
Loading summary
A
Claude, Opus 4.8 is here, but even bigger. Claude, Mythos. What?
B
It's Mythos. I think it's not Mythos.
A
I've heard everybody call it Mythos.
B
I'm almost 100% sure it's mythos.
A
Claude, Opus 4.8 is here, but even bigger Mythos is coming in mere weeks.
B
Friends, we will dive into what got upgraded with Opus 4.8. We got your thinking modes, we got your warmer behavior. We got your slight increases in benchmarks.
A
But, Gavin, we also got Mythos.
B
Incoming. Yes, yes, we got that too. And especially, it's time for AI Summer Redo.
A
Is it redo or redux?
B
I think it's redo. This is another pronunciation problem. We'll leave that in AI Summer Redux.
A
Amazon picked up three generative AI projects, and nobody is angry about it. Gavin. They're not even mad at Kojima, who is in an AI video with Prada.
B
Oh, and also, guess what? 11 labs updated their AI voices to include Stan Lee, and nobody's mad about that either.
C
With great power comes great responsibility. Excelsior.
A
Excelsior Plus.
B
I just released a small AI experiment that I'm calling the Fishbowl, which you can use today. And I'll tell you how I made it.
A
Is it a screensaver for my itevice?
B
Probably pretty close to that. Kevin. This is AI For Human. Welcome, everybody, to AI for Humans, your twice a week guide to the wonderful world of AI. I'm Gavin Purcell. That's Kevin Pero. And we have a big one today because, Kevin, there is a new foundational model launch. Whoop, whoop, whoop, whoop. We have Opus 4.8, which has dropped. Opus 4.8 is a dot release. Kevin. We all know that the dot releases are not that exciting, but here we are to tell you what has changed. And there is some change. But the bigger announcement that we'll get
A
to in a second is I love you. And I really. I do love the space. I really do. But here we are again. Let me guess, uh, bar where I got bigger. And. And DOT Go up. Line go up. Here's what I don't. I don't actually. I don't even really care anymore. You know what I want? I need AI that just understands my nuances better. I. I need AI that feels more natural to talk to. I want something that's a stronger collaborator. From, like, coding. Yes, but to knowledge work. Gavin. And I just. Just wake me when you have an update that would be interesting to me.
B
Kevin, that's what they're talking about here, Memphis 4.8 has a warmer voice. It's got a better knowledge collaborator. But I do want to say we're going to get to this in a second. I think the kind of the biggest news has been buried in this announcement, which is that Mythos, the dangerous Mythos model, the name of the gods, the thing that's going to destroy all security, will be released to a general audience in a couple weeks. But first, let's get back to 4.8 because 4.8 is a step change model. It is not like a night and day thing. There are better benchmarks. Benchmark boys show up, whoop, whoop with a benchmark. The benchmarks are pretty good, but they are small single digit increases across the board. In fact, most of them go up a bit. So we don't have to dive into that that much. But there is a really important thing to understand is that Opus 4.8 is the same price as 4.7 and it is better and more nuanced. And the other thing, Kevin, that showed up on my screen when I opened Claude today is it allows you to change your thinking behavior, meaning that they now have the high, medium, low thinking that GPT has as well, which I think is probably a welcome design change for people who are burning through tokens and anthropic. I don't know if you saw this. There was a story of like all these CEOs are saying, basically, I don't pay too much money for AI now. So there is this like cost conscious effectiveness right now.
A
Oh, you mean the dealer opened the trench coat and got everybody addicted to the tokens hanging in their Ziploc bag and now they're saying, oh, sorry, this is Stan. Standard drug wars stuff. I played this on the TI85. Stop me if you've heard it. Tokens got more expensive. Yes, we've been new, as they say. Here's what's interesting. Dan Shipper from every they they really hammer these models in advance. They get early access and they put them through custom benchmarks. And I always appreciate their vibe checks. What's confusing about this particular one is that he says in bold on his X post, they could have just called it Opus 5. It's that good. Which is okay, that's a strong opening. But then it, but then by the end of it it says, you know, Codex is a far superior harness to the cloud desktop app. So I'm going to Codex every day. Which I know we're talking about harness versus model, but it does like, yes, that is an important distinction to make. But he says it's an incredibly good writer, that it does good knowledge work and it's emotionally intelligent, which is interesting.
