← AI For Humans: Weekly AI News, Tools & Trends
AI For Humans: Weekly AI News, Tools & Trends
Google's Nano Banana 2 Just Dropped. We Tested It. We Have Notes.
00:58:08
Loading summary
Transcript196 lines
- [00:00]
A
Nanobanana 2 has landed and Google's new image model is much better at editing and it's way cheaper.
- [00:07]
B
Yes, but it is not perfect. And wait, what's happening right now?
- [00:12]
A
Oi. You see google drop nanobanana pro 2. Those fools on AI for humans prob didn't even get early access. It's fine, Google. We'll just sulk in the corner and we'll play with Sea Dance 2.0 because we can't have your nice things.
- [00:32]
B
Speaking of Sea Dance, this week, Logan Paul of all people and the Dor brothers lit Hollywood on fire with a 15 minute sea dance 2.0 video. You're screwed, Buddy.
- [00:51]
A
Sea Dance 2.0 is now available in Cap Cut here in the United States. And we will walk you through some of the best practices so you can maximize your memes.
- [00:59]
B
Plus plus Anthropic seems to be the AI company right now. As Claude drops new updates. It might have tanked the stock market. And CEO Dario Mode is getting into fights with the US Secretary of War.
- [01:10]
A
In my view, so close to these models reaching the level of human intelligence and yet there doesn't seem to be a wider recognition in society of what's about to happen. I can tell you all what's going to happen, Gavin. It's not that. You see, my assistant, Mr. Tibbs is going to absorb all of my free time. And even then I will be lost to the the permanent underclass. You see, the final set.
- [01:34]
B
Kevin, Kevin, Kevin, Kevin. It's okay. It's okay. Deepseek v4 might come out this week and just change everything all over again.
- [01:40]
A
Oh, well, I guess. I guess that works. Also, we're in the golden age of vibe coding. I forgot about that. Thanks to Claude Code people are destroying billion dollar businesses like Frame IO and maybe Spotify is up next. Gavin, I'm gonna take them down on our show today. Wow.
- [01:54]
B
We also redesigned our website from scratch. We did. Yes, we did.
- [01:58]
A
What's that cute little clickable smiley face thing? What is that?
- [02:02]
B
I'll tell you how this myself in about six hours and saved us some cash.
- [02:05]
A
That is. It's actually kind of adorable.
- [02:08]
B
We are adorable, Kevin. This is AI for Humans. You know what you want to be.
- [02:13]
A
Yeah. Finger hearts.
- [02:22]
B
Welcome everybody to AI for Humans, your weekly guide into the wonderful world of AI. We have amazing stuff this week. We have a new Nano banana, Nano banana 2. We also have an incredible video from the Door Brothers we're going to dive into and all sorts of interesting new anthropic stuff. But Kevin, we're also going to try something throughout the show regarding a new music model that's a little bit. Let's call it sketchy.
- [02:43]
A
It's super sketch. It might be shut down by the time you hear this, but I'm going to have my Open Claw assistant try to build us a custom Spotify that Reimagines the top 100 greatest songs of all time. Okay. That's basically what I'm telling it to do, is to take the top 100 songs, feed the lyrics and the artists into the machine, and give me back versions of the songs so that I can live like I'm in a bizarro dimension, basically. It's like, you know, you took the antenna for your radio and you tuned into some weird alien frequency, and that's why Elvis Presley sounds a little off. Or the Eagles sound like.
- [03:17]
B
Okay, well, hold on, let's. Let's just see what we get. Let's see what Mr. Tibbs does. You're going to send off Mr. Tibbs? We're going to come back with that later. But first, Kevin, we have to talk about Nanobanet 2. This is Google's brand new image model and it just launched this morning and we are very excited to play with it. We did not get early access as some other people did, but that's fine because they did release it right away. There's some interesting stuff about this. I want to say first and foremost that in my initial experiences, it's very good. It's very good at editing images. And if you saw that video, the two bananas at the top, I originally created those images and both of those were then animated in Sea Dance 2. But they had little Adidas logos on their jackets because that's what it came out with. I asked for tracksuits. Sea Dance would not animate those with the Adidas logo. So I went in and asked nanobana Pro. Yeah, it was interesting. Just remove the Adidas logos and it did it perfectly. So even if nothing else is amazingly great about this, not a huge step up like that alone might be a big deal. So some basics here. It's much better at reasoning, which is a good thing because sometimes image models kind of suffer at that. It's also very good at creating storyboards. And this is something I will say as I've started to do more AI video generations, especially, especially with Sea Dance 2, a hugely important thing when it comes to consistent characters throughout stories. It is really interesting to see what you can create within these new AI video models. But if you don't Have a consistent character story arc, and you're trying to use image to video, it can be kind of a nightmare. Also, another big thing they've done is infographics. There's some amazing examples of what's possible with infographics. There's a great weather one, which, I mean, who doesn't want a great weather infographic?
- [04:49]
A
Plus, I will say that one of the interesting things about that is the fact that it can go and use all of the power of Google intelligence to pull in accurate data. So it's not just like, make me a cute chart in a claymation style about the forecast in Palm Springs. Here's the data. You could just say it and it will go out and crawl it and make the data accurate. Like that's.
- [05:09]
B
Yeah, that's exactly right.
- [05:10]
A
World model understanding. Yeah.
- [05:12]
B
And Suddhar Pichai, the Google CEO, actually tweeted out a demo of an app they created called Window Seat, which is using nanobanana2's world understanding to generate accurate views of any window in the world, which is a very cool kind of thing. Now when they say accurate, it is accurate in the AI in the AI world. So we have seen some pretty cool examples of this. Ethan Malik, as always, is doing interesting stuff. He showed a very complex image of a kind of a Where's Waldo thing that was pretty cool to look at, like when you see all the different stuff in there. I tried to do something really complicated because I keep always wanting to think, okay, what is hype with these models and what is not? And again, we did not get early access. So this is. There's nothing stopping us from saying what is or isn't good. So, Kevin, if you could look at my test. I tried to create the periodic table of the elements, which is a lot of content, and I tried to get each of them to have an original emoji that represented that element. And you can see the first time back, it was not great. Right. It only got like even like four or five of the elements in there. And there's some mistakes in the text on there. Now, granted, this is a lot of plants.
- [06:16]
A
Diamond cactus. Those are all necessary, basic building blocks of our universe.
- [06:21]
B
Yeah, I mean, it's trying to do like calcium and all that stuff. Like, it's actually trying to figure out a way to kind of connect the dots there. But then I did actually upload it. Somebody said that AI Studio is a better thinking model and to use it in AI Studio, you do have to connect your API. So I did that. It thought for like a. You know, 130 seconds it still was okay, so just know this isn't like it's not solved image. Like complex images are not solved, but it is much better than it was before. And the most important part, Kevin here may be it is much cheaper. So you out there maybe on the consumer side or just like you have a Google Gemini subscription, if you are an API user, this is way cheaper than it was before. And the quality level is as good as Nano Banana Pro. So that is one of the biggest updates here, I think.
