Almost all statements Yann LeCun made back then (LLMs won't be able to do this or that) turned out to be wrong. They were right to fire him.
10% chance to be market leader doesn’t sound as good as 60% second place to them for example
It's his call for sure, but rather funny as an argument.
It is because even if it would pan out, it would mean temporarily losing AI leadership, reducing stock price, losing the ability to hire talent, and by virtue of that losing the opportunity permanently.
He was against LLMs, reinforcement learning, etc. Claiming that those are a dead-end. He was wrong. His own "JEPA" ideas have basically gone nowhere. Meta was right in moving on without him.
Zuckerberg wanted to placate his shareholders with a shiny new thing now.
would've stopped Meta from delivering a polished product like Muse sooner.
Muse is not polished, unless usability and programming standards have fallen so far in the LLM vibe coding era that a pre-alpha state is now considered polished.
Muse is imperfect, as is all software, but calling it pre-alpha is a bit much.
But even in computing, when there is a promising way forward that the basic reaearch has been done on, its hard to justify side quests.
...And because they have monopoly money. Basically, research labs are usually vanity spending for corporations with too much money. Xerox parc was funded by a monopoly. Bell labs was funded by a monopoly. Google funds its moonshots with its monopoly.
How does that work when the research team moved elsewhere? Re-hire talent? That takes time - the 1 thing you don't have in a fast-moving field like AI. Post sums it up nicely:
"Do the math—how much is it worth to halve training and inference cost given that the industry has investments approaching a trillion dollars?"
How much is these researchers' salaries, compared to what MSFT spends on compute? Chances are it's peanuts in comparison.
Companies will try to ruthlessly rid themselves of labor as long as those members of labor have agency and rights.
If someone disagrees they can convince me by explaining why a company wouldn’t choose to spend X on a machine they wholly own vs X on an employee they don’t.
There’s some room for argument that employees can do more than machines can but in the current AI zeitgeist arguing that we have machines with equivalent to human intelligence I don’t see how you wouldn’t see this coming if you believed the zeitgeist.
In a way this suggests that MS doesn't aim to be a leader in AI <the core tech>. Only in its applications + related services (read: profit to be had there).
Nothing wrong with that btw. But understandable that fundamental researchers then pack up & leave.
Ah, well, good thing we've almost entirely defunded academic science.
Seems weird when it's become apparent that LLMs aren't really special sauce.
The author states "I consider the scaling-only focus to be both misdirected and not fun.". I get the point but also totally get the focus on scale out for now.
For a corporation it is to make money. For a researcher it is to explore and illuminate. That exploration and illumination often leads to a landfall of new wealth generation is a happy accident that ensures that the monies are going to bet some of their capital in them, at least the wise ones.
Which is superior, the big f*king monstrosity or the elegant on-demand, small batches, reduce waste?
Meta will go in the shitter and Deepseek will make the small batches.
It does (Lean at least) incorporate controlled experimentation, but tightly controlled with a both a laser focus on what the goal is and tight resource constraints. And Lean (at least several writers on Lean) favors taking multiple paths with a a quick and low risk and a longer-payoff, higher-risk, higher reward path for some efforts—one source used the US Navy’s two-path approach to SLBMs with Polaris and Poseidon as a model for this—but that's different than “maybe there is something down this path” basic research, which is outside of the concern a production model like TPS/Lean and often completely outside the model of a product-oriented firm; basic research is something firms consume and incorporate more than generate, because until the basics have been proven out, the rationally perceived chance of any particular avenue paying off is near zero for any firm.
Meta will go in the shitter and Deepseek will make the small batches.
RCA developed the first good color TV cameras. They were monsters. Originally, electronic TV (as opposed to mechanical scanning) used Zworykin's iconoscope. This needed huge amounts of light. It didn't have a photo-multiplier stage to boost sensitivity, but it did accumulate charge from light over the whole frame time. There was an alternative approach, Farnsworth's image dissector, which also needed huge amounts of light. It had a photo-multiplier stage but only the light from the dot currently being scanned by the scanning beam mattered.
So RCA threw money at the problem. They developed the image orthicon camera tube. It combined the advantages of both approaches, integration over a full frame and a photo-multiplier stage. It was insanely expensive, costing thousands of dollars in the 1950s. It needed about twenty different voltages and multiple independent power supplies. It worked reasonably well.
