So fable will jump more often to Opus than it already did on original release? Working with fable felt like having to constantly fight against your work tool. Frustrating. Now they're making it even more frustrating.
Just a code review of my own project. Downgraded to Opus 50% of the time while evaluating the critical I/O and memory safety parts, the exact thing I wanted it to do.
I mean what do you expect when covering memory safety topics with a model that's not allowed to cover security topics? This seems totally expected. It'll be the same when 5.6 is released.
They're very similar models though, just with different safeguards and restrictions in placae around particular use cases.
I guess the underlying issue is that there is this model that is very capable, but it's being hobbled because of a fear of abuse. It may well be justified, but for a legitimate user any restriction just makes it a worse product and after all the puffery around how good it is (and some practical experience of how good it is) it's a pretty shit experience. "Here's our best model, no you can't really use it".
They're going to be verifying people's identities anyway. Why not put that bit of security theater to good use for once? I'm the author of project X, now let the model work on it, would you kindly?
This "only super special corporations get the model" nonsense is dividing society into haves and have-nots.
Fable is very strong for finding bugs. But you are explicitly not supposed to use it for cybersecurity. Even in the initial rollout I had it refuse and fall back to Opus when implementing a change password function
I can't roll my eyes hard enough at all the people who say this shit about Anthropic every day. I know I'll get downvoted. I know it's lame to complain about future downvotes. I don't care anymore.
Anthropic was correct in their assessment and early warning of Mythos's capabilities, and they did this rollout pretty well. They were not hype marketing. They were being genuinely cautious and honest.
The Trump admin was largely unreasonable with the sudden export control. (Though not entirely unreasonable.) The export control also had not much to do with Anthropic's pre-release warnings. See: GPT-5.6 currently being held up by the federal government.
> Anthropic was correct in their assessment and early warning of Mythos's capabilities, and they did this rollout pretty well. They were not hype marketing. They were being genuinely cautious and honest.
So what prevented them from putting in the sort of safeguards they ended up putting in without hyping it for months prior as being so good, it's too dangerous?
I'm not sure what you're saying. They spent ages adding guardrails to Mythos. Then they spent ages creating a whole new even more guardrailed version of Mythos called Fable. Then they added tons of classifiers so API requests to Fable would get rejected even if you ask a question like "what is a molecule". They put the thickest layer of bubble wrap around the model of any model in history. And then just today they made the classifiers even much more extreme than at the initial launch.
If they were truly honest in their beliefs of the potential risks of this model, how would their behavior have differed? I would expect exactly the behavior we see, if they were being honest in their belief.
Also note Dario here saying they shot themselves in the foot commercially with how they handled the rollout of the model - you can tell by his reflexive reaction how ridiculous he considers the accusation: https://youtu.be/v1wZwxY3CMg?t=2103
I am saying they could have not said anything about it being too dangerous etc. and just released Fable as a new model once the safeguards were in place and Mythos to trusted orgs as they did.
Instead they choose to hype for months about having a model that's simply 'too dangerous to release'.
In other words, why hype it beforehand instead of just quietly add the safeguards they ended up with anyways and release then?
https://news.ycombinator.com/item?id=48466313
Just a code review of my own project. Downgraded to Opus 50% of the time while evaluating the critical I/O and memory safety parts, the exact thing I wanted it to do.
And now it's gonna be even worse.