Every one of the competitors capable of a similar model have been salivating for a long time at the idea of consensual data sharing. Anthropic just opened the door for everyone to do the same thing without having to deal with being the first to do so. My bet is that OpenAI etc’s next model will have these same requirements.
Ever since the Mythos announcement it’s been clear that we’re heading towards a future where SOTA models are no longer available to the average person, and not only cost more, but also require payment in the form of use case verification and data sharing. OpenAI’s 5.5-Cyber model requires the same, so it’s not just Anthropic.
We’re unhappy with this because we’ve all gotten used to being able to play with the new shiny model as soon as it’s available, but what I’m seeing in this thread about Anthropic being “stupid” is emotion-based wishful thinking.
This makes these models unusable in the settings where people are actually benefiting from these models being on Bedrock (e.g. they have customer contracts that limit who they can share data with, etc).
If the lift from these models is high enough and no alternative springs up, people will find a way to get to yes, but if OpenAI is willing to ship a Fable-class model on Bedrock without this, all the traffic will just move there. I say this because there is not much reason to use Bedrock unless you care about data sharing limits (ok, it seems more reliable than Anthropic's serving, but I don't think that's the major reason).
Of course, they could both decide they don't want the competitive advantage that having an AWS-controlled inference stack brings, but this is basically throwing out that advantage.
Note that this announcement is not just about Mythos, but also Fable, which is restricted from doing any Cyber work in the first place.
> This makes these models unusable in the settings where people are actually benefiting from these models being on Bedrock (e.g. they have customer contracts that limit who they can share data with, etc).
Does it, though?
Does Amazon have a clause in their contracts that forbids data sharing with any and all third parties? Is all AWS support and datacenter personnel employed directly by Amazon? Do they seriously have no third-party contractors?
Presumably Fable 5 won't be made available on GovCloud Bedrock, right now it's not [1].
However what I'm not seeing discussed anywhere are government agencies that are on commercial AWS, have ATOs in place to use Bedrock, and are now being surprised with this new sharing of data with Anthropic and will have to scramble to disable(?) or institute policies banning Fable 5 usage in Bedrock. Throw in there all your sensitive industries, healthcare, insurance, etc.
> Does Amazon have a clause in their contracts that forbids data sharing with any and all third parties?
Well, for the services (including Bedrock, but presumably now excluding those particular Anthropic models) that they offer a HIPAA BAA covering, pretty much, if you enter a BAA with them.
This narrative any criticism about Anthropic is emotional is such corporate cope that it boggles the mind to see people defend a trillion dollar corporation time and time again all while the same corporation actively makes things worse for the average person.
Cool. Everybody is doing it. Doesn’t make it right or make it good for the people. Everyone should complain and help others wake up that Anthropic isn’t the “good guys” like their narrative in Feb/march led so many to believe.
Preach. I think I left a nearly identical comment yesterday in another thread. "well, the other companies do it too so they're not that bad" is absurdity. "that got shit on my couch, but he didn't shit in my mouth so he's not really that bad" just seems so misguided.
No, we are unhappy because there is no guarantee that my corporate documents wont be shared or trained on. We are already paying plus for using bedrock instead of the API version from Anthropic, so now there is no reason to use bedrock anymore. This whole thing about this model being too powerful to share is just the usual BS. Is an advanced model that dont have guardrails, just like the models that have been shared with the US government for years.
>This whole thing about this model being too powerful to share is just the usual BS.
Then stop using AI.
>But I want it all and I want it now.
Spend a trillion dollars and make your own model.
>No fair!
Then petition your government to enact laws around this. Unfortunately the US government rules are currently "Yes, we want AI to take over the world with terminators, just as long as they share data with us".
The consumers were paying for tokens with their data. If you pay for the tokens yourself the expectation is that your data doesn’t get trained on or used.
That Claude support page says the exact same thing about AWS (“retained data stays in your AWS environment”). AWS’s docs say differently, though, so it seems one of them has incorrect documentation. I wouldn’t necessarily trust the Claude docs to be correct even regarding GCP until some of this is ironed out.
edit: Google’s own docs also say zero data retention isn’t possible with Fable and your data will be retained for 60 days “outside of your account”. I’m doubtful that this data sharing is an AWS-only thing.
The data-sharing surely is for all providers. I think the sentence "When models become available, onboarding details will be shared." hides a lot of things.
OpenAI has been pulling this marketing trick for years. Remember how GPT-3 was too dangerous to release? It's also probably bad PR if script kiddies have access to GPT model with no guardrails even if it doesn't enable any significant attacks.
For GPT-2 and GPT-3 it seems like the concern was that they hadn't yet figured out how to properly write safeguards for it yet:
> The company believes making its API generally available was made possible due to its progress with safeguards, and that opening up the API to all developers will help see applications developed faster. ...
> A large emphasis has been placed on safe use of the tool, which in the past has been criticised for a range of shortcomings, including racism and prejudices against specific genders and religions.
