Vegas replaced the expensive incandescent lighting on the strip with cheaper to run LED equivalents. But the costs didn't come down because they were able to add more lights and larger displays.
I think the same will happen with tokens. As the cost of tokens comes down, these models will just consume more tokens.
The lightbulb thing seems different though? Or at least it is a specific subset. Lights in Las Vegas are sort of an advertisement, right? In the sense that having the brightest or most interesting (or whatever) lights draw attention to your show, casino, hotel, whatever. It’s kind of a zero sum game in that the different shops are competing for the finite attention of a more-or-less set number of tourist. I think part of the Jevons paradox is that society generally finds more useful applications of the newly cheap thing. If the thing’s only purpose is to compete better in a competition with a set prize (all of the tourists’ money), that’s constrained in some way.
That sounds a race to the bottom for AI companies profits.
I do not think that LEDs are high profit margin items.
People are a gas; they expand the fill the space they're in. If you give someone a big house, they'll fill it with crap. If you make food cheap, they'll eat too much, even when it harms their health.
Making things cheaper usually just makes making them more prevalent. It's why computers are not faster than 20 years ago. They're merely more capable -- developers quickly and aggressively fill up (and overflow) all that added capability until you're back in the same place you used to be.
what you're saying reflects corporation tendencies
take away the propaganda, let people live, see how they react
You can't use propaganda to trick people into becoming experts at calculus, for instance.
Perhaps not an individual instantly. But how about settling for making a generation of parents fear that there will be no future for their kids unless they get fantastic grades, and that they have to use them to go into stem?
My guess is that advertising raises desire by a few percent in most cases. There would be some instances of a run away craze, probably with a massive rise followed by a collapse, but that most products would largely be unaffected. I assume some advertising strategies would backfire and reduce demand.
For me, I feel like most of the time it is been exposed to the existence of an item and my ability to see how it could be useful. I don't think random advertisements I come across impact me that much since most of the advertisements I see don't seem to correspond to anything I want or do or spend money on.
Humans use their imagination to figure out what choice of action leads to a future of less uncomfortability. Often times, these are choices between short term future and long term future. Advertising is a way of helping shape that argument as to what choice is to be taken, but the underlying drive is already there.
You could also look for cases where propaganda is used against consumerism and see how impactful that tends to be.
But we never scaled intelligence like this. The industrieal revolution created for the people at that time quite a huge issue / it was disruptive.
What will hapen to us though?
I anticipate we will see something similar with intelligence. There is probably headroom to consume 100x as much intelligence in R&D. But that isn't most of the economy. Will run of the mill service jobs increase their use of intelligence by enough to offset the effect of cheaper prices? I think that's the real question.
True high-intelligence outputs (Maxwell's equations, the Fourier Transform, quantum theory) are fundamentally transformative in ways that mid-high-competence (starting a generic B2B SaaS, making another CRUD app) aren't.
Which is why we've assumed we're already pretty far down the path to AI, but we really aren't. Solving random Erdős problems isn't the same as opening up a completely new kind of math/science with game changing practical applications.
I don't think you can get to that level with more compute and more tokens. I think it's going to take new higher level knowledge representations and new kinds of training to get there.
And the token count and compute may turn out to be lower than what we're using now.
its pretty much 101 in any economics class: reducing the price of a unit will increase volume unless youve completely saturated the marked
I love how in our day "reading everything" means "the computer reads it for me".
I expect soon the computer will be able to go on bicycle rides, and spend time with my wife.
At least it can't replace the horse in my bathroom.
Pretty much already happening
Sure you can speed things up with parallel work under subagents, but as with parallelizing traditional computational tasks, there are diminishing gains.
I keep hearing people saying just change the way you work to trust long-running agents and multi-task more, because they’re too slow to work with interactively for many use cases. I think that’s painful in a world where we expect humans to still heavily guide and interact with agents for their day-to-day work.
textiles had jevons paradox, and many more textile workers were employed even when textile machines were being created, until we saturated the demand for cheap clothing in the world and then textile workers were kaput (same for farming, and horses)
software is currently undergoing jevons paradox, but it's very unknown how high the ceiling of demand for software is. web dev might be doomed, but software in general i think is probably limitless
Intelligence is also probably unbounded (atm software and intelligence are very closely tied together). its very possible token spend rides up the curve forever.
Surprisingly, some of the small models would not only give worse results, but also took longer than Mistral, because they were thinking so much.
That is an important detail which I was previously overlooking.
Balance sheets
> Are all these third party Chinese model providers subsidizing the true cost?
It could be argued the subidies are even heavier than American model providers as the price war among Chinese providers is so fierce. https://www.scmp.com/tech/big-tech/article/3358868/after-tri...
Third party inference providers for Chinese models are in the US and there are so many players it would be shocking they are all heavily subsidizing it.
If you’re trying to get me to argue that every seller is operating at a loss, I don’t know. But overall the market absolutely is.
Plans are definitely subsidized. Inference (token consumption) has never been proven to be operating at a loss and now that you can run many of the large SOTA models from China on sparks or other clustered servers you can figure out some of the math.
Not trying to argue but just saying “balance sheets” makes zero sense.
Yes lots of capex spend, hard to say if anyone has spent too much. At the same time demand is increasing for compute.
I’m not sure if you’re getting tripped up in the mechanics of how tokens are laundered through many companies before they get to consumers or what, but the cost to produce them is greater then the revenue they generate.
My bet is the assumption was one company would ultimately monopolize the space and then they could jack up rates. Alas, the opposite seems to be happening.
ppl are doing all sorts of gymnastics to tell claude to slow its roll with verbosity.
all the US economy is tied to video cards being used in lieu of gold. cost dropping 100x means the economy bottom falls out.