Anyone else find it weird how within 2 months of Fable releasing all the major labs suddenly had Fable-level models? Trying to think of explanations:
1) AI researchers talk and change companies often, so techniques circulate. This feels implausible because training and shipping a new model ought to take longer than 2 months?
2) Distillation - also implausible for the reason above.
3) Benchmark hacking. AI companies have ways they can dial up performance artificially, and will reach for that to maintain the appearance of parity.
Other reasons?
Edit: Most replies are ignoring timing. It's the near-concurrent release of the same jump in capability that I find suspicious; not the fact that labs can catch up eventually.
It's a combination of (1) and something you don't list: I think the frontier labs all have multiple generations of undisclosed models in continuous training. There is no "end point" when it's magically "ready". It's just getting better and better all the time. What they release with a name and a version number is just a marketing / branding exercise.
So what you experience as a "near simultaneous" release is just their decision of when to peel off a release from their current set of in-training models, likely based on how they perceive market and regulatory conditions. They likely see a competitor release and then baseline what they should release based on that and it takes a month or two for them to package it up and push it out the door.
What I can imagine is that for some of the labs, they are being forced to publish models closer and closer to the frontier of what they have in training. Effectively, "falling behind" is your forward pipeline shrinking. Google ran out of forward pipeline. So far Anthropic and OpenAI didn't - but probably, one is shrinking.
I've been hearing this myth since ChatGPT came out - that the labs have superAI that they're just slowly trickling out as competition forces them to.
Everyone in silicon valley has a cousin who's supposedly seen Anthropic's new unreleased model that changes everything forever. I distinctly recall sitting in a work meeting where a coworker was insisting GPT4 was AGI that anthropic was too scared to release. It's amazing to me that this playbook is still working at least somewhat on folks.
This explanation does so much without leaning into conspiracy that the labs are already sitting on the secret sauce but diluting it for the public or being left mystified when a lab drops out of the race for SoTA
There is a widespread belief that the nature of intelligence is scalar, like how a person can have 100x more wealth than another person. If this were true, then we’d probably see breakaway RSI from a single lab.
But I think we’re discovering that intelligence is about universality, not magnitude. This is analogous to how building a universal Turing machine wasn’t merely a matter of building a calculator that could multiply higher numbers. The difference is that with calculators we consciously theorized about what universal computation would require, then we built one as a step change. Despite it having low memory and slow speeds, the first one built was as theoretically universal as any computer we have today, in terms of the surface of computations it can perform.
With intelligence, it’s turned out to be less discontinuous, which I believe has convinced people that intelligence is a never ending exponential rather than an S curve approaching a horizontal asymptote. I suspect the LLMs we have today are the same kind of thing we will have in 5-10 years, but in 5-10 years we’ll consider them to be fully universal. At that point we’ll still have improvements in tokens per second and volume of context window, but not in capability per token.
At a certain point the roughness of the ball reaches a size threshold where the imperfections are smaller than the wavelength of light, and the surface takes on a glassy smoothness. Intelligence has similar milestones, almost like phase changes, I think, where capabilities are reached. Maybe it's like a superposition of many small step functions.
Humans, however, are highly variable, which may produce really varied and interesting results if they work together.
One instance of an LLM is the same as another instance, so while you may get more out of it by stacking more of them, I strongly suspect it falls victim to diminishing returns. 100 instances of the same LLM may converge on the same result as 10.
I think using different AGENTS.md can give the same model different perspectives on the same problem. For example a model with a well-tuned AGENTS.md by an expert mathematician approaching the same problem as the same model with a well-tuned AGENTS.md by an expert biologist can grind on the same problem from different perpectives.
It's worth a shot at least, as a microservices architect I have a bias that we aren't networking these enough, a single main agent session orchestrating multiple subagents is different from multiple main agent sessions with their own subagents coordinating with each other.
Crucially, does it make capabilities infinitely scalable? My comment just said that models may have a hard cap, and maybe doing specific setups like yours can make reaching it easier, but making the 'team' 10x larger after that optimal point may bring few to no improvements.
