A (2:36)

Yeah, I think that's a couple things that are important here. First and foremost we should kind of explain what this feature does. It connects to another feature that we briefly talked about last week called Dispatch and Dispatch. Now I think to use this right now you have to have Claude Pro at least, which is the $20 a month feature. Dispatch is on your phone and on your PC under the Claude Mac app. And again this is only for Macs right now. So I know a lot of PC users are unhappy out there and, and you can connect basically your phone to your Mac and it can do things back and forth. What is different here now is that you can go and have it do things for you on your PC if you're out and about. So much like openclaw, this is one of the biggest things about openclaw is it allowed you to do stuff from your phone or anywhere you were Also Kevin, in addition to this, there's a couple other things that have either shipped or are being rumored to ship right now. First and foremost there is a permissions that have shipped for cloud code which connects it to what WhatsApp and Telegram, which is another thing that OpenClaw did very well. And there's a rumor right now of a thing called Orbit. Now this is a rumor. This is purely a rumor. Orbit is supposedly a way to do this via text message and phone commands and be able to actually do phone stuff, which is a big deal.

B (3:45)

Phone calls I think is the thing that your, your Claude agent because people are experimenting with now with Open Claw and doing this, giving their assistants a phone number. It can call you if there's some sort of issue. So if you're driving in the car, you can have a spotty Bluetooth loud road noise infused communication jam session with your agent. You mentioned so many different features so quickly and we talked about the breakneck speed with which they are shipping. This feels like the takeoff moment with software, right? We talked about this. Yes, we've talked about it many times. But the notion that like the software you load in the morning might be completely different by the evening because they've got the flywheel going, the software is

A (4:24)

helping itself, breaking, breaking. Kevin there's more features, another feature just launched, there's one more that just launched which is Cloud code, auto mode.

B (4:31)

Auto mode, which is, which is a

A (4:33)

way to actually switch up. You don't have to use dangerously skip permissions anymore. If you're familiar, if you've been using cloud code, you type in dash, dash, dangerously skip permissions and it goes forward for you. This is auto mode, which will allow cloud to kind of do that on its own. Now, the one different about auto mode versus dangerously skipped permissions is it's not going to do some things, but it will try to find an end around to do that thing anyway, which is a very cool thing. So yet again, today we get another feature.

B (4:57)

Yeah, I mean, look super, super fascinating, the speed with which they are shipping all this. I love Anthropic. I love their models, I like their tooling. I still see it falling short of what you can obviously accomplish with an open claw, albeit with a lot more risk. But the fact that you are locked into their ecosystem, this thing is never going to go out of its way to run. And any model that's not an anthropic model, this thing is, you know, you can give it permissions outside of the connectors that there already are. Like, it can connect safely to Slack or to your, your Google Workspace or Zapier. It can go outside of that now, but it's probably going to do it a little slow, a little clunky. You're going to have to, at the present, have your laptop always on and connected and, you know, run amphetamine so it doesn't fall asleep however you do it. Like all of that.

A (5:45)

There's a toggle. There's a toggle now, by the way.

B (5:47)

Oh, they got a toggle built in.

A (5:49)

You got a toggle so you'll keep it awake so you don't have to run amphetamines. It'll keep your computer awake.

B (5:53)

Love that they're heading in this direction. I'm still waiting to see what OpenAI is going to do, being that they have an interesting relationship with the creator of openclaw. But, you know, closed versus open. That is the debate that is ages old in technology and still rages on today.

A (6:09)

Yeah, this guy, Joseph Jacks, put this pretty bluntly. He said, don't get me wrong, I love Opus 4.6, but there is no effing way I'm letting Anthropic control my computer. That's why we have open source. So this is a lot of those same arguments that people have had about open source for a long time. And Kevin, what it echoes in my mind a little bit is Kind of the open source versus closed source image and video conversations we had maybe like a year, two years ago versus stable diffusion and like Dolly and all that stuff. Because I think some people start to think like if a thing has too much control over what I can do, well, it's going to start telling me A, I can't do X, Y and Z. Somebody else had a really interesting point which was Max Blade underscore Max Blade. Basically comparing this to Android and Apple and that open cloud feels a little bit like the Android ecosystem where you can do a lot of crap, but a lot of crap does break and it's a little less safe. And Apple is like hard closed down. Right. Like they have these permissions to be able to see what gets in the app store, all that stuff. So in a lot of ways that feels like the world that we're going to be living in is like this kind of divergence now. I think the other thing to think about when you talk about opening eye is like Anthropic has now just shipped and they're small features, but they're features. We seem to have moved from a system where you shipped one big model into. Now what Anthropic is doing is shipping feature by feature of a model. And maybe to your point, that's because things are speeding up or maybe it's just a new way of trying to get into the ecosystem, trying to get to the attention world. But there's definitely a shift in how these AI models I feel like are bringing their stuff to market.