Then came color. That took three image orthicons (RGB) inside a box the size of a small refrigerator, along with optics to split the incoming image into three colors. Plus a rack of analog electronics per camera, connected through a large cable. Plus an alignment problem from hell, with a huge number of adjustment knobs.[1]
Scaling worked! From the 1950s to the 1970s, color TV broadcasting was powered by those monsters.
Now, independent of this, other people were working on other approaches. Vidicons, silicon detectors inside camera tubes, and finally fully solid state image sensors. Eventually, in 1983, after thirty years of image orthicon monsters, RCA produced a usable solid state camera. Today, a cell phone camera costs about US$10 in bulk, and it's far better than the monsters.
It took thirty years of low-priority development to get past the giant image orthicon cameras.
Electronic fluorescent lamp ballasts had a different problem. That was a suppressed invention. The magnetic ballast industry was dominated by Magnetek. Someone invented a solid state replacement, which was smaller and cheaper, and licensed their patent to Magnetek for royalties. Magnetek just sat on it, not wanting to gut their core market. The inventor hired Townsend, Townsend and Crew in Palo Alto and sued. (Read this in the law firm's brag book in their office.) Eventually, the inventor got a hundred million or so, solid state ballasts went into production, and Magnetek seems to be gone. (The company called Magnetek today, which makes crane controls, is an offshoot of a Litton unit, and seems to be unrelated.)
There are also parallels from the steam locomotive industry, the Baldwin Locomotive Works being the best known. Baldwin actually built the first Diesel locomotive. They made a few half-hearted tries at Diesels over several decades, then gave up and went back to steam. Baldwin ceased all locomotive production in 1956. The corporation lingered on for a while, not doing much.
[1] https://www.tvcameramuseum.org/rca/tk40a/pix/rcatk40aart.pdf
Suppressed inventions are very rare, but that's one of them.
One of the few others is Airadar, from 1973.[1] Not "AI", "air radar". This was a phased-array radar for small aircraft. Way ahead of its time. So far ahead that the USAF was upset about it being available commercial before it was in use by the military.
[1] https://books.google.com/books?id=NWzlTqj0gQ4C&pg=PA64&lr=&r...
According to official data from the U.S. Patent and Trademark Office (USPTO), there were 6,653 active secrecy orders in effect at the end of fiscal year 2026. Roughly 100 to 350 new secrecy orders are added every year. Many remain locked away for decades or more.
If the U.S. government puts a secrecy order on a patent application against the inventor's will, they are legally forbidden from talking about it. The inventor cannot talk about how it works, publish research on it, sell it, or market it. They cannot even discuss the exact classified details with a regular lawyer unless that attorney has the proper security clearance.
The rest is history.
That failed and SpaceX was then forced to develop its own brand new engines.
Where can I read more about this? I have seen people make this claim for a while, and when I looked deep, it used to turn out to be wrong.
Lex fridman podcast 2022
ChatGPT answers this just fine.
He also said they will be “ Writing aids, not much more" and they “Can't plan.”
No one said he knows nothing about building AI. He was just clearly wrong about LLM capabilities.
As far as I know, those things are still true. Unless there is text in the training material that the text in the prompt can map to, or there is text in training material that describe how to map text in the prompt to text in the training material, LLMs cannot process the prompt.
So may be he is wrong about what could be in the training material. But I don't think he is wrong in the fundamental..
This is sort of what I mention in my original comment. These claims, when pushed, turns out not to hold water...
AGI will require architecture changes not simply more data.
At no point do I ever confuse a human being with a Claude session.
Even if I'm somehow fooled or confused for a minute, the context window / hallucinations / Claudeish will immediately give it away.
Sure, we passed the turing test for chatbots but we are very far away from AGI.
We did?
> This is also taking his comments about transformers scaling out of context
> AGI will require architecture changes not simply more data.
[0] He speaks about Bitter Lesson directly at 9:57 but the earlier part gives better context: https://www.youtube.com/watch?v=21EYKqUsPfg
Yann LeCun has the syndrome where someone who sees the value of B (non-text modalities, i.e. physical reality) cannot stop themselves from constantly critiquing A (LLMs), even though clearly A has its own advantages.
I think he is a smart guy. But he constantly deploys the crackpot's PR playbook.
