The only reason I can see is because Amazon wanted something like this to happen. But I'm not sure what Amazon would gain from that, since they don't have their own competing frontier models.
My guess is that they liked the status quo with Project Glasswing and didn't want Fable to be public, especially if anyone is jailbreaking it into Mythos and using it for cyber
But then it backfired spectacularly and now it seems they can't use Mythos currently
There's a much simpler explanation: Amazon's business is selling cloud services. Amazon is constantly under threat of attack and anything that disturbs the balance between attackers and defenders is bad for Amazon. Amazon also needs to keep their AWS customers safe.
This is Amazon prioritizing their 100% stake in AWS over their 20% stake in Anthropic. It's also possible that Amazon knows things that are not public.
The fact that Amazon is willing to report this despite owning shares in Anthropic and being close to a liquidation event points to whatever they found being actually serious.
"Amazon's CEO knew what he was doing" is not a fact. That's speculation.
When it comes to highly technical, fast moving developments like frontier AI and blue team / red team perspectives, I could see any CEO getting out over their skis. Now mix in some incompetent Trump admin officials, including apparently Howard Lutnick. I am guessing many of these people don't understand the subject matter very well at all.
They would have no internally/externally defensible justification to stop the launch as they are partners/part-owners of Anthropic. They would have to let the rank-and-file keep moving on the Fable launch.
IIRC Anthropic models haven't all been available on day 1, so it does feel like a deliberate choice, especially since they are partners/part-owners like you say. No one would have bat an eye about some corporate PR thing "blah blah mythos is too powerful and too intelligent and so we've decided to focus our capacity on Opus 4.6/4.7/4.8 for now until we have the proper safeguards in place blah blah"
AWS has over 200 services, so that's a little over $2 per service. Yeah, a lot of it is built on OSS, but there is a ton of it, and there is also a lot of work involved in building the APIs and web UI, and making it scalable , secure, and resilient.
Now, you might be able to make a version of some small subset of aws services that runs works ok for a small scale for with relatively simple needs, for that many tokens, but I don't think that's what they were going for.
I’m on your side in that I would never take a contract to actually do this, but…
If we swapped out the IAM backend for something extremely simple like just private keys (one per allowed service or JWT-style list all services in the key), then we could have something that looks/feels pretty similar. With a 2$ token spend.
Not at all the same but it would look/feel pretty close.
For one thing, the financial barrier may not necessarily stay that high forever.
For another, such regulations could prevent a competitor from making the weights open for their model to try and disrupt the competition.
And finally, Amodei would no doubt want to be involved in designing the tests the AI needs to pass, and could (and likely would) design it in a way that Anthropic models would be able to pass easier than competing models.
> You are not a person or entity that is: (a) located in, organized under the
laws of, or ordinarily resident in any country or territory that is the target
of comprehensive U.S. sanctions
Seems to be pretty clear that it would include non-government entities in sanctioned countries.
IANAL, but it seems like the argument from Wickard v Filburn would apply to LE. They may not be taking money but they do impact the commerce of the market for certificates.
I disagree with that ruling, and I have some serious problems with sanctions against entire countries/regions, but it definitely makes sense that LE would interpret it as being impacted by OFAC.
It isn't just the US. China, Russia, the EU, and Australia and probably others are all increasingly trying to create virtual walls of various forms in the internet.
It is in the nature of nation states to assert control over national borders. That the Internet and the globalised flow of information it enables circumvents this is a historical anomaly.
If your primary audience is people who want to contribute to it, then the fun of hacking on it is more impactful than the benefit of the offering itself. That's not bad at all! But it also does not suggest that the offering itself is particularly impactful.
I suspect the fact that Github and Gitlab use ruby backends is a significant (although by no means the only) factor in the slowness of Github and Gitlab. So yes, being written in a language that is better suited for high performance at scale is, potentially, an advantage. Although if it is vibe coded, there's a decent chance there are architectural problems that offset the advantage gained from the choice in language.
Not literally. And I would hardly say it was a matter of language superiority. I love Ruby myself. But Github was a lot simpler when it was still just a Rails app.
But Rails was SSR by default, and most of the frontend was just Embedded Ruby (ERB) template files all over the place. And way back when, it was even relatively common to use Javascript supersets like CoffeeScript[1] and Opal[2]. The latter being Ruby that compiled to JS.
Yes, standard rails. Rails does not ship with components out of the box and the most prominent component system for rails was built and is maintained by GitHub. They have been moving to React for the last few years though, it seems to be more about hiring and ergonomics rather than capability.
Reports of Ruby slowness are often exaggerated. Around the time GH was built that was certainly true, but in the ways that matter Ruby is more than fast enough nowadays, even at scale. (Shopify is a rails monolith, as is intercom and a bunch of other massive services) Compute is/was pretty cheap and the economics of ROR mean that it's ultimately not that expensive to run. I think GH's slowness is a mixture of mismanagement and issues on the frontend.
But for sure a systems language is going to be far faster on paper and Rails is far from perfect and does have some performance foot guns you need to avoid. And yeah, architecture is everything.
the point is that argument can be made regardless of deed restrictions. the city generally does things it believes to be in the best interest of the city, so making a deed restriction with it is borderline meaningless with respect to authority. it is purely symbolic, but you would hope that people representing the city see the deal for what it is and respect it unless it's absolutely necessary to override it.. rather than treating the original owner like a sucker and putting up a data center
reply