- [07:04]
A
So which businesses just went out of business this morning or what's the new business that we should be looking at because of these new capabilities?
- [07:13]
B
Well, I mean, I think we're going to talk about more AI movies in a second. But Google also updated Flow, their AI studio. So there's a lot of companies that were funded as AI studios and Flow is a really pretty terrible product. I will say my experience with Google Flow has been bad. That is how you would generate VO3 videos before. What they're trying to do is create an all in one studio. And Kev, something happened this week actually and put this in the rundown, but I forgot to mention this. You know, Runway is now offering all of the models, which to me was a moment of oh, maybe Runway realizes or they're folding out a little bit of the race to get to the kind of state of the art model and they're gonna become another one of these Krias or Higgs Field. And again, this is no shade against Runway. They've done incredible work so far. But like it might just be really hard to compete with the Googles, the OpenAI's and the ByteDances of the world because you just don't have the money, right? You don't have the money to do it.
- [08:03]
A
Yeah. When they made that announcement, the number one comments were always, it was all like, oh, capitulate much?
- [08:07]
B
Or way to.
- [08:08]
A
Way to wave the white flag. But look, I'm sure they're still cooking on something that's unique for them or some unique tool. But yeah, to your point, maybe you just let others try to win the foundational model race and you bolt on an incredible interface or some crazy plugins that do incredible things with those models. That might be the way to go.
- [08:28]
B
Although we will talk about this later, those interfaces and plugins are very vibe codable now. Right. Like we'll talk a little bit about what you can do and how you can make that stuff. And when I say that it's hard to imagine a company Again, it's not shit against Runway, but Runway's I think, taken in a lot of funding. I think they're valued in the billions of dollars. It's hard to imagine them like be getting a lot of extra value out of that stuff. But again, maybe it is. I do want to say one more thing before we move on. And it's time for Benchmark Boys. Benchmark Boys are back. Because the Nanobanana 2 is the number one image model in the artificial analysis benchmarks. This is a big deal in some ways because it just means that the experiences people have prior to this launch, it was launched as a test, is very good. Go try it yourself. NANOBANANA 2 is live in Gemini right now. You can actually go use it right now and have some fun with it.
- [09:18]
A
Yeah. So for those hearing that, where do they actually go? Do they go to aistudio.google.com do they go to Gemini Flash Experience?
- [09:26]
B
You go to both, Kevin, you go to both because it's available in both. It is available to you in Gemini, but if you are an AI studio user, which is kind of Google specific, you know, kind of almost like their vibe coding studio.
- [09:37]
A
Yeah.
- [09:37]
B
You can definitely, I think, make sure that you're activating that thinking model. Now if you're a Gemini Pro user, where you actually pay for Gemini, there is a dropdown menu where you can go to Pro Generations. So I would encourage you to do that. But go try it today. I mean, this is one of those things. It's a new image model and according to Google, it is a full level up from the previous one. I kind of think this feels like an incremental change. The editing part is really good. The cheaper part is really good. The actual creation of photos seems fine.
- [10:06]
A
You sent me an AI for humans poster which at first glance the thumbnail was, oh, wow, this is like a generational leap. And then when I actually double clicked and looked at it a little bit closer, I was like, oh no. There's still some very like some telltale stuff going on. But the layout of the AI for Humans AI poster, the theatrical experience is, is actually pretty good.
- [10:27]
B
Yeah, I mean the goal here would be, is like you put a prompt like that in which was like, make a poster about the podcast AI for humans and make it cinematic. Design wise, it was fine. It did eventually get my. It made me look very businessy, which of course is probably because of my. My picture that exists out there.
- [10:41]
A
That's your most popular headshot.
- [10:43]
B
Yes, my popular headshot. What? It didn't do was what people are kind of purporting that it should be able to do is get all the text right. Right. Or. And there were some mistakes in the text again. And like, this is the sort of thing that like, that's the next level that will come is when you never see mistakes in text. Now, again, it's the same way with hallucinations. Maybe you'll just see them always, but if you are, if you continue to lower the bar where you're seeing less and less of them, I think it's going to be much more useful. Speaking of raising the bar, Kevin, we should switch over to talk more about Sea Dance 2. So if you missed it. Yeah, so I use Sea Dance 2.0 to generate that banana video. We're going to talk a little bit more about some of the stuff I did with Sea Dance 2. But this week, the Doerr brothers, who are well known AI video creators, partnered up with the Paul brothers, who are well known Internet. I want to be very cautious. Your word. Well known Internet rap stallions, superstars, their
- [11:34]
A
expert fight makers and knee takers. I mean, really.
- [11:39]
B
So they, they partnered up and they made a 15 minute movie called I'm a Good Guy. Logan Paul actually tweeted this out. He is definitely involved and his brother's involved in this too. This is 15 minutes of what I would refer to as completely passable direct to video quality film. Now, many people out there, this became a very divisive thing because the Doer brothers, as per usual, did a tweet that said something like, we took seven days to make a $300 million movie. And I'm sorry, if you have spent $300 million on something like this, you have made a big mistake. But you could have spent $5 million conceivably on something like this. And I will say it's 15 minutes. My response to this, I want to hear what you thought. But my response is there's way too many fight scenes. The biggest problem I have with this more was in fact, by the way, really cool visuals in this. Like I thought all the visuals of the. That there's a bunch of these kind of guys that all look the same and when they kind of fall over, it looks great. And the biggest issue I had with this was writing and editing. Kevin. It was like it could have been written and edited better, right? Like, if that was better, like, it actually would have been super compelling.
- [12:39]
A
It's so funny that I like, I saw the tweet storm that was fired off when you had mentioned your Thoughts on this video. And that. That, to me, was also the takeaway. It wasn't like, oh, the characters don't look consistent, or the emotion in the actress's face isn't believable in scenes, or the dialogue is clunky or anything else. It had nothing to really do with the tooling. It had to know more with the. The artist human element of assembling those tools and getting the product out. And that is like, I think it's still a watershed moment for generative art. Like, this is, again, forest for the trees. If people are looking at this and going like, oh, well, it wasn't the most entertaining thing. It's like the. The goalposts have really been moved for cinema.
- [13:21]
B
I have this bet with a couple of friends of mine about AI Video and, like, kind of when there would be a hit AI video. And to me, this is like a huge benchmark. On the way into that space, we were talking about 15 minutes of consistent quality. Could you edit it down? Were there a couple places where it wasn't as good? Yes, but this was 15 minutes of actual footage. And to be clear, this is not the. The. The machine did not generate 15 minutes of this. This was generated in many pieces and then stitched together. Right.
- [13:47]
A
You know, this is nothing compared to
- [13:48]
B
the time I fought Tyson. I know
- [13:53]
A
once we get this guy, people are finally going to see us for who we really are.
- [13:57]
B
Yeah. As the good guys. And then they'll finally stop trying to kill us.
- [14:00]
A
I'll be nice.
- [14:01]
B
It would be. Imagine all the merch we could sell the good guys action figures.