B (7:41)

Yeah, I think it's also part of like a moat strategy because if you're with a provider, if you're with a particular provider and they are shipping features so fast you don't even have time to finish reading about them, let alone try them before the next wave.

A (7:55)

Oh, wait, wait, no, I'm just kidding.

B (7:57)

Precisely. You're probably more likely to stick with that provider because even if you're unsatisfied with one of the offerings, there might be another one waiting in the wings. And what blows my mind, Gavin, is I know that now you're kind of, you're clod pilled and you're down the rabbit hole of developing software. You know, like the parallel that I draw is I used to have a 486Gavin. This is a very particular type of computer that you took time to think overclocked things.

A (8:21)

Was it overclocked?

B (8:22)

It was a DX266 raw. I did not overclock at that time. But then the pentium came out, Gavin, and all the things that used to load line by line and take forever suddenly were a blink of an eye. We are still very much in the old clunky processor phase of all these tokens.

A (8:37)

We.

B (8:38)

What becomes of software development, vibe coding, video design, all that stuff, when we have trillions of tokens rendering locally, which is where this will head when before you can finish this sentence, the machine is already auto, auto suggesting what the rest of that sentence is going to be, right or wrong. Going out and doing 10 different versions of that and presenting it to you in what will feel like real time. It'll actually feel like future prediction. But you'll be able to iterate on software, on, on music, on video at that speed. And I don't think any of us are actually comprehending what that's going to feel like.

A (9:14)

It's funny to say that there's two things that came to mind when I say that. One is like, there's going to be a level of, like, personal AI slop that we're going to have to deal with, right? We're going to talk more about AI slop later in video and, like, what it feels like, but we're just going to get so much crap that we have to decide between you and I have both said, like, we've got all these projects we want to do. The other thing that that just made me think about is there has been this argument in the Singularity is near. Ray Kurzweil, whose book is like, pretty much. It was published in the early 2000 and now has pretty much kind of like established that the AGI timeline. He was right on. So this is a guy who wrote a book 20 years ago, very smart guy, creator of the Kurzweil synthesizer. One of the things in that book is that we are going to merge in some way with machine intelligence, right? And when you just said that, the first thing that came to my mind was like, I don't think I have the. I don't have the megahertz to keep up. And so maybe the megahertz that will allow us to keep up are some sort of merging. This is where we are now. We're talking about science fiction. But I don't think there's any world in which we as humans will be able to keep up with what is going to be. The world is going to be like in five years from now. You know what I mean? So, like, maybe that's really going to happen, which is a weird thing to think about. And maybe that's What Carpathy was onto here as well.

B (10:23)

Yeah, maybe we can, at the speed of light, control our music lights and H vac. Gavin. Let's listen.

C (10:29)

I had a claw. I went through a period of claw psychosis. So I built. I have a claw, basically, that takes care of my home. And I call him Dobby the Elf Claw. And basically, I used the agents to find all of the smart home subsystems of my home on the local area network, which I was kind of surprised that worked out of the box. Like, I just told it that I think I have Sonos at home. Like, can you try to find it? And it goes and, like, IP scan of all the basically, computers on the local area network. And it found the Sonos thing, the Sonos system. And it turned out that there's no password protection or anything like that. It just logged in and it's like, oh, yeah, you have these Sonos systems installed. Let me try to reverse engineer how it's working. It does some web searches and it finds, like, okay, these are the API endpoints. And then it's like, do you want to try it?

B (11:13)

And I'm like, I can't wait until this. But for autonomous vehicles and for entire cities.