Although I'm also not sure about just how much better models can really get with this technique. Ultimately you're still getting the same model with the same training data, which are the important parts. Asking it to pretend to be something feels like it would just put a color filter in front of the conclusion the model has already predicted, or maybe alter the path to the conclusion slightly or pick a less likely answer that it still could've provided normally.
I agree it doesn't make the capabilities infinitely scalable, wasn't arguing with that point. It's just an experiment. I'm not talking about "you are an expert mathematician, go", I'm talking about an expert encoding their heuristics into the AGENTS.md base context. Routing the model's attention to very different aspects of the same problem in the early context.
FWIW I mean if I have an AGENTS.md that encodes my software heuristics (use an interface in situations like X, here's how we name variables, etc.) it generates far cleaner code than if I don't.
Edit- mostly pointing out that stacking 10 base models vs. 10 models with sufficiently different base context isn't necessarily the same attention routing. I suppose I was thinking about tasks that don't have a concrete single answer.
> 100 instances of the same LLM may converge on the same result as 10.
Not in the highly verifiable domains. There you can take it from say 80-90% maj@x to 99% pass@n. Math, some parts of programming and cybersec are examples of highly verifiable domains. (e.g. if you're searching for a linux LPE, that's expensive to search but easy/cheap to verify - just have a token in /root and have the model retrieve that token)
Verifiability makes it easier to understand how well the LLM works, but this doesn't counter my hypothesis. If X number of instances get 99.0% on an objective, verifiable metric, is there any guarantee that 10X will get 99.9%? The fact that we are reliant on new model releases to push capability in big ways, and that people running gigantic clusters of LLMs end up beaten by new models implies that the capabilities of a given model have a hard upper limit, and that it may not even take much to reach it.
High temperature makes the LLM pick more out-of-distribution tokens, but the choices its presented with are still the same or same-ish. I'm not convinced that the more random outputs don't end up averaging to roughly the same conclusion after enough passes.
Yes, given enough time I can answer all the questions in an IQ test correctly. We measure human intelligence in a time-limited setting and score relative to the performance of other humans doing the exact same task. Problem is brains can’t be scaled. To scale humans we need organizations, but human organizations also don’t scale well with increasing headcount.
LLMs scale well in almost all dimensions. Context window (working memory) can be a bottleneck but for humans you can’t scale it at all.
I think they dumb down their public models to be only slightly better than the competition. And the real competition is China, so the current state of the Chinese models would define the baseline.
I think one evidence is that the US has more than 5x the compute of China. With that difference in training speed, it should be impossible for Chinese models to close the gap that easily. It's also very unlikely that they sell the same public models to their private customers (military etc). We also know they talk about "unpublished internal models" for things like the last HuggingFace hacking incident. So it's not a bad theory.
> I think one evidence is that the US has more than 5x the compute of China. With that difference in training speed, it should be impossible
How could we really know how much "compute China has" in reality? Is it possible that whatever estimates people has come up with for both China and the US might not be 100% accurate?
I'm not an expert but I think this sort of thing is relatively traceable for two reasons. One, datacenters are difficult to conceal. Two, the supply chains for many of the relevant materials are difficult to conceal. Some of those supply chains still require western components, I believe, so if you know how much of X component was sent to china, you know how much compute they have.
In China and in the US most owners of computing power have to quickly gain from it, as obsolescence hits hard. In China a consensual will emitted by powerful companies may convince the central power to subsidize efforts towards int'l market domination: R&D, including dataset building, learning... Maybe even also low prices obtained by selling at a price inferior to the costs...
Its possible no AI lab has any unique edge, and success is a combination of (a) having access to GPUs (b) having access to large amounts of data (c) know about the handful of techniques to build an LLM, of which nearly all are likely open source and documented in papers.
So the cycle of growth is (a) and (b), get more GPUs and get more data and you have a better model.
Yea, this reads as LLMs are a pretty obvious technology to develop(for the highly intelligent researchers who are there). Also there's probably a lot of actual divergence in model capabilities and skills that concealed by the fairly narrow set of tests we run them against nowadays. Like wasn't Grok 4.20 super targeted at non-coding tasks.