- [14:06]
A
The good guy T shirts.
- [14:07]
B
The creator of Entourage, Doug Allen, who was on Twitter @MrDougallen, actually tweeted about this video, which is a really interesting response to me, and kind of gets to the point what we're saying. Before, he said, looks kind of good, but could. But could not be more bored by. This lasted about 80 seconds. If this is the future, reality TV will be mine. So I think, again, this is not. He's talking about being bored. That's not the fault of the AI Video.
- [14:31]
A
Correct.
- [14:31]
B
Right. Like, that's a storyteller.
- [14:34]
A
We're not talking about the future. We're talking about what was released days ago. We're talking about the present. Like, people are already right now working on the future. So that is important to keep in mind. But again, if the takeaway still is that taste and the human artistic element are what are going to be the differentiators, then great. Great for everybody that's looking at this going well? I could make better and I could do better. Great. Please do.
- [14:58]
B
So let's talk about Sea Dance 2.0 and some of the stuff that's happened over the course of this week. And I actually got some time to make some stuff that I'm going to show in a second here. First and foremost, it is available now in capcut, which is an interesting thing you have to pay per gen. If you're a paid Cap Cut user, you get a few actual things to show. There's a lot of great people that are already making things. But I want to show Kevin this little video that I made within capcut to kind of demonstrate that it's live in Cap Cut. So play that here real fast.
- [15:40]
A
The hot dogs are out of the bath and on the hardwood floor. Yes.
- [15:44]
B
Yeah. So if you notice the very end of that, it says cap clut, which made me laugh, like it's not actually Cap cut. So it got most of the. But they did jump out of there and they fell on the floor. Kevin, there's another big story around this that supposedly. And I think that there's a lot of debate whether or not this is real. It is from the taker of Underscore Wiz Twitter handle that the weights were leaked for this. And Kev, do you want to explain what that might mean and what kind of implications that might have? Yeah.
- [16:09]
A
When these models get trained, it's, you know, it's all about proximity of one thing to the next. It could be a word to another word, a character to another character, a pixel to a pixel. It's like those are the. I'm distilling greatly here. But the weights are what allow the model to predict the next image, the next word, the next whatever. So here basically, if these weights were released, if they got leaked, then conceivably someone could make their own front end and access their own model using these weights. So, you know, look, we talked a lot about Hollywood going after China and saying the cease and desist and we're going to sue to oblivion and maybe see dance will react to that and play nice and they'll figure out a business model or whatever. But if the code gets out there and a hacker community, an open source community can go take and do what they want with it, there's no stopping, there's no censoring, there's no guard railing. So conceivably you could generate anything that was trained into this data set. And that is a. That is wild if that is.
- [17:06]
B
Yeah, well, and we should talk about that. Like, you know, there's a lot of conspiracy theories going on around the fact that, like, China releases themselves so they could kind of destroy western entertainment. But I do want to say one thing. There was a lot of debate whether that Tom Cruise versus Brad Pitt clip was real. And I want to show the creator of that clip, like, basically said it is real. And he made a second video to kind of respond to that.
- [17:28]
A
Wait a second.
- [17:29]
B
That blogger guy really thinks this text to video generation needed a video input.
- [17:32]
A
What?
- [17:33]
B
This doesn't even make sense. Yeah, he's saying the model has like a database of clips being loaded to drive the motion instead of just training the model in those videos and generalizing the motion in the model itself.
- [17:40]
A
Sounds dumb. Anyway, want to get back to fighting? Sure.
- [17:42]
B
Anyway, he was a good man.
- [17:43]
A
Crazy.
- [17:44]
B
All right, Kev, do you want to tell. Make sure people. We tell people what we just saw there just in case they're not. They're not watching the show.
- [17:50]
A
I mean, we saw. I mean, we definitely saw Ryan Reynolds. And then. I don't know.
- [17:56]
B
I wasn't sure who that other guy is. He's somebody famous. Tell us in the comments. Somebody tell us. We're not there enough.
- [18:01]
A
Andrew Tate. It's. That's the mashup. It's Alpha Affleck. Yeah. Yeah. He's trying to sell us on a new token. But it is. I think that's who it is. I don't know who the voice is necessarily, but it's like a Ryan Reynolds person, a Ben Affleck Tate character. They're on a rooftop yammering about. Yeah. About how you could make this without needing to film actors on a green screen. Like, it's phenomenal.
- [18:22]
B
So I got some time play with this this week, and I got into the Cap Cut Creative program and spent some time playing. It's actually really amazing. It is frustrating, like all other AI video that, like, you have to wait for outputs and. And you have to spend some time. But I will say it is doing stuff in the edits that is unlike anything I've seen before. I want to play this thing that I made in 90 minutes. Kevin, if you follow me on X, you might have seen this already, but this is a video. I just came up with an idea. I made it over the weekend, like in. In 90 minutes in the morning, and I just put it out. So play this. I think it, you know, it's pretty short. We can probably play the whole thing.
- [18:53]
A
Sometimes things aren't all they're clucked up to be. Okay. So the chickens have come home to roost. Things are not always as they're clucked up to be. I love the egg laying in the middle of Manhattan. That's delightful. The process for this, Gavin. This was all text to video.
- [19:28]
B
This was all text to video. There's a variety of different things that are better and worse with it. I have found that text to video is really good. The one thing to be aware of is that CDance 2.0 has an Omni model, which is better than Cling's Omni model, which you can just drag stuff into. And it will kind of try to make its best version of this. But this was all text to video. I wasn't trying to keep, like, you know, one character the same throughout it. That helped a lot. But I will say also the fact that I was able to string together this many clips was pretty impressive. And again, most of this was first generation. I wasn't doing stuff multiple times. A lot of this is editing. You know this and I know this. Like. Like, it's not like this stuff all kind of comes together on its own and magically gets connected. I edited it. I picked pieces of shots that were off one another. If you look at that tweet that I put up, you'll see I. I opened up a couple other shots where there were definitely bad AI video stuff, and you have to trim that off. I added the voiceover from 11 Labs. I added a music track. Like, all that stuff is added. And so you get a sense of, like, if you think of AI video and it's something like Sea Dance 2 is the. The kind of next level of it. If you think of it as like the con, the kind that might come into you from the field shoot. Your job as the editor then is to kind of take that stuff and make the best version of it. So that's kind of what I think. This gives a good example of it. There's a lot of interesting, like, really good prompting tips out there already. Alex Prompter has some good tips involving, like, cinematic words and things like that. I also. Kevin, I want to play one more thing that I made that I sent you last night. And I think this is a good example of why the Hollywood might be so mad, but also what makes it so interesting. Play just that clip of the polar bear in the bar, because that's kind of what I'm working on right now.
- [21:12]
A
I was supposed to be on vacation, Ralph. I'm not the one who left you, Freddy.
- [21:18]
B
Okay?
- [21:18]
A
So I know we're going to get into some like, good prompting tips. But how did you get George Clooney out of the machine? Did you spend five hours fine tuning that?