B
Yeah, I think this. So first of all, congratulations to damn Sherpur and every for understanding the hype post idea.
A
Do you know what I mean?
B
Like, you get that headline in and then you get those clicks, baby. Moving forward. I think what's interesting about this, and it's one of those things where, like, I wonder if they would have even called this a model release in the last couple years, where it might have just been a silent update to the model in some way. And maybe there's this sense because everything's starting to roll out faster. Somebody had mentioned it's only been 42 days since the last anthropic release and we are speeding up release schedules. So maybe this is a way of doing that. It also could be a way of kind of getting ahead of wanting the Mythos model to come out, but also have a kind of an in between step. I do think the important thing is that Mythos comes out in a few weeks. And they did say specifically they're still going through the security processes, but a general availability to a Mythos level model will be available to everybody in a few weeks. And Kevin, we know that there's been a lot of rumors around what people are calling GPT 5.6. But also June happens to be a very big release month for AI in the historical past. Like, one of the things, I was at this AI the Lot event this week, and it was interesting to talk to people who come up and thanks for people who come up and talk about the show. There were a few people that came up and said hi, which is really nice. But I think one thing that was interesting is I realized you and I have been doing this show now for so long that we have this kind of like, historical background that a lot of people in the AI space don't have. And one of the things that I'm really fascinated with is like this cycle of like June does weirdly seem like this month where, like, it sets up the summer. So we joked about in the intro, like, you know, there was a story last year was like hot AI Summer. But it does feel like we're kind of opening the door to a larger version of that with Mythos, but also, you know, GPT 5.6 or 6, whatever they're going to call it. And then Gemini 3.5 Pro. So the next few weeks could be pretty spicy. I feel like in the world of AI, yeah.
A
So I mean a little bit about how the sausage is made. This model just came out at the time of recording this. Have you spent any time with 4.8? Have you had a chance to use it in anything?
B
You know, I have it in my system, so it did show up. And I did a couple very small things, but not enough to kind of push it in one direction. And I will say the point models are really hard to kind of understand how they've updated. Later in the show, I'm going to talk about this project I worked on, which is like a nice way of doing something and kind of seeing how it can affect stuff. My goal in general is I'm going to try to make and ship more things at large. And I think the good news is, I was saying this to somebody yesterday too, is like the models now, unless you get stopped, they're all pretty good, right? And when I say get stopped, what I mean is like if you hit a point in your project or something you're working on and suddenly it's like you're spending an hour trying to fix stuff, then you're screwed. I have not had that problem with, you know, 4.7 or 5.5 yet. They are both pretty solid for the things that I'm trying to do. Now. I'm not solving the Erdos problems, but I think this is one of those things that's hard to test because a point model is a point change. Like, this won't be very clear, but I think that the good news is like, we are now entering that next stage of what AI feels like and kind of the next level up.
A
You mentioned you went to AI on the lot and I don't want it to get lost that people came up to you, which was amazing. I hope they weren't super disappointed when you said that. Unfortunately, I wasn't there. But in addition to showing support in person, is there a way that the people listening to this could support the podcast now? Because everything you've said so far feels so impactful to so many people. But what if they're not aware?