Why is everyone ignoring the pattern that has existed since training models became a thing? At first it sucks. Then it's better than humans. Just by using it you generate training data that makes it better over time.
GPUs might explain the remarkably concurrent timing. Data access doesn't really explain it unless all labs simultaneously got access to some treasure trove of data.
To the best of our recorded knowledge, nobody ran a 4-minute mile in the five millennia prior to Roger Bannister in May 1954[0], but more than 2,000 people have met or exceeded this achievement since. In fact, his record stood only briefly, being bested the following month by John Landy.
The moral of the story? People work in parallel on the same goals, they build on best practice, or sometimes just need to see something is possible (reusable rockets). Having achievements cluster like this is normal and expected.
The simplest explanation is that 'Fable-level' doesn't mean anything; it's just hype, and there's not much difference in capability.
All you need to have Fable-level AI is to announce it, and have enough fans shift from insisting that model Y is the best now, way better than model X.
> The simplest explanation is that 'Fable-level' doesn't mean anything; it's just hype, and there's not much difference in capability.
Couldn't be further from the truth. The models can be tested and statistically evaluated.
I ran a massive Fable max code review on my lone lisp codebase. Now that I have switched to OpenAI, I decided to run an equivalent review using Sol max and compare them. I'm keeping all data so I can thoroughly evaluate their performance in multiple areas such as correctness, rigor, performance, security, maintainability, consistency, among others.
Fable pass is 100% done and I'm around 70% done with the Sol pass. Preliminary results are already becoming clear: Sol is capable of reproducing around 70% to 90% of Fable's performance. Haven't tested open weight models but I'd wager they have the same performance as Sol if not lower.
It seems Fable is still king, I'm afraid. It's undeniable that OpenAI is providing huge value here: up to 90% Fable performance at multiple times the usage on a subscription than what Anthropic offers us is a phenomenal deal. However, if one desires the best model, to me it looks like Fable is still it.
If I sounded certain, it was not intentional. I made sure to hedge my statistical claims with "seems" and "looks like". I'm no AI lab, I'm just a random subscription user trying to get the most value out of them.
I'm just saying it's not wise to simply put all these models in the same bucket and say any differences are due to vibes or hype. They are clearly different. We can and should scrutinize the testing methodology but it's not exactly fair to just ignore the results.
I don't intend for my benchmark to be private. The core component of my test is my parallel code review skill which is already on my GitHub. I'll be publishing the results on my website when it's done. Anyone could take the skill and reproduce the test using multiple models against any codebase out there, then analyse the depth of each model's findings.
In theory that's what benchmarks are for. If you're assuming they're "benchmaxxed", note that new benchmarks have been released after the model came out that it did well on without being trained.
Do you have any links to credible claims or independent benchmarks that found they were a step down? Or a specific task that worked worse for you?
My private benchmark tasks, and independent evaluators I've seen all overwhelmingly showed improvement.
Every model released for the past four years has had claims on the internet of getting worse. But transcripts are permanent so it should be easy to give a side by side of an earlier task that is now worse. I don't ever see people do that. Instead I see that every single task on a computer that is verifiable is now night-and-day better.
I'm genuinely curious if you've used them yourself or you're judging this based on internet commentary?
I'm solidly in the "they are benchmaxxing" camp. This became very apparent with GPT 5.6 Sol. It, too, was widely hailed to have near-Fable level intelligence. But I used it non-stop for a week and realized that they had mostly just dialed up the relentlessness meter to eleven, most likely via heavy RLHF.
Last week I gave it a small-sized auth ticket to work on, then stepped away. I came back later that afternoon and found that it had worked for 3+ hours and written 25,000+ lines of code. I skimmed over the code and it looked like a small fix followed by a massive number of additional checks around it, including static analysis tooling.
I gave it to another GPT 5.6 and said "check this code and see if it addresses the ticket". It looked at it and said that 98% of it was garbage and should be thrown away (its own words). I then gave it to Fable, which said it was massively over-engineered. Fable's theory was that the agent implemented the fix first, but then compacted and lost crucial context, forgot what the original task was about, and kept going. After many compaction cycles it was completely lost.
Some people complain that Opus 5 stops before finishing a task. But to me, that behavior is vastly preferable to what GPT 5.6 Sol does.