More specifically:
* he said that autoregressive transformer is fundamentally flawed: if it makes one error, all further generated tokens make no sense; that's, obviously, bullshit: there are many ways to implement backtracking, and there are many papers demonstrating backtracking at a time he did that lecture. And, of course, it just doesn't matter - reasoning model can just write "Oh, that's wrong. Let's try a different thing". - it's just not derailed by one mistake.
* he said that GPT-1000 can't possible have understanding of physical world, when GPT-4 already existing at that time could solve a problem he described.
So it's not like he was bad at predicting things, he's intellectually dishonest.
The idea that someone should be fired just because they were wrong seems to be something that would cause issues if one wanted to think long term.
This approach will make people afraid of being wrong. It will lead them to avoid challenging dogmatic assumptions in large organizations and inevitably lead to group think.
People really shouldn’t be punished for saying wrong things. Long term it’s better for a company if they aren’t punished for that. Helps to let people freely criticize dogmas or propose new ones.
You're being uncharitable. He would have been right if Chain of Thought hadn't been pulled out of a hat, and no one saw it coming.
At this moment no one can know if there will be a future step-change on LLMs similar to reasoning, and conversely, no one can be absolutely sure there won't be - so it shouldn't be a firable offense because it's not obvious; even though it looks like it in hindsight.
Even a nobody like me was playing with a sort of CoT with llama finetunes before o1 came out.
Only this year I'd say we picked all the low hanging fruit that has been obvious for a while that was needed.
For verifiable tasks like coding and math, you can keep going, doing more and more RL. You have the AI generate hard problems, then try to solve them and reward correct solutions. We basically have RSI for verifiable tasks, we just need to keep the wheel turning.
Meta AI is actually frontier lab now with LLMs. Muse Wonderfull is usable with multi agent orchestration (where other labs struggle).
I do not think that would happen with pure research and without single focus!
Already happening: https://news.ycombinator.com/item?id=50397994
Obviously we'll see how it pans out, but other approaches feel a bit like trying to ignore the bitter lesson.
Is that observed scaling part of some equation or law? Is there a limit to how far that scaling goes before a ceiling or diminishing returns hit? Or is it simply based on more GPUs/parameters/compute plus how much human knowledge is in existence and will scale predictably forever?
And you can in practice often see a large network follow an expected training trajectory for loss/entropy.
In practice data is getting more sparse than compute, so you are usually overtraining using more tokens and being suboptimal.
These empirical observations on quantities which aren't "intelligence" since no one has yet defined intelligence.
I wonder about this a lot.
There are all these publications from OpenAI and Anthropic that are solving complex math & science problems.
Are they just "brute forcing" it? Able to try a lot more "known" or reasonable approaches given existing knowledge?
Or are they showing "emergent intelligence" and able to try genuinely new approaches that can't be expected from the knowledge in their dataset?
There's a mathematical sense in which it cannot be: the data-processing inequality (https://en.wikipedia.org/wiki/Data_processing_inequality). The extent to which that sense is totally counterintuitive to how things normally work in ML nowadays has led to investigating different foundations for information theory: https://arxiv.org/abs/2601.03220.
Are you smart if you see that cognitive scientists had something figured out, connected it to something in your field of chemistry and made a breakthrough treatment for some disease?
Does any of our existing intelligence build on completely unknown information? You could argue that any new solution/emergent intelligence is an association of existing information in new ways.
It also brings us to what is brute forcing? Humans are brute forcing stuff most of the time. You will have 10 stupid thoughts for every good thought. 10 stupid ideas for every good idea. It's very rare, if not near impossible, that a non-mathematician with no prior knowledge, goes into a room on day 1 of deciding to think about a conjecture they don't know anything about, and solve it on the blackboard. They need to use their collect associations, formalize their axioms, think of first principles (known information), and associate it.
So to answer your question you need to answer:
1) What does "smarter" refer to? Is it accumulation of information about the universe? Is it solving tasks useful for the human in the world? Something else? Is it acting as an agent with it's own goals and interests in the universe? What if the goals are hard for humans to understand?
2) What does "genuinely new approaches" refer to? Is it fine to use math axioms to prove things? Does interpolation between two existing solutions count as new? At what granularity of concept does novelty happen (Does using two higher abstraction concepts, like combining some new, modern neural network architecture with a database count as a new thing), even if the underlying lower abstraction of information as storage, latency and compute is not novel?