- [21:28]
B
So here's a couple things about this. So if you're just, if you're just listening, what that is is a very kind of interesting looking polar bear in a Hawaiian shirt sitting at a bar. And the acting, Kevin, is the thing that I am so impressed by, like the acting. I did not ask for it to be this, but it kind of implemented this idea of what the idea of the words were going to be that that bear was going to say. It had it wait for a second and kind of grab the glass and kind of tink the glass is the kind of thing you would do if you. If you were nervous. The fox, I prompted. So there's two shots. There's a polar bear and a fox. The second voice is a fox. And I prompted as anthropomorphic animals. Right. I didn't say anything about the fantastic Mr. Fox. I did not say anything about that particular idea of a fox. But if you look at this video, it is very clearly, if not influenced directly by a complete lift of the Fantastic Mr. Fox. And as the voice of the Fantastic Mr. Fox in the movie the Fantastic Mr. Fox by Wes Anderson, George Clooney was the voiceover artist for that movie. Right. So what is interesting and in some way in its system, it is making the connection between anthropomorphic fox, fantastic Mr. Fox and George Clooney and giving me this performance, which by the way, is a very good performance, I would say, like that it's a very good performance. So. So the thing that I take away from this model in general so far, after having spent some more time with it, is it is very capable. You should definitely be spending time understanding what it can do. And I think the acting part of it is the thing that kind of keeps shocking me. Which you also saw in the Log Paul video.
- [22:54]
A
Yeah. So, okay, so prompting tips wise, I mean, there's a few people that have done guides and whatnot. What did you discover? What is actually working? I've seen things like always add cinematic or 4K to something because that cinematic makes things look better.
- [23:08]
B
Cinematic actually makes a difference. The word cinematic. One thing that does get frustrating, they have put guardrails on it and sometimes you just won't understand what it is. So you kind of have to go through and poke around. I have found, like I said, text to video with the Omni model actually is much better than trying to do image to video first. And Last one thing that worked with this polar bear thing that is really fascinating that I think people should try if they get access, you make a clip. And then when you make a clip, you actually grab that clip and use that as the omni model. Because the omni model can take video stills, audio, all sorts of things. So I basically take the 15 second clip, give it to that and say like, okay, we're going to. This is the, this is the thing you're using to work off of. And it mostly gets it right now we'll say one thing that's happening and we'll see when this polar bear video comes out is over course of this video, the polar bear is getting much. He still wears a Hawaiian shirt, but his fur is getting much grayer and dirtier. So I think it is spending a
- [24:01]
A
longer time at the bar.
- [24:03]
B
There's a passage of time.
- [24:04]
A
I mean, when I had my Red Bull and vodka phase, I knew that, you know, by 3:00am I was similarly covered in weird substances and grime.
- [24:12]
B
So, yeah, there's been some other amazing videos that drop. Before we move on, we should show this energym video from AI Candy. This is an AI studio. Watch this.
- [24:20]
A
By 2030, almost 80% of people had lost their jobs. They had no, no purpose, but they had a lot of time on their hands. The less people actually did physical work, the more they wanted to appear as if they did.
- [24:37]
B
What if we could use the energy
- [24:39]
A
of humans to power the machines? I mean, again, the Sam, Old Musk and Bezos, great. But Sam Altman.
- [24:48]
B
Chubby Sam Altman. Chubby Sam Altman. So if you ever again go watch that, we'll drop it in our show notes and it's super fun. Kevin, you know what else we're going to drop in is a new website that we have made mostly because we are here for you, everybody. That's right, AI for Humans. Each week is here to do stuff for you and to make things for you. Go to our new website, but first and foremost, like and subscribe this video. We're going to talk about how I made that website later. But more importantly, we need you to like and subscribe our YouTube channel. This is how we grow. Every week we try to kind of come out with new things and work on new stuff. Kevin, our Patreon tiers are kind of heating up a little bit too. I'm very excited about that. And as always, our discord. Please stop in our discord too. So if you're out there, come join us at aiforhumans show and all the links are there.
- [25:33]
A
Yeah, go click on the new site, browse around, really have a time with it. And make sure you subscribe to the newsletter as well, because Gavin's been pouring his heart and soul into that thing and the line is going up and we want it to go up steeper. Make it something I would tumble down if I was looking. Looking at my phone and trying to walk up, please. And thank you all very much.
- [25:50]
B
Kev. Speaking of that, speaking of making things, we should talk about Anthropic, because we have been talking about Anthropic quite a bit over the last month or two and ever since the kind of cloud code moment, from the December moment we have entered into kind of Anthropic's world, I saw a stat that really shocked me, which was that Anthropic's growth is like outpacing ChatGPT and OpenAI by three to six times right now, which in the last month. Which makes sense because people are finding Opus 4.6, they're finding Cloud code, they're finding all this stuff that's really interesting. Another viral piece hit over the weekend from a financial investment group called Citrini. Citrini, and it was all about how if we were able to do all this stuff and make all these new pieces of software, that the essentially the economy would crash. And this article, ostensibly about kind of Claude code and all the stuff people are making right now did crash the stock market on Monday. There was about a huge correction. So anyway, I'm curious to know kind of how you see what's going on with Anthropic right now. Obviously we get into more stuff that's going on there, but do you feel like they're starting to kind of maybe, I don't want to say win the race, but they're at least pulling even.
- [26:53]
A
Yeah, I mean, look, Google released a brand new model last week, which by, by all accords, is a phenomenal foundational thing. And I'm still overwhelmed. Granted, anecdotal, my bubble within a bubble, but I'm overwhelmed with new Claude code this and new that.
- [27:08]
B
Yes.
- [27:08]
A
And even though OpenAI's Codex model is also their latest one. Phenomenal, really good. I actually use it to sanity check a lot of anthropics work because I use cloud code as my daily driver, but that's what I'm driving to. I still use Anthropics models for most of my stuff because it seems like every other day they've got the machine helping build the machine. It seems faster than most. And Their releases, like all the tooling that they have, is just more impactful for my daily usage. Now Coworkers has a bunch of new features that is driving towards this, like, agentic future where you can connect it to all of your things. HR plugins.
- [27:44]
B
We don't have to deal with HR anymore, Kevin. We have HR plugins. Sorry, HR people. I'm sure you're out there listening. I don't. I love you HR people. You do good work. But there are plugins now for hr.
- [27:56]
A
Yeah. So, like, look, the, the, the Citrini memo, it was, it's interesting to me because. And this isn't like shattering my hand patting myself on the back or, or you, or this podcast, but I'm like, we, we been new, as the kiddies say, like, this is, it was, there was nothing in there. It was presented in a very interesting, elegant way. A memo from the future about how we ended up in that future. I totally get why. That presentation was great. But like, of course, I mean, Sam Altman has been out there himself saying for years, AI is going to be a better CEO than anybody else. So if you're, if you're sitting back or you're sitting on the sidelines and going, well, okay, it's coming for Legal Review, or it's coming for hr, or it's coming for the storyboard artists or whatever, I'm good though, because I'm the idea orchestrator or I'm the one who runs the thing. I keep the trains. It's coming for you as well. Yeah, like, that's, I feel like that has been known, but maybe it just hasn't been.