Yeah I found the timing on Sol especially curious since it came right on the heels of Fable. I've had mixed results with it - sometimes it seems great, other times it makes mistakes so stupid I cannot understand how it ever gets anything right.
Explaining it as a difference of effort would explain both.
Agreed, but my suspicion is tied to the timing. Catching up eventually is to be expected. Having similar jumps in capability ready at the same time is odd.
There's also a bit of selection bias going on here because we forget about labs that don't have a jump and just focus on the ones that do. Notably Google is definitely not having that capability jump.
Maybe "readiness" is quite a flexible category? You're mid-training for your next model; a rival releases something; you clear the boards and release the model without completing the training run?
Touche, aborted training runs probably do happen often. Closed model providers have zero incentive to announce a new model with less-than-best benchmarks.
My theory is that it all boils down to better data and longer post-training period. Cursor got curated data from the trillions reactions of real world developers in real jobs. xAI bought is and used it for its post-training and got Grok 4.5 . Longer post-training on the powerful Colossus cluster helped it get Grok 4.6 , although both versions use the same model with the same number of parameters. Thus, both must use the same pre-trained model as a baseline. See also an article infers the training and release timeline of popular models featured a few days ago here on HN.
Chinese labs must follow similar trajectories plus their specific efficiency improvements. That also explains the jump from DeepSeek 4 performance in April and July releases. They both use the same pre-trained model as well.
Now GLM 5.3! And they explicitly confirms my theory:
"Scaling post-training is all we did for GLM-5.3. With GLM-5.2 we built the stack: IndexShare for efficient long-context processing, SAO for RL on long-horizon tasks, and slime for large-scale asynchronous training — all running on the long-horizon task environments we have been accumulating. Over the past month we kept scaling on this stack: more environments, more diverse tasks, and more compute spent training on them."
The release of Gemini Flash 3.7 just 3 weeks after 3.6 confirms my theory, IMO. Only post-training refinement and reinforcement learning (RL) trajectory optimization could yield such high improvements using the same baseline pre-trained model. Flash, MoE models are basically so efficient that the AI labs can put them in a continues post=training loop.
This is basically the answer, they generate A LOT of synthetic task rollouts in parallel, then use RL on the resulting reward signals to improve the model. Add scale to this and you have a Fable class model.
keep in mind fable = mythos which as been "done" since february. so the gap is not 2 months, it's more like - techniques probably started "working" in late 2025, now are trickling down to 2nd tier labs 9 months later.
Yes it does, it just means all the companies come out with similar models around the same time. If what they were doing was completely novel, it would take a long time to repeat. As it is now each company releases a new model every few months, and every couple years the "leading" company changes.
Who are these task producers? Are you saying that Anthropic, et al delegate the RL part to third party companies that do it for pretty much every other AI company as well?
Yes they're called RL gym companies and there's a whole ecosystem of them. You hardly hear about them because their only customers are AI labs and RLVR is where the improvements are coming from at the frontier right now.
Note that RLVR is incredibly compute expensive but it's CPU as much as GPU.
Why would you release a model if you are the current frontrunner? Only when a competitor pulls ahead, or comes close enough to actually get traffic, you prepare a new release.
I think this is the main one. The benchmarks from this are heavily cherry-picked, and they also widely publicised their performance for 4.5 while downplaying the fact the benchmarks were "accidentally" in their training set
Yeah, I’m not convinced that there are any models as smart as Fable. Opus 5 definitely isn’t for all it has great benchmark scores. Fable displays judgement in a way I haven’t seen from any other model.
My experience with Fable is that it eats all my tokens and returns something I didn't ask for.
I realise this might be a skill issue.
I prefer models that are less "smart" but faster. Do the thing I asked you to do, immediately, and if you can't tell me and we'll work it through. Iterate faster not smarter.
what we're going through is the same thing as smartphones, the limiter is compute.
it used to be snapdragon came out HTC rushed out a janky phone everyone went omg htc is goat, then in the next few weeks and months others would impliment better versions and people would not notice those as much, finally sony would release a polished phone right as the next snapdragon cycle came.
eventually compute gains leveled off and apple won on taste.
nvidia/tpu is the new snapdragon. Anthropic and google both peaked on the first training run on a new tpu cycle.
you should expect amazing things within a few months of each other from everyone with access to chips and willingness to use them on a training run.