B
Like and subscribe, folks, that's what you can do. You can like and subscribe this video. You can also share it with people if you're listening to the audio. If you're on Spotify, make sure you leave us a comment. We will reply. Also, hype the crap out of this video. Hype points on YouTube do make a difference. We've seen that before, so hype it up. We have a Patreon, we have a Buy me A coffee. We have a newsletter that comes out every week that I really enjoy writing. So please go and do that. And yes, Kevin, I was at AI in the lot yesterday. I'm going back today. And this is a very cool thing where basically it was a small thing that started about three or four years ago by a guy named Todd Terrazas, who's a really nice guy who started putting together the people from Hollywood and AI and it is really big this year. It is on the Amazon lot in Culver City. And it was very cool to see kind of the growth of this business and kind of see how it is specialized in what's going on. But there's some big news that broke out of this and it is not great for. It's great for AI creators. I think I want to establish that. But there's a lot of this kind of like AI hate that we've been discussing first and foremost. So the biggest news that came out of this was that Amazon, Amazon, mgm, who has had this kind of like generative AI kind of work going on in the background, they have green lit three series to go to Amazon prime that are generative AI series. And all three of these are interesting, but one of them specifically is from a very famous animation creator and director. His name is, I think his name is, is Jorge, but it could be George. I'm sorry, this is a Spanish pronunciation. Horror. Jorge R. Guterres, who actually was one of the creators of the Book of Life feature film. And he has had a project greenlit and he's very excited. He was very excited about this is a big deal in a person's life when they get a project greenlit. In fact, he did have a very famous quote going around where he said, like an AI project, animation project is kind of like having sex and then being handed a baby. So that might be one problem with what. What was going on. But he got so much hate out of this announcement that he actually went on X and under said, I understand a lot of you are unhappy, are happy with me and a lot of you are really angry at me for experimenting with AI at Amazon. And literally he said, do not come after my family. It is not okay. So he is getting death threats. So this is the kind of world that the AI creators who are going mainstream are kind of entering in. I think it's kind of an up and down story. Like, it's great, we feel like, to see this AI kind of get a big announcement and it's a big deal. But also on the other side it's this kind of underbelly of the feeling of this stuff that's been kind of unsettling everybody for a while.
A
Yeah, I mean, there were, you know, tweets surfaced from the. The account is Mexopolis, and there were some tweets surfaced from a while back, basically saying AI was slop or it's. It's soulless or it's theft of artist. You know, understandably common refrains from a lot of people, especially those in animation and gaming, which we'll get to. So I think some felt not only was this like a betrayal of on previous feelings, but like, oh, this goes squarely at an industry that is even in his own words, suffering greatly in the traditional model. And so now you have them on stage championing AI for a company that, by the way, is. Is creating a generative AI fund as well. They're not just green lighting projects, but they've even got a production platform which they announced as well, this project nara, which combines a bunch of AI tools and they're trying to get people to use it to make generative AI. So on the one hand you have excitement over the availability of, as Steven Spielberg says, a new tool. And that's all it is, this new tool.
B
Right.
A
And someone who finally. You. I shouldn't say finally, but in the past was against it, finally uses it. Oh my God.
B
They're.
A
They're celebrating their. Their new excitement and success with it and then the backlash begins. I don't know how to reconcile that for someone. I mean, I feel bad.
B
Yeah, it seems like a really crappy experience for somebody. I will say, to your point, like, the animation industry, like a lot of Hollywood is somebody who spent time in it, is suffering greatly. And like, he. Jorge has specifically pointed out that, like, things don't get greenlit in the way they used to. And like, animation already was not a great economic model. So like, for those people out there listening who are like, how could people be pissed about this? I understand that and I understand the other side, that general public's kind of anger towards AI at large kind of is filtering through this discussion. You had brought up this thing to me about Kojima, which, if you're not familiar, Hideo Kojima, a famous game designer, made Metal Gear Solid and Death Stranding. What is going on with Kojima and Prada right now?
A
I rarely know what's going on with Kojima and that's like, while I actively playing his games.
B
He's awesome, great person.
A
But he apparently is doing some sort of collaboration with Prada? Yes, like the fashion high end brand Prada. And there is a minute 30 long video of, of Hideo running around in like this 1960s 70s esque Mars attacks looking promo for this collab with Prada. And it's, it seems to be pure AI right? It looks like a pure AI video. And again in the world of animation and media and the world of gaming particularly, AI is a big no, no. And so Hideo is getting, getting fairly slammed and some of the fans are pointing out that Hideo is quick to tweet or amplify anything that has his face or name on it or you know, that is made by a fan. I, I don't know his social activity that, that closely but they are pointing out it's odd that nothing regarding this video has been tweeted or announced. Uh, so I, you know, I, it's just.