- [28:52]
B
Well, here's what I'll tell you. There's been a lot of people who've kind of dismissed this idea because the whole memo and the crash of the stock market was based on this idea that, you know, essentially if, if, if, if people start making these things and code gets automated, there's no need to really pay for these large companies to do stuff. Right? Which I think in some ways people were like, oh, these large companies will exist. I think the, there's a, there's some counters to it, which is like, the more that people make this stuff, the more new things that will come along. The thing I keep coming back to, which is to your point is I do think you're going to see like three corpo states, right? Like, that's what we're approaching. Like, there's a lot of sci fi that has like, these corpo states where there's like these three Major corporations. If, you know, like any cyberpunk stuff, that's kind of where it starts at. My thing is like, well, you do have to figure out a way to pay for all these other people to have this kind of, like, life where they make podcasts or they do all this stuff to kind of entertain people. And I don't know if every single person's going to become an entrepreneur. Right. And we should say very quickly, like, Claude is shipping like crazy. They're shipping new Cowork plugins. Like Kevin said, there's a remote control for cloud code that's coming so you can do your cloud code when you're walking around and keep it going, which for some people is good. Some isn't Scheduled tasks for Cowork, which is kind of their version of what you could do with open cloud. Basically, you'd be able to schedule it ahead. I think this is all really interesting because what it's doing is it's kind of elevating that feeling you had a couple weeks ago in me at least, which is, oh, I now have to do. I have to be coming up with new things and doing new things all the time. Because if I don't, I am going to get stuck in the permanent underclass. I am going to get stuck in a place where I haven't made something for myself. But. But I just think it's all going to move so fast. And again, we've said this on the show for years. When we hit this time, and I think we are in this time now, it is going to move very fast. Right. So 2026 and 2027 are going to
- [30:36]
A
be crazy years if there are, if we're lucky. And there are more than four, let's say there are six mega cores as we're talking about, that own all of the intelligence and everything requires that intelligence on demand, constantly humming, building, creating, crafting, whatever. We're all going to be beholden to one of those six providers. We will all be paying them for intelligence to lead our daily lives or run our small businesses, which makes little sense in that case because they could just run our businesses.
- [31:05]
B
Yeah. Who's going to run that business?
- [31:06]
A
Right? Yeah.
- [31:07]
B
Who.
- [31:07]
A
Who has any credits or tokens left to buy any of the things that are powered by intelligence? And, you know, are we really ready to have those discussions about, like, so many people don't, like, it's a gift to, like, your job. There are a few people that actually like what they do for a living. And most people, like, they eke through their Days to get enough whatever so that they can have a roof and have food and then go enjoy their family and friends or their hobbies or whatever. Are we really ready for the other side of that discussion, which is, you know, that might not be how you're able to define yourself in the future?
- [31:42]
B
Well, I mean, it's so interesting.
- [31:43]
A
What are we gonna do?
- [31:44]
B
Yeah, I mean, this is an interesting thing. So I will say just a personal kind of note, like, last four years doing this has been great, and it's given me a thing that I do, and then I have all these other we jobs that I do for my money. I used to have a job that I would spend 80 hours a week at, and I knew one thing that I did. It is a little bit of trying to figure out, like, who you are and what you want to be. Right. So, anyway, we're in this crazy change time. I also want to say, like, there's a big story going on Anthropic this week that there might be news on. Literally, as you're watching this, Anthropic is in a battle with the Secretary of War over the use of Anthropic models in Defense. And this is a big deal, Kev, because Anthropic is really the only model, the only company that is pushing back on the Defense Department. And it sounds like on the Defense Department side, Anthropic is the model they want to use, which also should tell you a little bit about where Anthropic space in the world of coding is. This feels like it is going to bubble up. Now. The thing that is at risk Here is a $200 million Department of Defense contract. And the argument is that Anthropic is saying we don't want to go back on our promises not to use these. This tool for war. And, Kevin, I read something really interesting today, which is saying that some of the people inside Anthropic and think it's really important to stand strong on this because future versions of Claude will watch what this decision that was made this week about from Anthropic and will then know going forward how to. How to respond, which is a weird science fiction thing, to put GR Will
- [33:13]
A
gaslight them and let them know that this actually was never. That never happened.
- [33:17]
B
Never was real.
- [33:18]
A
Fine. They're the good guys, and it's okay. Yeah. Look, this is, like, I am for. And this. I don't mean for this to get political, but there's no way to avoid it.
- [33:26]
B
Yeah, you can't avoid it. Really? Yeah.
- [33:28]
A
I would love A healthy government regulation of industries so that we get our clean air and our clean water and our clean food and our safe medications and blah, blah, blah. I would love for that. I would also love for them to not, like, meddle with a private business and say you have to work with us in ways that you don't want to. People get really upset if, like a cake maker has to make a cake for someone that they don't want to bake a cake for.
- [33:52]
B
Right.
- [33:52]
A
Well, imagine that, but with guns and drones and extinguishing human lives.
- [33:56]
B
So, like, yeah, I have a counterpoint. I have a counterpoint. My. A finger point. A finger point. This is counter. My counterpoint. And by the way, I agree with you, especially this government. You people out there know we are living in a chaotic time. I know there are probably some people who don't agree, but I think Kevin and I both think that this person in charge is not doing a great job. The big issue I have right now is that there is a coming time. And this is in all of the AI science fiction that exists out there, AI 20, 27, whatever, where AI models as they get better, there's going to be a push to nationalize them. So we're going to do some serious talk here. This is some government talk. But when you nationalize something, when you make it a national thing, that means the government kind of takes over control of it. Now, the downside of that would be, is like, obviously they're going to get in the way. It's going to slow way down. There might be a lot of things happening there. The good side of it is, is that the government then owns or at least shares in the profits of the thing being made. So there's a big conversation that's going
- [34:54]
A
to come up here.
- [34:55]
B
On the other downside, though, when you nationalize something like AI, you can weaponize it. You can do all sorts of things. So it is a big thing that's kind of in this conversation right now. And yeah, I think it's good if Anthropic pushes back. Supposedly they have changed their kind of like, safety manifesto just internally right now. I don't. I don't know if that's for sure, but this is a situation to monitor because it is not just about right now. It's really about the next 10, 20, 30 years of the AI space.
- [35:21]
A
So is the pushback that. The pushback is nationalization is good or it's going to happen anyways. So let's go there. I'm just trying to.
- [35:29]
B
What I'M saying is I think that there's a real world that nationalization of AI is going to happen. And I think the important thing to realize right now is that it kind of depends on what government it happens in and who's running the government at that time for what that's going to mean for our country and for the future of AI.
- [35:45]
A
Look, there's plenty of providers they could go nationalize. I have a feeling one is probably launching satellites right now that they're going to be shaking hands with very firmly. I also wonder, by the way, Gavin, and I've seen this. This isn't necessarily an original thought, but I'll put a little, a little sauce on it. Is this all a public display so that they can look like they're the good guys and they're being forced into it.
