What TF is Claude Mythos?

Randy is out, so we run the show as a two-person check-in on the latest wave of AI hype and fatigue. We break down the Claude Mythos rumors, why it is reportedly being gated, and what any of this means for real-world users.

• Randy and Jenna being unavailable while we cover a busy week
• Feeling jaded after years of AGI warnings and overhyped releases
• What Claude Mythos is and where it sits after Opus and Sonnet 4.6
• Leaked benchmarks and the Humans Last Exam jump with tools
• Why Anthropic may limit access due to zero-day vulnerability risk
• Whether gating meaningfully stops bad actors or mainly builds hype
• What improvements actually matter for day-to-day Claude users

Josh

Jenna

Jack 

Randy 

Transcript

Just Us Today

SPEAKER_01 0:00

Okay. Jack Carden. It's you and me. Me and you. It's just us. It's just us. Randy's out. And I guess it is. I guess it is. It's about time. At least buy me dinner first. Yeah, I know. All right. All right. Well, it's just us. Randy can't make it. We don't know where he is. I think that he's somewhere in the back cave. He comes up for air every now and then. And you know, he'll grace us with his presence, but not today. And Jenna, I'm pretty sure, is wrapped up with work, but we've got a

AI Alarm Fatigue And AGI Talk

SPEAKER_01 0:37

lot to cover. A lot has happened this week. I feel like we're always here prognosticating and being afraid of AI because the next model is gonna approach CLI. Um, no, is it CLI? AGI? AGI.

SPEAKER_00 0:54

Yeah, CLI is the command line interface, man. I knew that. Yes, exactly.

SPEAKER_01 1:00

I feel like we're always talking about the next thing is gonna be smarter than humans, is gonna outdo humans, and we're gonna be left behind. And that was like two years ago. And I feel like we've gotten to the point where it's like, okay, who cares if the thing is conscious or not? But here we are again. We're facing Claude Mythos. I'm like over it. I'm jaded. I don't know about you, Jack. Like I feel like, you know, you're we're gonna raise the alarm and then uh a few weeks later they're gonna put it out for everyone to use, and then we're gonna go, okay, well, that was overhyped. I don't know. What do you think?

SPEAKER_00 1:34

I almost feel like if a company like Google, if this AI model can find a bunch of zero-day vulnerabilities in it, then you should just let hackers have it already at this point. Like release mythos to

What Claude Mythos Claims

SPEAKER_00 1:47

the public. For those who don't know, Mythos is Claude's newest or Anthropic's newest model that they're going to try to release after Opus and Sonnet 4.6. They've leaked some benchmarks that, of course, groundbreakingly shatter the previous marks. Like right here, I'm I'm looking at it. Whereas Claude Opus 4.6 scores 53.1% on the HLE or the humans last exam. Mythos with Tools scores 64.7%. So it's a 10% increase in things like computer use, general knowledge, general skills. So it's slated to just wipe Opus 4.6 off the map just like it wiped 4.5 or what

Gated Release And Zero Day Fears

SPEAKER_00 2:33

have you. And something that's interesting about this is that Dario Amode, Boris Cherney are only letting certain very prominent Silicon Valley tech companies have access to this model first because they feel as though it has the potential to allow hackers to find zero-day vulnerabilities in the tech stacks of Apple and Cloudflare and Microsoft, all of the above. So they're gating the release of this model so that all of these big tech companies can essentially get their shit together and patch all of these vulnerabilities before any bad actors have access. And between you and me, I think that if any of these hackers somehow gain access to Mythos anyway on its release, I don't think it's really

Hype Versus Real User Value

SPEAKER_00 3:22

gonna help you that much. So truthfully, I don't see the whole point in gaining this other than to build hype as the next model that will kill all of the previous models and completely obsolete them. And we've seen this happen time and time again. A part of me thinks it's overhyped. I think, again, it will improve natural accuracy, it will get better at doing everything that we've already seen the trend line improve on. So I don't know. I'm I'm cautiously optimistic. I, for one, as a Claude user, am looking forward to having any improvements to my current tool, to my current stack. I'm probably going to use Mythos because my company pays for all my tokens and a max plan, so I can afford to kind of be a little bit frivolous with my token usage.