We haven't seen willingness from google to do that. So its currently xai,oai,anthropic, and probably soon meta.
That's exactly what Anthropic said was going to happen!
Their big bet is that models are going to keep getting sharply better, not that they're going to quickly reach a plateau of quality that they can then defend.
They will get sharply better in tasks with verifiable domains...
math and coding
Gradually the labs will start engineering verifiable sandboxes for wider domains like videogames
This strategy will hit a plateau in about 18 months and then we're back to diminishing returns and incremental progress along other dimensions (like accelerated inference using ASICs)
RL can do behavior cloning, but really needs good simulations or verifiable environments to get to superhuman levels. That currently exists for math, coding, and a lot of videogames. Soon there will be good enough simulations for robotics.
There's a lot of domains where that simply isn't the case (like bio)
You get much better supervised data in bio/chem though. These data companies have people working on exactly that.
While it's not going to give you an "alphago" effect, it is still enough to work at human levels, augmented with the general knowledge of an LLM, together making it super-human.
Yes. But there is also no other choice for people in these professions. The underlying job has been automated already. What's left is automating the last leg.
If you consider a 5-year outlook, it is also a very temporary job unless you're like a specialist neurosurgeon or something, as one of the examples in that article shows:
> The on-again, off-again nature of the work is not just the result of company culture; it stems from the cadence of AI development itself. People across the industry described the pattern. A model builder, like OpenAI or Anthropic, discovers that its model is weak on chemistry, so it pays a data vendor like Mercor or Scale AI to find chemists to make data. The chemists do tasks until there is a sufficient quantity for a batch to go back to the lab, and the job is paused until the lab sees how the data affects the model. Maybe the lab moves forward, but this time, it’s asking for a slightly different type of data. When the job resumes, the vendor discovers the new instructions make the tasks take longer, which means the cost estimate the vendor gave the lab is now wrong, which means the vendor cuts pay or tries to get workers to move faster. The new batch of data is delivered, and the job is paused once more. Maybe the lab changes its data requirements again, discovers it has enough data, and ends the project or decides to go with another vendor entirely. Maybe now the lab wants only organic chemists and everyone without the relevant background gets taken off the project. Next, it’s biology data that’s in demand, or architectural sketches, or K–12 syllabus design.
I think model level is more a function of the state of hardware. Once it exists and is available (and if a lab can afford it), then they can train their own 1T, 5T, coming up next 10T model.
It's just model size and heavy RL, sometimes they overfit on specific tasks.
RL can get you very far, prior models did not have such a focus on RL for agentic setups.
Look at deepseek, they improved it just by doing a lot of RL and you can see it from how it behaves. You provide very little information about a task, but since they are trained on similar tasks, they come up with a lot of assumptions and details on their own, because they were trained with such an info during RL.
> Anyone else find it weird how within 2 months of Fable releasing all the major labs suddenly had Fable-level models?
Frontier model release cycles generally take around 6-8 months anyway. OpenAI and xAI (or however you spell it, branding almost as bad as X/itter) were probably working on their next generation of models already, and Anthropic just beat them 2 months to this release.
You also say "near-concurrent release of the same jump" - but 2 months isn't "near-concurrent", it's a full quarter of the normal release cycle.
I don't think that the other explanations you gave are implausible, though - for both human circulation and distillation, you can apply those during a training and development run (with reduced effectiveness). Reasonable to imagine those as bumping them up another few points to bring competitors from "a little below Fable" to "around Fable".
I'm pretty sure both Anthropic and OpenAI haven't necessarily been secretive that they have internal models that are much more capable than commercially available ones.
It's probably a mix of all of that plus simply always keeping one in the chamber to 1up everyone else when the time is right.
Anthropic finished a new pre-training run, Opus-sized models got enough of a jump they could have released Fable as Opus 5... but the economics of Opus models weren't where they wanted.