(This moves the argument to "is being really good at predicting the next token the same as being intelligent?" which has happily diverted many minds for years now)
> for every doubling of model size the number of training tokens should also be doubled
Well, I don't think many claim that increasing the model size while also increasing the training tokens doesn't make sense.
The problem is how to get the training tokens...
And the answer to your question is basically just “synthetic data”.
Where's the money coming for those 10X and 100X ventures?
Trying to estimate equivalent "parameter size" of the brain when we only superficially understand behavior of individual neurons let alone how intelligence emerges from the connectome seems too speculative to be a particularly useful metric.
LLMs need an awful lot of handholding all the time, meanwhile humans get pretty good at it on their own.
Is this true?
I barely notice any improvements ever since... 4.6? At least for software development.
Opus makes the same sort of mistakes and has the same sort of strengths ever since that point in my perception. My impressions is that increasing the number of parameters clearly has diminishing returns.
In fact, the best improvement I got from Claude was Haiku 5.5. Good cheap workhorse to implement plans.
As in your statement to me indicates that you clearly see the scaling working over longer timespans. You chose a relatively recent 2026 model as your baseline after all, not a 2025 model or earlier. You clearly see that the 2026 model is better than ones in the past.
Might i suggest that you perhaps just don't see the improvement as clearly when comparing iterations that are fairly close together? If you're not anecdotally seeing improvement between an early 2026 model and one from ~6 months later perhaps just give it another iteration? Personally i do see 5.5 as much better but i can understand not seeing a lot of difference to 4.6 since they literally aren't that far apart in releases.
No, you can't. Because you invalidate your argument by adding later:
> Personally i do see 5.5 as much better
It has to be one or the other.
Either each minor release Anthropic farts is a major improvement from what came before, or I don't see much difference because it has been very little time in between releases. Holding both arguments as true would require me to be excessively apologetic for a company that desires a 2T IPO.
And to be clear, I do think that there is an argument that not much time has passed since Opus 4.6 and Chat GPT 5.3 Codex days.
Part of this conundrum may be because the models reached a good enough point for software development. Maybe walking 80% of the way was easier than the final 20%. I am nowadays more excited about how cheap models can be useful for it (hence my positive remark about Haiku 5.5 and Chinese models in general).
Chinchilla Law says no, we wont. We're likely well past the point of diminishing returns but MOE hides a fair bit of it.
We already know that power laws apply to scaling so 10x the size is going to result in minor gains - where costs become a real factor (as if they weren't already).
FTA "The core philosophy there is scaling, and aggressively not research."
This is where the real gains for AI come from - good enough and CHEAP is going to win the day here (brute force, mixture of models). It's the engineers displacing the academics - and that will mean a bloodbath for revenue (see DeepInfer for what the end game looks like).
Bioinspired algorithms may eventually lead to vastly improved intelligence per a weight too.
I don't think we should do this, but I feel doomed in that I suspect that we "can".
I'm incredibly skeptical of anything that goes "oh, we need to split AI into specialist systems".
I suspect that a lot of the delta between "our best models" and "base models they're trained on" lies in relatively low amounts of parameter changes driven by SFT and RL.
Model capacity is not the limit - the limit is our ability to drive those changes efficiently. Even things like "good research taste" could be represented as tiny in relation to total model capacity, if they could be trained into a model efficiently.
I forget who said it, but a trenchant insight is that current neural networks are generalized function approximators and essentially they are doing through brute force what biology accomplishes elegantly. Our hardware (both in the sense of our brains in terms of elegance and our computer chips in terms of raw throughput) is amazing.
Thinking tokens have nothing to do with parameter count and where some of the biggest gains have come from.
We're really past the point of scaling parameters being viable. We simply lack the content to make "bigger" better at this point.
I suspect that the "elegant" workings of the human brain are a pile of hacks. Evolution crammed the brain full of patterns that are a good fit for the specific problems it typically solves - the way curve detectors and scale invariance are a good fit for CNNs.
We don't have millions of years of evolution, but what we have is enormous datasets and compute at scale.
The one thing that is clear is if it can become more cost effective it’s world changing. Even at its subsidized standards now it’s world changing. And it’s reasonably classified as AI IMO, but it’s not self aware.