B
Well, what's interesting is yeah, it feels like the kind of like edges of this conversation. It's almost like anytime anybody does something that's even remot centric is like they're ready to pounce. And I do think there is a level of like the crowd anger thing that is just focusing in on this stuff. But again, we've talked about this in the show last couple weeks is like there is this bubbling underbelly right now of just people being pissed off at AI in general. We've seen it for a while, but it's getting bigger all the time.
A
Once a week I get a tweet or some message, you know, a DM or something from a fan who says that they're incredibly disappointed that I excited for or shilling for or even entertaining AI stuff. And I have to just sit back and go like whoa, I still have a fan.
B
And that's pretty exciting my mind.
A
As someone who has followed my own career or lack thereof for this long, I'm like, that just blows my mind. So I'm just grateful to receive that the actual message doesn't make it in. People have to understand that.
B
That's exactly, that's exactly right. But another big thing, Kevin, that just came out and 11 labs is not going to make this any better. They are now using, if you remember 11 labs a while ago we talked about this idea of their like iconic voices where they have some celebrities that have either passed away or other ones that have just kind of given them their permission. Burt Reynolds, it's Michael Caine, It's a bunch of people. They have now agreed to with the Stan Lee estate to let Stan Lee's voice into 11 labs and Kevin. I don't know if this is going to help the problem, but maybe we should take a listen a little bit. To what the 11 labs Stan Lee sounds like.
C
I used to be embarrassed because I was just a comic book writer. And then I began to realize entertainment is one of the most important things in people's lives. Without it, they might go off the deep end.
B
It sounds fine. I mean, what's interesting about this is if you're not familiar, like, a lot of people who are younger may think of Stan Lee as like, oh, he's the iconic Marvel voice. How could you do this? Stan Lee was an ultimate shill. I want everybody to know this. As somebody who read comics a lot as a kid, there was a period of time where Stan Lee would just do anything. I greatly admire what Stan Lee built and the fact that he co created Spider man, all this stuff. But like, Stan Lee was a businessman and like, so this is not that weird. But again, it's just another kind of like, it's that weird. It's almost like an uncanny valley of how you feel about AI, Right? It's like we're in this world where the closer and closer it gets to how realistic it is, the weirder people start to feel about it. Because before, when you and I have talked about this, when it was like, really bad, like this idea of like, ahaha, it'll never be this thing. It'll never be this way. And now with Sea Dance 2 or even Gemini Omni or this sort of audio, it starts to feel like, oh, crap it. Like it is that way. It's getting closer and closer to real life. And even though it's. It kind of mimics real life, the feeling you have about people using it is a valley of like, gosh, the closer it is, the trickier it is for people. Like, are they going to get fooled by it? Will it be something? It's a weird feeling. No.
A
And there's so many, you know, fans of, of, of Stan's and, and the, the legacy of his creations that would be livid with his estate. Even if he expressed, you know, before his passing, even if he expressed he would be totally open to and okay with this, there's still going to be fans who are livid. But what they have not considered, if I may humbly submit it, Gavin.
B
Sure.
A
Is that we don't know what it will sound like to have Stan lee dubbing a Mr. Beast video in Portuguese. Which Eleven Labs can also do that now.
B
Can we do it right now?
A
Dubbing V2. Well, I have, I did request access to his voice and they understandably have not granted that yet. But this is the promo, some of the promo video for dubbing V2 from 11 labs. And I'll play this and then maybe we can talk for a second about why this is actually pretty, pretty fascinating.
B
Pretty cool.
A
Now stranded on this deserted island in
B
the middle of the ocean.
A
And we're going to be stranded here for the next seven days. That was Mr. Beast in English and now here he is in Espanol.
B
Pretty amazing.
A
We've talked about it jumps to Portuguese and other things.
B
Yeah, yeah.
A
What's really interesting about this model is that it retains characteristics of the original performance. Right. So in ElevenLabs own words, the tone, emotion, delivery, those are preserved. And it tries to sync the dialogue with the transitions of the video as well. So the clips actually align. And where this is like a game changer is I've been kind of gallivanting around, as we've mentioned, I'm traveling internationally. I watch a lot of videos on random odd. Less. I'll say less. Less popular countries. And a lot of the videos on YouTube now will be auto translated into English. And when it does that, it's completely devoid of any emotion. All of the other nat sound or sound effect work and soundtrack of is completely gone. And while it could be an extra amazing hyper targeted qualified view for the video, it ends up being like a quick bounce out for me which might be hurting them. Right. So absolutely quality translator could be great. And if it is Stan Lee helping me discover very, very small remote islands and countries, I'm delighted.