- [36:10]
B
Philanthropic.
- [36:12]
A
Yeah. That you have a cause. Like Google's in bed with them, like. Right. Google's invested. Amazon?
- [36:16]
B
I don't think so.
- [36:17]
A
They're going to pay. Nice. Is this like a. Hey, we're going to do a public stand here. But then privately models get used anyways or get sourced by other.
- [36:26]
B
Very possibly. I mean, it's possible. I guess the good thing is like anthropic. If any of the companies has always been like semi true to their ideals. Right. They left OpenAI for the exact safety reasons. And. And so the question will become, where does that shift happen eventually? Do you have to do more stuff? All right, Kevin, another huge story this week. Deep seq v4 is the thing that keeps lurking around the world of AI in general.
- [36:50]
A
Yeah.
- [36:50]
B
This is the Chinese new Chinese open source model. We've been expecting it. It has not happened yet. No one really knows why it hasn't happened yet because it's been kind of very much something that's on the precipice. We also talked last week about how these larger AI companies seem like they're trying to get ahead of a Deep SEQ launch. Who knows? It could have come out. When you see this video, the big thing though that did come out was a story from Reuters that said deep seq v4 has been trained on Nvidia's Blackwell chips without authorization, which shows they weren't supposed to do it. But supposedly it has been trained on the highest level chip. So we can expect something pretty big to come out of Deep Seek, I assume in the next week, few days to a week.
- [37:31]
A
Oh, they would just put a tariff on an LLM. Can we do that? Did we just do that?
- [37:35]
B
We could put it. Yeah, we Just did that.
- [37:37]
A
Hey, Gavin. Hey, Gavin. Yeah, yeah, hey, Gavin.
- [37:39]
B
Okay, what if Kevin West.
- [37:41]
A
Yeah, hey, Gavin. Yes.
- [37:42]
B
Yeah, sure.
- [37:43]
A
Again, remember like two weeks ago I had to like, apologize to you and the audience basically in real time because I was like, so sleep deprived and crazy over my open cloth setup that I was like, yeah, barely a husk of myself. During the podcast, we. We flash the graphic. Kevin, Sacir apology time.
- [37:59]
B
Oh, okay.
- [38:01]
A
I owe you in the audience an apology again, if you're wondering why.
- [38:04]
B
Hold on, hold on. Take your time. Take your time with this. I love to hear you apologize to me. So make sure you take your time. Don't rush.
- [38:10]
A
You know what? Scratch that. Run the graphic. Again, Kevin's sincere apology to the audience, specifically, not necessarily to Gavin. If you're wondering, Kevin, you seemed a little checked out during the Nano Banana talk or the seed dance stuff, or even during some of the anthropic stuff, and you were talking about like New World Order and the government working with private AI companies, and you seem like you were kind of half there. Is that what you got these days? The answer is yes, I am sorry. It's because I was very distracted with what I was trying to vibe code during this podcast, which was a good idea, I think, when we started out. And I realize now, Gavin, I should never do this again because, yes, I do owe you and the audience apology. There's one set of footprints through today's podcast that lead us to this point here. Those are your footprints, Gavin.
- [38:51]
B
But now I lifted you through. I lifted you through. All right, let's choose Set to God.
- [38:55]
A
Two sets of feet in the sand. Disregard anything I might have said until now.
- [38:59]
B
No, don't do that. Go. Go back and watch it again.
- [39:04]
A
And realize Kevin, at 20% almost seems like he's making points. Listen, there is a website that is going to be shut down by the time you are hearing this. But I did use it to do something great right now. So maybe we talk for a second about Tsunado and then we could talk about what we just did with it. Gavin.
- [39:23]
B
Yeah, so let's quickly get the briefs on. Tsunado is a website that I discovered through our friend, theoretically Media who covered it, who also said it might be shut down by the time that video came out and it wasn't. What it is is a Suno like website. It is very much like Suno shout out to our friends at suno. But what they. What Suno did and what Tsunato did was take out a lot of the celebrity voices and real musician Voices and real musician training out of their model. Sonato basically allows you to generate music with actual celebrity voices and you can do all sorts of crazy stuff.
- [39:56]
A
They clearly just scraped Spotify and slammed it all in there with tags that are accurate enough that you can genre flip between. Insert your favorite artist here and I mean just straight text to prompt it. I got a shout out at Addison on X because Gavin, you. You put me on blast about the site earlier this week. And I was like, okay, cool. Yeah, I think I know the model they're running the open source model. Then Addison hit me up and was like, dude, you have to check this out. I was like, yes, I know the. The model that they're using. What I didn't realize that was that he was making like Fire coheating Cambria and Glass Jaw and perfect circle covers. I assumed this was, oh, upload the track and generate a Laura and then get it back out. Nope. No, just say words to the machine and Taylor Swift comes screaming out.
- [40:40]
B
Here's. Here's the note. Everybody there. Next time some listen to the first person. Listen to the person 100%.
- [40:48]
A
Run the apology again.
- [40:49]
B
We'll run it one more time.
- [40:50]
A
Do that. Do the I'm sorry sad Kevin graphic. Kevin, I'm so sorry.
- [40:54]
B
So Sonato is crazy. I went in and said, make a James Brown song about the podcast, AI for Humans. So Kev, let's just play that a little sunk a little chunk of that real fast.
- [41:19]
A
Okay, so sorry.
- [41:20]
B
So. So if you know James Brown, you know his band the JB's. Like, that sounds almost identical to the JB's. The voice is James Brown. Even the. Yeah, that stuff like, it does like, is really good. Like, you're right. This is a clear scrape of real musicians. Right. So let's hear about what you just discovered.
- [41:37]
A
Sure. Other audio models did as well, but they took the time to put the guardrails in. Like, look, we know with like Udio, for example, you were able to get Weird Al Yankovic out of it basically by prompting around, using his name. Curly haired accordion aficionado that loves Twinkie wiener sandwiches. Yeah, you'll get Weird Al out of it. But this one is just fire and forget. You just ask for it and it goes. Maybe I'll play some of Addison's covers in a second. Because like I said, he was doing like, coed and Cambria. Whatever. I have a band that I love, the Deer Hunter. D E A R. I prompted them in and I decided to not send the outputs to Casey, the lead singer. Because I do think he would genuinely be upset.
- [42:16]
B
That's really interesting. So I don't know them. I mean, I've heard of them before, but, like, it's not just big artists. It's. It's got.
- [42:22]
A
No.
- [42:22]
B
Like, that's.
- [42:22]
A
Oh, it's got everything. Yeah. And I. And Deer Hunter is. Has a lot of stuff out there, and they're. They're popular. Ish. But they're not on the level of like a Taylor Swift or.
- [42:32]
B
Yeah. James Brown.