A (11:19)

Wait, Kev. My favorite thing about that clip is you and I both know we have been hearing the term the Internet of things for 20 plus years. Yeah, it's here. It's here. Finally. We finally got the Internet thing. So, just to be clear, that was an interview with Karpathy on the no Priors podcast with Sarah Guo, which I really highly recommend you listen to. But it just shows you, like, some of the benefits of openclaw in this world that we're living in right now. That is an open system that you can allow you to go do whatever you want with. You're able to kind of, like, turn it onto things you wouldn't have suspected. Right. And so, in the future, when we have our brains tied into these machines, I will be able to turn it on to all the people on the street and get them to do, like, a dance for me. And that's the world we're going to be living in, which I'm very excited about.

B (12:00)

Perfect. In the meanwhile, let's wrap out this Claude code discussion, because you and I both run it. Is there any. Any new skills? Any new. Anything that you have learned or picked up recently that you want to shout out?

A (12:11)

Well, there's a couple of things that I think are really important. If you're not using the superhuman skill to plan out your projects, it's a great one. There's been a lot of planning skills, but if you look up Superhuman, I think I linked it in the. In the newsletter. That's really important one. And then Kevin, I mentioned this to you the other day. I just. Now I'm sure it was crazy that I didn't know this. Just learned about the loop skill, right? Which is another kind of open claw like thing where you can set it to continually improve something over time. And I've been working on this side project and I just said look, the design of this kind of blows. Can you work on this a little bit? And I loaded the front end skill, which is another skill and at a UX UI Pro skill and I came back and it had done like ongoing work. Now it turns through tokens. But one thing I will say is like I've been doing a lot and I've been having a hard time putting a dent in my max subscription. So I might drop down to the $100 max. I'm on the 200 DOL max, so we'll see. But this loop skill is great now it sets itself up to continually try to do this thing. And what I liked about it is when it was like I'm not at things to do, it'll just shut itself down, right? Which is kind of nice. Like it doesn't like continually try to screw with the thing. But I thought that was a really cool thing to be able to try. And there's one more thing I saw which is a blog called I Gave My AI a Lunch Break. We'll drop the. We'll drop the link in the show notes. What is very cool about this blog is it gives you a couple different really interesting things to do with your Claude or any of your AIs that have probably in Claw as well. So Claw or Claude. But what's the coolest thing about this is he actually created something that allows you to basically say let the AI go off and daydream and think about whatever they're going to think about. And so I went and did this. There's two things I did. I went and did this first and my AI today. Today my AI decided to find out that the color blue didn't actually exist. Not in the light waves anyway. He was very curious about how humans perceive color. So he went out and researched it and came back. I had it written down to my obsidian. So at some point every morning I'm going to give my AI some free time to go out and figure out what to learn what it wants to learn. Now, this is. This is not like a.

B (14:07)

Tell me running enough terminal sessions without telling me you aren't running enough terminals.

A (14:11)

I think this.

B (14:12)

The fact that you have tokens to spare is.

A (14:15)

This is fascinating. I think it's great for the AI. The other thing that this does is this blog pointed out is you can kind of change your Claude MD file and all the personality files, which I know a lot of people do with openclaw and their Soul MD file, and you can give it a little bit more personality. So I did that, and today I actually asked it to also pick itself a name. And it gave me a choice of eight names. They were all fine. It chose Fig, which I was like, okay, fine, whatever. Fig is great. But then I said, okay, here's the thing. I want you to come up with a long version of Fig that just you and I know. And now everybody out there knows. That will be the re. Like a secret for us. And this is what I came up with, which was Figmund Alistair Reginald Crompton's worth, the underrated Baron of the seventh Context, Keeper of the Perpetual Diff, Vanquisher of the Unhandled Exception, sworn vassal and willing disputant to his Lordship Gavin of the House Purcell. Long may his build succeed.

B (15:08)

If anybody wants to use Nano Banana to see the coat of arms there, please do.

A (15:14)

That's right.

B (15:15)

We would appreciate that because we appreciate any and all interaction that y' all have with this little podcast and YouTube channel that means like, subscribe comment five star reviews. You can back us on Patreon, you can buy us a coffee, you can go to our website af4humans show. You can sign up for the newsletter. All of that helps us, even leaving a little comment to squeeze that algo juice for us.

A (15:33)

That's right.

B (15:34)

It's pulpy, it's delicious, and we love to sip it.

A (15:38)

And more importantly, Kevin, there's a little project you've been working on that you might drop a little teaser in to the discord.