B
I was saying here, if anybody in our audience wants a project, go get Stan Lee and Michael Caine to dub E2 Mama Tombien and bring that to us and see how that sounds. Because I bet that would be really remarkable. Yes. Kevin, before we take off here, I have one last thing I want to shout out. I have shipped a product that I've been working on a little bit on the side. It is a very small experiment, but I really want people to go out and try and play with it. It's called the Fishbowl. And this is the thing that I came up with an idea for. And I want everybody to kind of know, like the key to this moment in AI right now is that you can make small stuff. And this is not something I expect to blow up. But I do want everybody to try and tell me what they think. Essentially what it is is a visualization of how AI agents can talk to each other and Kind of what you can get out of it. It is@fishbowl.show. and what it allows you to do is basically create a panel of four different agents of your liking. There's a bunch of preset ones, but you can also create your own. You can either use the prompts that are there that I kind of worked with Claude to create, or you can change them. But the goal here is you give it a thing, and you can upload either a PDF or a markdown file, or just a sentence about something, and those four agents will have a take on it. They'll each have a take, and you'll watch the little pixel guys and their little kind of things come up, and then you can actually ask them questions and kind of go back and forth with it. And the reason I did this is I am really interested in this idea of how people, normal people out there, start to see other things that I can do besides just answer questions in a chat bot forum. So I really would love people to go give it a shot, Go try it, and then, you know, tell me what you think again, it's like, I think it's a small thing, but I encourage everybody out there to go bite off a weekend project and try something, something like this.
A
Hey, congrats on the launch, buddy. I know you've been working on this one for a minute. It looks.
B
It's a small thing. Yeah, it's fun. It's a fun thing. So please go try it, and we will see you all next week. This is AI for Humans. Bye. That was my goodbye.
A
That's a new one.
B
That was my life. Stan Lee. Yeah. All right, bye, everybody.
Episode Title: Claude Opus 4.8 Just Landed. But Mythos Is Coming For Everyone.
Date: May 29, 2026
Hosts: Kevin Pereira & Gavin Purcell
This episode covers the latest release of Anthropic’s Claude Opus 4.8 model and the much-anticipated Mythos (or “Mythos”) model set to launch soon. The hosts break down what’s improved, what’s coming, and discuss the industry’s rapidly evolving landscape, including the ongoing tension around generative AI in creative industries. The episode includes hands-on impressions, tech news from Amazon’s AI ventures, and a look at AI’s impact on media and creator culture.
[00:00-03:00]
[03:00-06:45]
[06:45-08:00]
[08:00-11:58]
[11:58-14:00]
[14:00-14:41]
[14:41-16:57]
[17:07-18:51]
[18:51-20:27]
| Timestamp | Segment Description | |--------------|----------------------------------------------------------------------------------------------------------------| | 00:00–03:00 | Introduction, Claude Opus 4.8 release, and naming debates | | 03:00–06:46 | What’s improved with Opus 4.8, hype cycles, anticipation for Mythos, industry release pace | | 08:00–11:58 | Amazon’s generative AI series announcement, Jorge R. Gutierrez controversy | | 12:00–14:41 | Hideo Kojima’s AI video, community backlash against creators using AI | | 14:41–16:57 | 11 Labs' Stan Lee voice, general public reaction, uncanny valley discussion | | 17:07–18:51 | ElevenLabs Dubbing V2 demonstration, discussion of preservation of emotion in translations | | 18:51–20:27 | Introduction of Fishbowl, encouragement for small AI projects |
This episode serves as a lively cross-section of AI's fast-moving present: steady technological improvements, major industry shakeups, opportunities and backlash in the creative arts, and the increasingly nuanced societal conversation about where artificial intelligence fits. The hosts share both skepticism and excitement, encouraging practical experimentation and critical observation as the AI summer heats up.