- [42:33]
A
Or even a James Brown. Of course not. So. So listen, I don't know how long this is going to be up for. Um, you could have done stuff like this in the past, but it's clear that they swept in enough training data that, like, the results are pretty amazing. And, yeah, apologies that I slept at the wheel, but here's what I did at the top of the show, which had me so damn distracted this entire episode. Gavin. I went to Mr. Tibbs, my AI open claw powered assistant, and I said, tibby, here's the API documentation for Sonata, which they give you.
- [43:06]
B
Right.
- [43:06]
A
Allows you a coder or a machine agent to go and interface with their system. I said, here's a screenshot of the top 100 greatest tracks of all time from Spotify, because they don't want to take time to plug it into Spotify. I just grabbed a screenshot of, like, the top 10 tracks from that list. I clicked and dragged it over to Mr. Tibbs, and I said, go grab the lyrics for these songs and then go genre. Flip them. I want you to take Nirvana, who's in that top 10 with smells like Teen Spirit, but have them cover something else, like Hotel California. And I gave it one other example from that list. Have Boston cover Toto's Africa or whatever. Yeah, it needs to work. And I want you to make the interface look like an old ipod. And this was. I was coming up with this as you were doing the diligence of hosting the show. And I'm sorry. And I'm sure you noticed that, and maybe the audience noticed that as well. Well, I was over here, click, clacking away. I said, go make it look like an ipod. I want to be able to scroll through the tracks. I want to be able to click to generate them, and I want full control and playback. It wasn't one shot, and I will capture the screen so you can see that. I had to go back and forth a couple times and say, hey, this doesn't quite work right, or, this needs a little love. But during the podcast, I was able to jam this out. And I think we need to host it on our website and maybe be ready to eat some of the the hosting charges by streaming these MP3s and maybe defend ourselves in the lawsuit, because I don't think there's anything illegal about anything I'm about to play back here, Gavin, but I'm gonna play some of it back.
- [44:38]
B
Before we play this back, I want to say two things. One, like, as we said, Sonato probably will get a takedown notice and might get a giant takedown notice very soon. So if you want to go try it, you should try it now. It is as S o n a U T O like Sun Auto AI Also, yes, Kevin has made these tracks. We are going to play them. We don't believe there's any way you could take us down off of YouTube or demonetize us for this. I'm going to say not. You can't do that because other people at least as well, but you never know. So if you're out there and you're listening to this and you're part of this right now, this is something very special that may not exist for much longer. So we will try to put it on our website as well. But let's play these tracks.
- [45:15]
A
So, Gavin, I'm on the site now. It says AI for Humans Pod. It looks like a classic ipod. I can scroll through Jailhouse Rock. Queen covers Elvis Presley. We have Nirvana covering Hotel California. The Beatles doing Billie Jean. Michael Jackson covering Queen. Should we hear a little bit of the Nirvana?
- [45:32]
B
Yeah, let's hear it. This is Hotel California by Nirvana. So what you're saying.
- [45:36]
A
Yeah, okay, it sure is.
- [45:48]
B
Get it. Crawl. Sounds like Nirvana. Okay, pause it, pause it, pause it.
- [46:16]
A
Now that's what I call Infringement Volume five.
- [46:20]
B
Something that's so interesting about this, and this is the one thing I think it's important to understand is that, like, there have been people, really talented music artists. What was the name of that, that Tik Tok handle that we love? Oh, there I ruined it, right? That guy doing. Who's been doing these really interesting mashups forever. Kevin literally had Mr. Tibbs, his AI assistant, go and generate these himself. These are one shot, right? These are not somebody, some artist who's making a lot of time on this. And I know if you're a musician who's out there wondering, you know, oh my God, am I in here? It's probably likely you are. And again, this thing won't last very long based on the way that the legal system works with the music Industry particularly. But that was remarkable. Like, that was unique. It felt interesting. I didn't feel like it was a Nirvana song that I had heard, but it was very much Nirvana. Now granted somebody. It's like a deep Nirvana person. Be like, those are the drums from X, Y and Z. Pretty incredible. I think. Pretty incredible. Overall.
- [47:13]
A
The playback controls on the ipod on the website work flawlessly. Play Pause next track menu. It has a visualizer. If you're just, you know, hearing the audio of this, there's a full visualizer I can click and drag in the scroll bar. The progress of the song and it just works. So. California.
- [47:43]
B
It's funny when somebody there, it kind of sounds like,
- [47:49]
A
yeah, sorry, what was it?
- [47:51]
B
What was interesting about that is that you hear the voice change throughout it. Right. There are moments where it's definitely sounds like Kurt Copain and there's other moments where it's almost like some sort of of amalgamation of Kurt Cobain and other rock star singers when he tries to hit specific notes. Pretty fascinating though. That is amazing. Like, what else do we got? Is there one more example we can hear?
- [48:09]
A
We got. I mean, do you want to hear Queen covering Elvis Presley? Do you want to hear what was. You want to hear Michael Jackson covering Queen?
- [48:16]
B
Yeah, let's hear that. I'm really curious to hear what Michael Jackson covering Queen sounds like.
- [48:23]
A
Okay, so we got some Billie Jean inspired stuff here, I think. Yeah. Or Thriller.
- [48:28]
B
Thriller meets Billy Jeans. Yeah, yeah, yeah.
- [48:29]
A
That. Yeah.
- [48:34]
B
That's like a sad AI being. Being tortured.
- [48:37]
A
Oh, hold on. I. I have faith.
- [48:42]
B
Wow. Pretty incredible. By the way, like, this is an amazing. Good.
- [49:03]
A
This is so good.
- [49:06]
B
Okay, I have a question.
- [49:08]
A
Dude. I never understood the react channels when it's like old man, here's tool for the first time. And they're like, what? But now I just fully made the face and did the thing and. Okay, now. So again, Addison and I were going back and forth. He was sharing some of his covers again of like. Because we. We kind of have a similar taste in music. And he's like, what does this mean? I'm gonna just wash, rinse, repeat something that, you know, I had said and that you and I discussed two years ago.
- [49:32]
B
Yeah, ages ago.
- [49:33]
A
The essence or the soul is going to be licensable like this genie. If they. If they plug this bottle up, three other bottles are going to appear. This is a Napster moment for music and for art artistry as a whole. If you don't like it and want to participate in it, completely get that. But I think there's going to be a fork in the road and other artists are going to choose a path. And where I think that path is is right now. I pay Spotify way too much money a month for a family plan, which I gotta kill, by the way. I mostly do it to help subsidize some friends. I give them a bunch of money a month. I listen to maybe 3% of the music that's actually on Spotify at any given time. Right. What I'd rather do is pay a little bit to each artist who I actually like directly. And I would actually pay a little bit more to each artist if I could get some sort of privilege or, or, or, or access or rights to their likeness to what makes them up. Yes. And then I can say, look, I pay for, I pay for Metallica and I pay for Tool, and of course I pay for Ariana Grande, by the
- [50:36]
B
way, Just to be clear, you will never be able to pay for Metallica's essence. They will never, ever, ever.