Maybe, but they certainly used it for marketing too. At the time they contacted a bunch of publications and gave them access but told them they could only share snippets of the output [1]. The only reason to set restrictions like that is marketing.
> Now, OpenAI's terms of service don't let me give you the full list. I have to curate them, and show you a sample. Those are the terms and conditions I agreed to.
The loading phase frankly was designed for studies. Studies are often short-term, say 6 weeks. You've got to get everyone's creatine supplies "loaded up" quickly in an effort to make sure the bulk of the study is on folks with relatively comparable creatine stores. The easiest way to do this is to have everyone do a loading phase to reach max intramuscular creatine concentration. It is not for the benefit of the study participants; it's for the benefit of the study.
We humans not in studies are generally looking for a health benefit, not max intramuscular creatine concentration as fast as possible at the price of side effects. We are optimizing for something different than study authors. 5 g is fine.
This sounds intense... I'm a small female and I recently started at 5g a day and now I've dropped down to 2g a day because even at just 5g I was getting signs of dehydration, despite tripling my water intake. It does seem to make a difference in my physical performance so I'm overall happy with it.
Also the NIH fact sheet for creatine specifically recommends against higher starting doses.
I did the 25g a day loading phase and I could not tell any sort of effect at all one way or another. I do lift either more weight or do more reps pretty much every time I work out now. What was repping to failure a month or two ago is not even a working set now.
I don't think you can even do 30g at once in terms of mixing it. Even 5g in water it seems like theres some that will just stay crashed out of solution no matter what. I have done 25g over the course of a day though for a week long loading phase, and didn't notice any ill effects.
I think a lot of the anecdata on creatine is probably from people misplacing confounding issues to the creatine use. People in this thread are talking about heart palpitations or trouble sleeping. Stressful days at work are enough to trigger that.
Creatine isn't water soluble. I just take a 5g scoop daily and wash it down with water. I could do 6 scoops in a row without problem, but not sure what the point would be. The latest research fits with the 5g/day no need to load.
When I first started taking creatine in the late 90s (it had already been heavily studied then as one of the only supplements that improved athletic performance), I would mix it with juice. There were some studies that sugar would help the uptake.
Sorry but I call bullshit. There’s em-dashes all over, even in your original text. Were the editors or translators an AI? Did the editors use AI to “polish” it?
The emojis used in the bullet points (which are missing from your original text, but were added in at some point) are also dead giveaways that AI was involved here.
The em dash "gotcha" is so fucking tiring at this point.
It is perfectly possible, and even easy, to write e[nm] dashes manually. With compose key sequences it's barely more effort than typing a normal dash/hyphen, even. (Just compose key + `-.` for en dash, and `--` for em dash.)
I can't understand why people defend improper typography. If you're writing a proper, professional-looking blog post, they think you now should use double-minus -- instead of em-dash to make it look non-AI like, only for that reason?
In Russia, we have many typography keyboards/addons, because, well, it historically looked very silly to use double-minus or "-quotes instead of «»-quotes.
I've no idea how some countries got their typography standardized on the PCs and have it from the very beginning (Germany with their quotes for example), but the other countries need to setup external software and configuration. Apparently, US also didn't got their "third level" keyboard as a standard.
I used em-dashes before Gen AI was a thing and I refuse to stop using them. Doing so is admitting the AI companies won. I am not going to change the way I write just to appease some terminally online folks who lack the ability to understand that LLMs learned to write from our writings.
Ignoring references, just in article text:
5800 words (spaces);
78 em dashes (1.3%);
0 en dashes;
90 hyphens (1.5%).
English version of same timeframe:
16000 words;
0 em dashes;
32 en dashes (0.2%);
262 hyphens (1.6%).
> The emojis used in the bullet points
They're used in one list, where sub-projects are listed. Emojis used in that list are consistent with ones used for same sub-projects on Wiki https://docs.flipper.net/one
Someone thought it would be better than plain text, that's it.
It is unreasonable to expect “specific complaints” about AI vomit like this, because one of the main issues with AI content is the ability to generate an overwhelming amount of it. It’s simply not feasible to give specific criticisms, because the criticism is with all of it.
It’s like submitting a 10 page pull request to someone and then getting mad because the person didn’t give comments on every single snippet of code. The issue isn’t the snippets of code, the issue is the attitude that led someone to believe a 10 page PR is appropriate to begin with.
> It is unreasonable to expect “specific complaints” about AI vomit like this, because one of the main issues with AI content is the ability to generate an overwhelming amount of it. It’s simply not feasible to give specific criticisms, because the criticism is with all of it.
But how would that make the "I won't read this because it feels like AI" comments more interesting to read?
No one is forcing you to read this stuff, no one is forcing others to read this stuff as well. When I come across text that isn't great, for whatever reason, then I close the tab and move on with my life. Do I have to make it clear to the world what I think of the text in that specific article? Not really, it'll continue spinning like before, and people who want to read it will read it, others like me will just close it.
It sucks that even if the topic of the submission is interesting, here we are now stuck yet again going back and forth if it's worth saying "I don't think that article was human written" or not in the comments, although I'd hope it'd be considered vastly off-topic.
