Anthropic just put a Mythos-class model in the public's hands (Claude Fable 5). Here's what your shrinking security head start means for leaders, minus the panic.
Oh my god such a token hungry monster. If you’re a vibe coder, probably isn’t meant for you as much. There won’t be a significant difference vs opus. And with such prohibitive pricing who will pay? Even enterprises and biggest companies ever are pulling back claude code. I guess cybersecurity attacks that are state sponsored or bad actors are probably first in line. At least anthropic proactively called them out, haha!
I *wanted* to be excited for this release but there's no way I'm going to pay for it after the 22nd! I tested it for a bit (and when I say a bit, it was REALLY a bit haha) and it consumed 50% of the 5 hour window.
I'm curious to see how more experienced AI folks use it :) thank you Joel for this breakdown!
Dang that’s crazy! I may try to only have it be pulled for certain high level verifications, not everything. Was it faster or just better in responses overall for you?
Worth watching the timing on that. A free tier that meters later tells you capability isn't the scarce input, the compute behind it is. One argument I keep coming back to: AI's next three to five years are set by transformer purchase orders, not model capability. Which is why a tool that finds 1,596 vulnerabilities from one prompt eventually shows up with a meter attached.
How does this release feel for you? Negative? Positive? Uncertain?
It's exciting on a number of levels, Joel.
How it fits into Anthropic's broader strategy and what people do with it remains to be seen. The economics signal a welcome shift.
Yeah! I’m also interested in how the other major ai labs will step up to the challenge
great article :) realistic outlook and great perspective. thanks a ton for the shout out as well! 🔥
For sure!
Oh my god such a token hungry monster. If you’re a vibe coder, probably isn’t meant for you as much. There won’t be a significant difference vs opus. And with such prohibitive pricing who will pay? Even enterprises and biggest companies ever are pulling back claude code. I guess cybersecurity attacks that are state sponsored or bad actors are probably first in line. At least anthropic proactively called them out, haha!
I was thinking about that! Not sure where I’d need more than opus other than to look for bugs after a creation
Probably we’ll see use cases emerge as people who can experiment with it more!
I *wanted* to be excited for this release but there's no way I'm going to pay for it after the 22nd! I tested it for a bit (and when I say a bit, it was REALLY a bit haha) and it consumed 50% of the 5 hour window.
I'm curious to see how more experienced AI folks use it :) thank you Joel for this breakdown!
Dang that’s crazy! I may try to only have it be pulled for certain high level verifications, not everything. Was it faster or just better in responses overall for you?
Not faster, like at all. And not surprising in terms of responses but then again I didn't test it to its max capacity!
Interesting!!
The same AI that finds the flaws can help us fix them faster if we build the workflow now.
That’s huge as well! Have you tried it yet?
Worth watching the timing on that. A free tier that meters later tells you capability isn't the scarce input, the compute behind it is. One argument I keep coming back to: AI's next three to five years are set by transformer purchase orders, not model capability. Which is why a tool that finds 1,596 vulnerabilities from one prompt eventually shows up with a meter attached.
That will be so interesting to follow! And now… mythos gets pulled off the shelves