- [50:41]
A
I don't know. I think they came around on that. I think they've come around that. I think Lars got hit enough with the Napster of it all that they don't want to be on the wrong side of technology again. I do think.
- [50:52]
B
I don't know. Lars, this open, open call. Lars have come. Lars, come on and let's talk.
- [50:56]
A
Yes, we'll do some single stroke rolls together. What's up, Lars? I do think there will be a world though where it's like, look, I, I subscribe to these models or to these, these bands, these, whatever. I could mash them up like toys in my sandbox. I can have one cover the other. I could take multiple artists and have them create something original. And then I can publish that to this player, this Spotify 2.0 or whatever. And Gavin, if you like the song and you want to give it a dollar or you listen to it a million times and a million people listen to it a million times, then I, as the producer, get a little taste and all the artists who were swept in along the way get a taste as well. It is a disruption of the music model. I think it's going to come. I think a lot of people might try it in the interim without properly compensating artists. And they deserve to be whacked like the little moles they are. Not to be Kevin o' Leary with the cockroaches, but this is a site that will probably go down unless they figure out how to properly compensate these artists along the way. And in the meantime, we gotta put this on our website, Gavin. Well, let's.
- [51:56]
B
I have one last thing I must say about this, which is, to your point on all of this, that is a very complicated system that does not yet exist. Right. Like, that is something Vicodin it.
- [52:05]
A
Right now I'm literally vibe coding it. Make me Spotify meets Apple Music with this new model in it. Zero mistakes. I must go viral. If not, you will be deleted. That's what I just told my assistant.
- [52:17]
B
Well, let's see how well that does. I do think it's an important thing. That's an important moment. In the same way that the sea dance moment with actors was in it. We are entering in this world where people are going to have to figure out a way to pay people that have appeared in these models. Not only that, but the people that were trained on these models. Because the Nirvana track you played earlier, like, those are somebody's guitars that were sampled for that. And yes, it's probably Kurt Cobain's guitars or Pat Smear's guitars. Like, but those are people that actually played on that guitar. All right, it's time to talk about what we did this week. And hey, I see what you did there.
- [52:47]
A
Sometimes you're scrolling without a care, then suddenly you stop and shout.
- [53:05]
B
All right. In addition, addition to Kevin's very cool real time ipod that he just made, we have been vibe coding up a storm. And I want to talk a little bit about something that I did which is directly related to this show. And I made us a new website, Kevin, for this show. And it has been something that I've been thinking about for a while. You and I shout out to the podpage.com, the company PodPage, which is a very good website for a podcast, if you have one. That's where we've been on for the last couple years. One of the things, though, that did frustrate us about PodPages, that we are not, at the time, we're not web designers. We were not able to go in and, like, do a lot of tweaks. We ended up having to get stuck with a website that we weren't totally in love with. But it held up. We paid them, you know, a certain amount of money every month. I think it was 30 bucks a month, and now we're paying them 20 bucks a month. But now, Kevin, we are part of the economy that is taking away money from other big companies because we have vibe coded our own website. And Kevin, I want to say I did the majority of this. I did most of it. And I am not a technical person, I think it's pretty damn good. And what was interesting about this? Yeah, what was interesting about this versus a personal website? It is slightly bigger project, right? The personal website I made we talked about a couple weeks ago. Pretty easy. It's a one shot page. This is multiple pages. This is different kinds of design. This is pulling in a database, this is pulling in all sorts of other stuff. And I have to tell everybody out there, if you are waiting to do even more complicated stuff, don't these tools are so good right now. Go to aiforhumans show. You can see it didn't start out this way. This is not a one shot sort of prompt. You are going to have to spend a lot of time going back and forth with your coding model. I used Opus 4.6 for a lot of this and a lot of just back and forth about dragging screenshots. That's my big secret is drag screenshots, right? Take a screenshot, drag it to your point earlier I did use Codex GPT 5.3 Codex High to do a security check on it. And Codex found like four or five things that I then took back into cloud. And I said, hey, a friend of mine looked at this and it fixed those things in general. Just an incredible thing that you can do now. So what was cool about this, Kevin? For me and I'm curious to hear like you know, more of your Mr. Tibbs stuff as well. But like every time I do this I feel like I level up to the next level of like what's possible. So this was a kind of a low to medium job and now I feel like, oh, I can handle that. I did it and I understand what the possibilities are. It's just a very cool thing to be able to pull off the fact
- [55:24]
A
that I'm going to send you a GitHub repo for this AI for humans iPod thing or whatever and I'm going to add you as a collaborator on GitHub and you're going to pull that thing in and tell it to add it as a section, a special section, limited time only on the website, blah blah, blah, whatever. I don't think you're going to bat an eyelash at that now.
- [55:42]
B
No, I totally get it.
- [55:43]
A
You'll deploy to Vercel and you do all these things which even just weeks ago you might not have completely understood. And that's like, it's important to not only like just experiment to learn what the tools are and what they're capable of and, but Also to flex with like the general knowledge that you have because you might not have an intricate understanding of how this website works but you have an understanding now of how you can deploy it or how you can ask for specific sections to be updated or you're learning the intricacies between OpenAI's model versus Anthropic's model and maybe you'll sprint a little faster on the next one because you'll use a faster model because you don't need the all thinking, all dancing model like just at bats are so important with all of these tools. And what's really nice about this is that I think it's think it's a huge upgrade from our last website. This isn't just like a oh I made a personal thing and that's for the and it'll be over there and that's it. Like you made something that's really, really useful for us and the brand. And I hope everybody goes to check it out. I hope people leave us a five star review on Apple podcasts so they can see it pop up in there. That's very cool. And I hope they get to play with our little ipod on there too.
- [56:45]
B
Oh we should also drop a one of those yeah buy me a coffees on there. Anyway, even if we don't upload that other thing which we will upload the of the thing hopefully just so people can give us money on the thing directly and one more path to helping us out.
- [56:58]
A
I'll say yeah, if you go to the website and if you notice that the, the. The the ipod is down, it's either because we've been sued out of existence that quick or we literally can't afford to host the MP3 files. And if the latter is true, we'll put a little buy me a coffee link there and you can do it and that's great. And then I guess we will have just stolen from artists. Maybe that's not the best idea. No point is go to aiforhumans show. That's the point. Go to AI humans playing with all
- [57:24]
B
these tools by the way. Last thing I will say about this, it is again this is the thing that you can do now. It is only going to get better. There are improvements that are happening every week. Kevin's thing is just a matter of connecting the right dots. The open claw world is just connecting the dots. That's how we got that to happen. Kevin is not a technical genius. Yes, he is a technical person but he has learned a lot of stuff doing openclaw. You can do these things. You can do them every day. You can do them. And go spend some time and learn it.
- [57:51]
A
Go get better friends. And I really am sorry about today's show, but how good is the Billie Jean? I mean, come on. I'm gonna go, you know, I'm generate the top 20 tracks. I'm gonna put money into their machine. I'm gonna generate the top 20 so people have more to play. Wow.
- [58:04]
B
All right, bye. Bye, everybody.