Being the masters of distribution that they are, instead of announcing a massive price hike, they just introduced a new tier and promoted Sonnet-sized models to Opus.
That's why every Opus after 4.6 has had such mixed feedback: smaller model with more RL can only make up so much ground, especially on vibes (which are hard-to-impossible to build a reward for)
(I mention all of this because if they'd just released Opus 5, no one would be asking "why is it a few months later everyone caught up to the latest release"... that's always how it works)
> 1) AI researchers talk and change companies often, so techniques circulate. This feels implausible because training and shipping a new model ought to take longer than 2 months?
The assumed timeline (2 months) is slightly wrong because Fable (Latin) is essentially the same as Mythos (Greek) albeit with protections against cyber and biological misuse.
Mythos (Preview) was publicly announced in April 2026 [1] which means other labs have had 4 months to catch up, not 2 months.
Assuming everyone had access to Mythos from the start, your expression, similar to other folks would have been "Mythos-level intelligence" and not "Fable-level intelligence".
Fair point. Still a very quick turnaround considering the other labs would have to figure out both HOW to train a Mythos-level model and then do the work (and Grok is the last to catch up), but certainly more plausible than a 2 month window.
I suspect that because each RLVR episode injects ~1 bit into the models capabilities, and training on a reasoning trace injects ~megabyte into a models capabilities, distillation is powerful enough right now that they’re all basically the same model
I think a lot of it is just time. The quality of a model is E * C
Where:
E = Efficiency, and efficiency gains come from quality of data, quality of algorithms.
C = Compute (Size of model, flops of train run)
So a better company can train a bigger and better model with less required compute which let's anthropic get there first. If another company does the same thing with a worse: model architecture, kernel, optimizer, etc... They will get there as well if they just run there train run with more flops for longer
Mythos was actually ready about 6 months ago. So if you have 6 months later or hardware setup and time to train you can get a lot done.
Could it be that there's no magic formula, everybody uses the same known ideas, the same computation power, the same training data? if that's the case, we can imagine that models will be commoditized.
There is a herd of companies all running a race. The technology is known. They all have roughly the same resources. It’s not unexpected that they have similar cycle times for model development and that those models will be of roughly the same quality. Then layer in corporate PR demands and you see all these models landing within weeks, sometimes days, of each other to keep the model developer’s name associated with “frontier” development.
It's about chips with a large enough scale up domain. Larger domain allows for bigger model, which is what's driving this jump. You've got to get the chips, test them, tune kernels, then start a big pre train, mid & post-train, and only then do you actually get the model. So it takes time. Anthropic got there first partly because they use different hardware (TPU I think, maybe Trainium) which had larger scale ups earlier.
It was said at the time that xAI acquiring Cursor was very smart because it would give them access to years of agent coding traces from millions of users.
$60B in SpaceX stock for Cursor was a bargain
Data + compute + being competent and smart enough to ship.
fwiw I don't think these are yet Fable level - the difference tends to get discovered in the long tail of tasks - but they're close enough, they're cheap, and the length of the frontier exclusive window is narrowing
Well, Opus 5 and Fable are the only models I don’t constantly swear at and call stupid, which seems like a pretty good moat to me.
My guess is all the commenters (you are the 4th person I’ve seen say this) saying ‘Anthropic has no moat’ haven’t actually used Fable or even Opus 5 yet. Sol is laughable by comparison, and Grok… lol.
Not really, it depends. Sol is better and useful in some areas. Definitely not all.
Fable is gimped just by those "guardrails" that silently downgrades you to Opus 4.8. Not only do you pay extra for Fable but your caching can be easily messed up. It also doesn't just find all the bugs or is bug-free. Sol has spotted lots of Fable issues and vice versa. Fable also costs 2-100x as much.
> I don’t constantly swear at and call stupid
That's not a judge of anything. There are models that may be stupid and you can swear at it, but if they still get the job done for 1/10th the price... maybe that's all you're paying for.
No, it has happened to almost every other "sota" model before. There used to be a meme with a circular arrow going through Anthropic, OpenAI, Google as a hype circle. Now we can drop Google and add a couple of Chinese companies.