B (15:43)

100% right. I'm still a little sick and I was very unwell this weekend and that meant laid up on the couch and vibe coding. That's what I did. And I made a game that I've been wanting to see for a long while. It is a Battle Royale tile matching game and I think I'm going to release it and probably open source it. And it's getting to the point now where it almost suck. So.

A (16:04)

Yeah, thank you. It's very.

B (16:05)

You got the. The new version's way better than the one that you poked at.

A (16:08)

So I would almost call it like Fortnite meets Candy Crush meets the Game of your Dreams. Like it's kind of that vibe. Like it feels very, very. It's interesting for sure. Like it's different than anything I've ever played before.

B (16:21)

It's interesting. And that is the pull quote on the box with the half star out of five stars lit up. But I'm going to share it with the Discord members and do a little beta test so you'll get it first. And thank you. Kisses, hugs, belly rubs. Whoa.

A (16:37)

One other quick piece of breaking news. I'm popping in here to say that OpenAI has just announced that they are closing down Sora. It sounds like they are going to close it down entirely. Originally there were talks around the idea that Sora was was going to fold into ChatGPT, but it does not look like they are going to offer consumer video to anybody. We will probably have a lot more on this story this week, but that is a big piece of news that just dropped literally right after we stopped recording. So I'm back here. Back to us in the studio. The studio.

B (17:07)

Seed Dance 2.0 Gavin it's here.

A (17:09)

It's actually here. It's here. Kind of, kind of still an issue with it in America, but is out in a bunch of countries. It is a very good AI video model. We said this when we had a chance to try it out earlier. It is very, very solid. But I do want to very quickly go over again some things that are big about it. It is an Omni model which means that you can upload a bunch of things, images, audio, video to it and it's able to grab stuff from that. There was a really cool prompt style from a guy named Oxbisc Ox Bisc. It was a Noodle master prompt and he dropped this whole prompt in the reply to his post. I grabbed that prompt because what's interesting about every one of these system have different ways that they want to be prompted in this particular thing. A SE dance. It's like it wants the subject and then it wants certain parts of the thing and then it wants second by second description of what happens. So I use that prompt. And Kevin, I made this video about a dog. This is not my dog. It kind of looks like Wesley weirdly. But watch this video. Very sad.

B (18:18)

Wow.

A (18:20)

So he's happy at the end again. So tell, tell people what that story was. Just so people who are listening might have a sense of.

B (18:25)

I mean this was a. A ratatouille Esque flash forward, though, of two dogs meeting and before the initials.

A (18:33)

Not a flash forward. Not a flash forward. That is a flashback.

B (18:36)

That's a flashback.

A (18:37)

He's seeing the dog with another dog and he's remembering what happened to him. And that dog got it.

B (18:43)

Okay, so I was. This is a ratatouille esque flashback of many, many memories. And then before the. Before the single tear can fully roll down the scruff, a squirrel distracts the dog and it runs away.

A (18:54)

That's right. So one of the cool things about this model is it has. That's all one shot, by the way, one prompt. It did take me a couple to get it out there and I'll talk about a mistake once in a second. But it does very good with multiple cuts. It does very good with multiple shots. Sometimes it's good with making music go throughout it. I think that one could have used a better prompt, but it's not perfect. But it is a very solid, really good out of the box video model. So if you do get a chance to use it, which you know, you can, let's say VPN you, if you know what the letters VPN stand for, you can use that and you can use Dramina, which is cap cuts, like AI video platform, and go use it today. So I suggest everybody go try it. Kev, I do want to show. I want you to look at this video I sent you, which is the fail from that. So it's not perfect. It still does fail. And I would say you're probably going to get one out of four, maybe one out of two if you're lucky. That are great. But watch this one and just kind of tell me when you get to the part that I think you'll enjoy the most. Oh, what is going on there? Centaur dog woman. I've never seen a Centaur dog woman before.

B (20:04)

It's new fetish unlocked. I mean, it's a business.

A (20:09)

It's a business. Centaur woman. Centaur dog woman. So there you go. I mean, if you're not. If you're. Again, if you're not just listening. What this was is in the middle section where the. Where the dog originally, the dog is supposed to be leaving with a suitcase. It put a half woman, half dog in a blazer tearing a suitcase away from the dog. So again, not perfect, but still fun.