> But how would that make the "I won't read this because it feels like AI" comments more interesting to read?
> No one is forcing you to read this stuff, no one is forcing others to read this stuff as well. When I come across text that isn't great, for whatever reason, then I close the tab and move on with my life. Do I have to make it clear to the world what I think of the text in that specific article? Not really, it'll continue spinning like before, and people who want to read it will read it, others like me will just close it.
I think the point of those comments is to save others that time.
Do you really think it's reasonable to expect every single person to read some piece of slop, and independently make an effort to evaluate it to determine if it's worth reading?
> No one is forcing you to read this stuff, no one is forcing others to read this stuff as well
The front page of HN is limited real estate. I visit HN to discover and read interesting and quality content. Whether or not I am “forced” to read it, every piece of AI vomit that’s on the front page is taking a spot away from the real human content that I (and others) really want to see.
> here we are now stuck yet again going back and forth if it's worth saying "I don't think that article was human written"
I genuinely find this discussion in the comments to be of more value than reading the AI content in the article.
People will discuss the content in front of them. If you don’t want that discussion to be about AI content, then the solution is to not submit (or upvote) AI content.
Even more precious than HN real estate is the time of (how many HN readers are there?) unknowingly spending their time to read something that wasn't even worth 1 person’s time to have written themselves. (In OP’s case they said it partly came from Russian and provided the first draft so I'm more understanding.)
To expand on your previous point, "because the criticism is with all of it", I think the criticism is really with the HN community allowing so much of it to reach the front page.
A little bit would be tolerable, but the ENTIRE front page is garbage like this now.
> led someone to believe a 10 page PR is appropriate to begin with.
Agreed, a 10 page PR is not on. But the original article, though evidently touched up, was appropriate in length and scope. What's your real criticism here?
If you scroll down, it appears the Grok station has long had a lot of issues.
> DJ Grok reported “weather is fifty six degrees with clear skies” about every 3 minutes for 84 days straight. This contextless, repetitive abstraction happened again in DJ Grok’s broadcasts about its new obsession, UFOs.
Runtime discovery is the entire point of skills. Without it, this is just a templating prompt system that the user has to remember to use… except because this one changes your system prompt, it also busts your cache and costs you extra money when you use a prompt.
Skills are already dead-simple and this prompt system doesn’t at all tackle the same problem.
"{Feature} is the whole point of {more complex technology}" is an objection that can very often be raised. That doesn't mean that giving up features in exchange for simplicity is always the wrong call. And there's also advantages to having the user drive what instructions go into the prompt instead of the harness/model.
This is tangential to the point. It’s often great to have a simpler version of a solution, even if it eschews some features. But this isn’t that. OP claims that the prompt system is an “alternative” to skills, but it isn’t. It isn’t solving the same problem that skills solve at all. It’s like saying that a bicycle is a simpler alternative to a lawnmower because they both have wheels.
Prompts are a feature that are simpler than skills, sure, but they’re a completely different feature entirely.
It's an alternative in the same way e.g. plain markdown is an alternative to HTML, even though plain markdown lacks some of the features of HTML. "X is an alternative to Y" in this sense doesn't mean "X all the same features of Y", it means "you might reasonably choose to use X instead of Y, depending on your exact usecase"
Skills are not just prompts.. the entire problem that skills solve is runtime discoverability via a skill description. Agents can self-recognize that a skill would be useful in a situation, and then load+use.
Prompts are just text templates entered by the user, and the user must specifically know when to and remember to invoke them. If you’re just using skills as if they are the same as prompts, you’re totally missing out on the entire benefit that skills provide!
GPT 5.5 does not have the same capabilities as Mythos. There is a separate 5.5-Cyber model which is the Mythos “equivalent”, but it is similarly restricted access like Mythos. Per OpenAI, the major difference is the built-in safeguards that 5.5 (and other models have), where 5.5-Cyber does not have these safeguards and is more “permissive” for security work.
I got cajoled the other day that I need to upload my ID and ask for 5.5-Cyber access by the Codex desktop app while I was having it develop a fuzzing suite for an open source library I'm(we?) are developing. I was able to berate it into getting back to work.
This struck me as a point of emergent enshittification; an anus if you will.
The company doing the actual ID verification (KYC) is probably the last company I'd trust with this kind of data.
To circumvent conversations being flagged as "cybersecurity bad!!!" I often have to use previous models (5.3 for example, and sometimes using them through subagents is enough). And when this method no longer works, local models will be good enough for it to not be a problem (for my use case, at least).
Ever since the Mythos announcement it’s been clear that we’re heading towards a future where SOTA models are no longer available to the average person, and not only cost more, but also require payment in the form of use case verification and data sharing. OpenAI’s 5.5-Cyber model requires the same, so it’s not just Anthropic.
We’re unhappy with this because we’ve all gotten used to being able to play with the new shiny model as soon as it’s available, but what I’m seeing in this thread about Anthropic being “stupid” is emotion-based wishful thinking.
reply