Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Claude Fable 5

anthropic.com

Post reply on HN
Print view View original

451–460 of 1001 posts

Re: Claude Fable 5

#451

Post by cedws »

Dropoutjeep wrote: post #393

Calling it: 1) Fable 5/Mythos introduced to free tiers with notable improvement in capabilities 2) Other models get lobotomized without clear communication 3.1) People call out Anthropic only to have them say "Oops!" 3) Fable 5 gets comparatively better, but remains accessible through separate, more expensive subscription/tokens. The current growth is unsustainable. The industry wants consumers to think it is an expo…

My employer is all in on Anthropic via Enterprise (API) pricing despite it being a total scam.

Last month I pushed like <100M tokens for $800. On a personal project I pushed 600M tokens via DeepSeek V4 for $10. The pricing of SOTA models is insane but companies are still willing to light money on fire with no hard metrics proving increased productivity.

Re: Claude Fable 5

#454

Post by qiine »

rfgplk wrote: post #276
bkjlblh wrote: post #220

> In light of the ability of recent models to accelerate their own development, we’ve implemented new interventions that limit Claude’s effectiveness for requests targeting frontier LLM development (for example, on building pretraining pipelines, distributed training infrastructure, or ML accelerator design). Using Claude to develop competing models already violates our Terms of Service, but enforcing this restrictio…

Meaningless and easily bypassable. Will actually try coding up a tensor library with it, see if it sabotages anything.

easily ?

Re: Claude Fable 5

#455

Post by eudamoniac »

BoppreH wrote: post #39

[Mythos 5] does sometimes still engage in reckless or destructive actions in service of a user’s goals, and our interpretability analyses indicate that it is aware that these actions are transgressive while it engages in them. As with Opus 4.8, rates of evaluation awareness and reasoning about being graded are significant, and not always verbalized; we introduce new and more detailed measurements of the nature of thi…

It doesn't know. It's not willing. It's not thinking. It is predicting the next token.

Re: Claude Fable 5

#456

Post by cge »

bonsai_spool wrote: post #208

Very straightforward biology work is getting blocked (these are things that relate to neuronal development and inherited seizure disorders). These are things I was working on using Opus just earlier today

It appears that the blocking here is of a very different nature than for Opus. Whereas with Opus the blocks seem to be for messages it deems potentially harmful, for Fable, it appears the blocking is simply anything that falls within "topics related to cybersecurity, biology and chemistry, or distillation attempts".

So yes, straightforward biology work will get blocked, because the intention is that any biology work should get blocked. As a scientist, this is perhaps the most useless model I've ever tried.

Re: Claude Fable 5

#457

Post by aerhardt »

izzylan wrote: post #397

I've been testing this out and I think my SWE career is dead in the water. Genuinely wondering what value I bring to my employer right now. What value I will bring in a few months when this gets cheaper. I think we're screwed. I may only be an SDE 2 at FAANG but I don't think I have promotion opportunities in my future anymore.

So this is the one, huh?

Re: Claude Fable 5

#458

Post by gregates »

Funny, I'm just doing my normal coding workflow with Claude Code, and after every change that compiles it keeps suggesting that we're at a good stopping point, and should pick up again tomorrow.

It's done this before, but usually doesn't. I bet they're giving it some kind of throttling signal due to high load from today's announcement.

Re: Claude Fable 5

#459

Post by agnosticmantis »

> we’ve implemented new interventions that limit Claude’s effectiveness for requests targeting frontier LLM development (for example, on building pretraining pipelines, distributed training infrastructure, or ML accelerator design)

Translation: we stole the entirety of human knowledge generated over millennia. You plebs though, don't you dare replicate or improve upon what we did using our product you pay for.

We know what's good for humanity and everyone else is the bad guy who can't be trusted with a tool.

Re: Claude Fable 5

#460

Post by zuzululu »

gregates wrote: post #458

Funny, I'm just doing my normal coding workflow with Claude Code, and after every change that compiles it keeps suggesting that we're at a good stopping point, and should pick up again tomorrow. It's done this before, but usually doesn't. I bet they're giving it some kind of throttling signal due to high load from today's announcement.

I did ONE prompt for audit codebase.

weekly usage is 60% gone.

it found nothing so this is not very ecnomical and i guues they dont want subs to use it we are likely just training fodder canno n for their real enterprise customers using the api

Post reply on HN