I think a lot of your general traditional researchers are fundamentally looking to understand ourselves through machines as proxies in a sense. Something self aware that senses, learns, adapts, so on. The current path isn’t there or necessarily clearly the key to achieving that. That may be the disconnect. I think LLMs may give us some hints at that but not the full picture.
Still incredibly powerful for things we once considered nearly impossible.
I see it that most of legacy research is not obsolete but can be significantly amplified by AI.
Race down the transformer highway as fast as possible and explore other AI methods.
One need not preclude the other.
Going close to as fast as possible while saving a stream for the other approaches makes sense as an approach, but let's not ignore that there's a compromise/tradeoff here. Might not matter in orgs with near infinite resources
For an example of what this looks like at a more pronounced scale just look at the discourse around European models and how far behind Minstral are perceived to be.
Look at what happened to Meta where they squandered a lead, had some mixed results, and reorged to more product certainty and away from fundamental research.
Both of these approaches require quite a bit of institutional support notably.
I should also say that I use ChatGpt daily and enjoy it for researching algorithms. I tried to set up Claude last week and it/he was so overwhelming as to be unusable. So I think frontier models can be "unwieldy" or something for average use.
It
So yeah, LLMs keep getting more powerful/useful but because of an industry wide pivot not a "straight highway" as imagined in the GPT-4 era, when simply going bigger and bigger seemed like a reasonable roadmap for the foreseeable future.
Moore's law: Computing power is multiplied by 10 every 6 months? Can't wait for the optimization and rationalization phase, where we'll recover a trillionth of that power to run powerful databases.
Feels like this is leaving major breakthroughs on the table for the competition. Sure, transformers might get us much farther. But if someone figures out an approach that unlocks a performance boost or better resolution then they are going to kill the competition.
Feels like the American industry is high on all the capitol being shoveled into the furnace without caring how efficient the furnace is. At some point someone is going to make a breakthrough and burst that bubble. This could crater the value as we are currently betting on the low power efficiency of the models as the barrier of entry and point of profit. Of course that would lead to cheaper AI but is that profitable vs local models once (or if) it becomes affordable to do so?
This is a false dichotomy if I ever saw one
Why not hedge and just keep an ai skunkworks team working? These companies are big enough that they can have these people far away from the main work without distraction or without starving those teams of bandwidth
US rightfully placed a bunch of labor/environmental regulations around manufacturing which moved to China but stupidly didn't place restrictions on goods manufactured with severe labor and environmental violations.
We are repeating the mistake again.
The 'US' isn't a monolithic thing. US market leaders convinced US politicians to open US markets to China, and the lack of labor and environmental regulations was a feature, not a bug in the plan.
Yeah, we all know that.
Who voted for the politicians? Who clamored for cheaper goods and fell for the lies?
Anthropic, which is more open about this has shown the numbers, it works.
More and more, the talk of using generative models (e.g. transformers) is coming up to replace existing models in all concepts. Anything from understanding data, image recognition, labeling, ranking, etc.
As someone who is not an MLE and really doesnt understand these models well, it seems like the transformer approach is a "one size fits all" solution that may not be optimal for all problem categories but is good enough that its not worth investing in specialized models to get incremental improvements.
Which many credible, AI researchers believe is already starting up an increasingly steep scaling wall of diminishing returns that eventually ends at a brick wall.
A Copernican/Lindy baseline suggests the growth rate in LLM intelligence will peak in around another 2-4 years and I'm leaning on the low side because while the growth of LLM intelligence per dollar and per month are still growing in absolute terms, the growth rate of both seems to already be falling as power-law constraints in compute and data bottlenecks take hold.
But the whole problem is that these things aren't acting like other technological revolutions like the internet or Google where lots of profit comes from relatively little spend. They scale linearly maybe if you are super smart and good about your design.
Meanwhile, citizens all over the world are fighting data centers tooth and nail because we are hitting a scaling limit in terms of the resources people will allow to be put into this highway project and its equally unwise and untethered from reality to pretend this doesn't matter in a strategic sense.
I think that we're well into diminishing returns already. The models have improved, but to me the far bigger advances have been in feeding them contexts, building them into loops, etc.
I don't even think much more construction is really necessary in data centers. We can do a lot more with the power/HVAC we have, especially once the ASICs start coming online.