It's not an explanation of why it happens, I am just pointing Fable is not an exception, it has happened with almost every other model release by all these companies over the last 2-3 years.
Researchers moving between companies (and other ways that techniques get leaked) is the largest cause of this IMO. It's happening continuously, so I don't see why the timing makes it implausible. A really underrated strength of Silicon Valley is California's ban on non-competes that allows this to happen and ensures robust competition between model providers both for talent (increasing salaries for workers) and in the marketplace (reducing prices for consumers). If OpenAI had been located in New York instead then Anthropic could never have succeeded, for example.
But I think the other reason you didn't mention is the timing of new compute coming online. Compute is the major factor limiting the training of these models and new datacenter investments are bearing fruit at around the same time.
> It's the near-concurrent release of the same jump in capability that I find suspicious; not the fact that labs can catch up eventually.
When everyone's improvement (or at least, everyone's rate of increase in parameter count) is so rapid, "within 2 months" shouldn't be seen as "near-concurrent".
Okay so everyone is blaming diffusion or spying or whatever but we all use all of the models on our various projects in aggregate and they get to all read the code each other is generating. I do this with research tasks and local random stuff too.
So why do people have this idea in their heads that it's all some sorta secret sauce they are taking from each other?
I didn't mean that - I meant that when, for example, Anthropic started, then later finished their Mythos/Fable pre-training run that people at OpenAI and elsewhere would have heard about it, probably knew some details such as the size of the model etc - people from these companies go out and socialize with each other, attend parties, share houses ...
So, it's not coincidence when they respond to each others models with something roughly equivalent - because they know what each other are working on.
I think it also shows that breakthroughs are not driven by innovative and research but mostly by scaling.
If this is the case, makes sense that frontier labs with similar access to compute driven by funding on same order of scale can produce improvement largely on similar pace
> 2) Distillation - also implausible for the reason above.
DeepSeek V4 Flash 0731 is a distilled version of Fable into the original V4 Flash (announced before Fable), to the point that it also says load bearing and what not.
4) Algorithmic improvements are either relatively easy to find if you already know the system can do better, or they don’t provide an edge that can’t be overcome by increasing training compute.
One company making a big release both reduces the risks of training a big model (you know it can work) and increases the risks of not doing so (you are bleeding market share).
Maybe research is sufficiently public and simple to reproduce or the next steps of how to improve things are sufficiently obvious to the smart people working on frontier AI.
I predicted this exact event several months before Fable. ,not in a provable way, but the reasoning was related to a paper I read from here that I basically self-internalized as variability knowledge. Two very similar papers, one unfortunately named.
I also stated recently (in informal conversation), based on the performance posted, that said variability was only applied to specific fields of information.
So allow me to make a more provable prediction:
There will be another significant jump related to full field converage, followed by another and from there (we'll call this v3), it will then be capable of automating ASI.
not suspicious at all. They are all doing the same scaling of test time, training data so getting similar results.
anyone with access to capital can produce frotier model. hell you can just ask chatgpt how to create a fontier model. recipe is not a secret despite what these 'labs' pretend
Google is providing more TPUs to SpaceX and Anthropic than to its very own DeepMind. Most of that capital investment is going to Cloud, not frontier model development.
Maybe compute is the real moat (chinese possibly skip around it with distillation), xai is buildouts have been insanely fast (colossus 1 - 100,000 H100 GPUs brought online in 122 days lol) so maybe that explains them catching up
asked grok to give a compute estimate for each:
- SpaceX / xAI: ~1.4 GW (owned Colossus clusters)
- OpenAI: ~2–3 GW (mostly rented/cloud)
- Anthropic: ~1.5–2.5 GW (multi-cloud + xAI lease)
1) AI researchers talk and change companies often, so techniques circulate. This feels implausible because training and shipping a new model ought to take longer than 2 months?
2) Distillation - also implausible for the reason above.
3) Benchmark hacking. AI companies have ways they can dial up performance artificially, and will reach for that to maintain the appearance of parity.
Other reasons?
Edit: Most replies are ignoring timing. It's the near-concurrent release of the same jump in capability that I find suspicious; not the fact that labs can catch up eventually.