It's like we invented the wheel in 2022, the crowds are fighting over capital to build wheel factories, and savvier tech firms are quietly patenting automobile designs while everyone's focused on making slightly rounder wheels.
We definitely know that.
Opus is 5T, not 1T.
> We'll soon start hitting 10's of trillions
Fable is already 10T.
> "Foot down, there's a straight highway ahead don't divert". I have the view this makes sense too.
The roadblock they ignored is politics.
Three days ago the states where most of the new datacenters are being built (Texas, Ohio, Georgia and others) elected governors who campaigned on halting datacenter construction. The highway metaphor implies money in => intelligence out. Turns out there's at least one other ingredient, and we just ran out of it.
And we’ve consumed pretty much all the data, and that part is getting harder as data is now expensive, Balkanized by owners, litigated and charged for. Couple that with model collapse as models eat their own output (not solved for any models, despite more than a decade of the problem), and add that lots of previously human curated datasets (stack overflow style), and it very well may be the models are not getting much better.
The coding/math breakthroughs were caused by training models on the subset of their own outputs which pass a verifier, see reinforcement learning from verified rewards.
https://arxiv.org/pdf/1912.02292
There's plenty more. Read Google scholar
https://youtu.be/lKoJJxjUfdw is a really excellent lecture that changed how I think about it.
The LLMs still hallucinate way too much when there's no computer code involved, but the computer programmers in here won't be aware of that.
Like maybe a couple dozen people on an explore team?
Opus is better (not necessarily "very much") than Sonnet in coding. That's because of a gigantic engineering effort to make it better at coding. That is a regression: we've gone from "sparks of AGI" and systems that were said to be "genearl-purpose" to systems that do one thing well and everything else... well, there's no real measure of how better Opus is from Sonnet in replying to your emails so it's hard to say.
Same goes for OpenAI in maths. Industry wide, there is a regression from general-purpose to special-purpose systems, like in Good, Old-Fashioned AI. LeCun is absolutely right to complain that this is not the path to general intelligence. I don't think his preferred approach, JEPA, is going anywhere either, but at least he's trying. Other luminaries of the field have expressed similar doubts, e.g. Richard Sutton's tweet was in my news feed yesterday saying that brute force is ngmi and algorithmic advances are needed [1]. Schmidhuber I think has said multiple times something along the lines of "there is no intelligence without physical intelligence".
We are very far from that. The recent trend for Vision-Language-Action Transformers (VLA-Transformers) in robotics has conclusively demonstrated that you can't solve the real world by overfitting to a set of discrete tasks, and there is no source of data that we can harness to repeat the feat of LLMs. Algorithmic advances will be necessary, but as the article suggests, these have now be put on hold.
There is a dumbing down, slowing down of research as all resources are spent to try and scale what we have now. What we have now is not enough for intelligence, but that doesn't matter. The industry's goal is not intelligence, it's making more money than anyone has ever made before; in fact more money than all humans put together have ever made before. That's the current race: who will win the money game. Obviously that's not likely to do anything for intelligence.
____________
[1] Brute force computation is not enough for intelligence. Clever new methods are needed. The science of AI is far from finished. IMHO
https://x.com/RichardSSutton/status/2119142310894334262?a=ry...
The transformer based approach is seen as a highway to race down in terms of training and building better models with minimal organizational distractions allowed while every company puts its foot down on this highway. Exploratory research into other AI methods is highly discouraged for now and organizations are encouraging transformer based scale out and iteration as a top priority.
As an example Meta literally fired it's previously head of AI who held views that transformers won't scale. He wanted to veer off the highway and heavily fund exploratory new types of models. They got someone willing to scale the current method out. Essentially the messaging it "Foot down, there's a straight highway ahead don't divert".
I have the view this makes sense too. Current top models are mere (mere in big datacenter terms) trillions of parameters right now. We definitely haven't hit the ceiling and everyone's noticing "Hey this Opus 5.5 with the increased parameters seems much better than the last one". Well that's a trillion parameter model. We'll soon start hitting 10's of trillions and 100's of trillions of parameter models. Unless there's a ceiling which nobody expects they should keep scaling just as they have been. I really do agree that it'd be insane to veer off this highway with exploratory research at this particular point in time. Foot down, get to 10 trillion parameters, see if the scaling holds, if so off to 100 trillion parameters. If the scaling continues to hold we'll be asking the AI to do the exploratory research for us soon enough.