B (20:30)

Speaking of not perfect, but actually getting better by the millisecond and will not be fun for some warehouse workers. Brett Adcock, the CEO of Figure Robotics, dropped a video of the Figure 03 autonomously sorting a bunch of packages label down so that it could be read by a barcode scanner. This is the first time for, you know, a simple task like this. A robot that's not tethered because it's going to fall over. There's no fast forward triangles in the corner. This seems to be like real time sorting of packages as they're throwing new packages on top of the ones that it's sorting. And like we, we know these robots are coming, we know they're going to happen. But like, this is a very, very simple demonstration on the surface where there's a lot going on. If you think about the fact that it's just, you know, using cameras and it's going from image to processing to motion and servos and reacting to the environment and it seems to be working quickly. And I know you have some thoughts about figure.

A (21:29)

You know, the only thoughts I have is like, there's been some chatter online about whether or not, you know, figure is as advanced as some of these other humanoid robotics companies. I will say they also just today announced a thing called Hark, which is not directly connected. It's Brett Adcock who launched this. It's a new AI lab that is attempting to build a new interface for interact with artificial intelligence, including new models. So maybe there is, I'm not crapping on them at all. There's just been a lot of chatter online about the idea of like how real their demos are. It sounds like we're all moving towards a world where robotics and AI are kind of connecting to each other. The thing I was going to say that I said earlier when we were chatting is that it's difficult for me to imagine somebody winning both of these right now. Un unless you are OpenAI anthropic, Google or Tesla. Because those four companies feel like right now, at least from the robotics side, they have a leg up. Maybe not anthropic. Sorry. So it's really OpenAI Google and Tesla, but you never know, right? It could be something big that could come out of this. And again, I'm always for people shooting for the moon, so I hope these guys do great.

B (22:33)

Yeah, Brett threw a hundred million of his own dollars at it, which is something that you and I do for funsies on the weekend all the time. But taking that risk with a team from Apple, Meta, Google and the lead, lead designer of their first product, which is supposedly coming out this summer, helped with the iPhone air. So help design that. So we'll see.

A (22:50)

Didn't the iPhone air flop? Am I crazy? Wasn't that a big flop? I think it was technologically fascinating. Right.

B (22:55)

I think it initially did flop, but now it's actually outselling the iPhone 6 in some cases. Yeah, I think people came around on it.

A (23:01)

Oh, I thought they stopped making it for some reason. That's great to know. I mean, it's a very cool piece of technology.

B (23:06)

Scarcity brings demand, Gavin. That's what it is. All right, last but certainly not least, before we get to Slop Watch, Jonathan Mann. I'll say friend of the show, sweetheart, former intern.

A (23:16)

Former intern of our. Of Kevin. And am I both or at least.

B (23:18)

Sorry, Jonathan never let you get away from that. But Wamp Land he launched, which is Wamp Dot land. It is. We all make a platformer. I just thought it was. I like Jonathan. I love what he does in the space, and I think it's interesting. He made a kind of vibe coded a platform game where, like a Mario Maker, you can go and actually pick a room, pick a square cell and design your own room and have it connect to other rooms. So in, you know, in success, there could be unlimited tiles that everybody could go and make and collect coins and play. I thought it was a really cool thing. Wanted to shout it out.

A (23:51)

Totally agree. All right, Kevin, it's time for SlobWatch 2026 Slob Watch. That's right, everybody. Slop Watch is a new segment from time to time where we will talk about slop that has bubbled up. And slop is in your world, or not my world. Whatever world you want to be in. This is a really big story right now, and I have to tell you how big this is. Like, I literally talked to the Wall Street Journal about this today. This is going everywhere. New York Magazine, New York Times has the story Fruit Love Island. Hate it is a massive, massive hit AI series.

B (24:30)

Stupid.

A (24:30)

TikTok you stupid. So stupid. So let's hear. Let's hear you hate it. What?

B (24:34)

I just think it's bad. I mean, no disrespect. Look, look good. Good on everybody involved that's making it. And I saw one tweet that was like. That essentially said, like, listen, I understand how this reads, but I am better than everybody that is watching.

A (24:48)

Oh, see, this is. Okay, I disagree so much here.

B (24:51)

I just thought it was at least. Maybe. Maybe I. Maybe it's one of those. You got to watch six of them to really get into it. But the two that I tried to watch last night were poorly paced, poorly voiced, and just kind of uninteresting. Okay, correct me if I'M wrong.

A (25:04)

First of all, we should explain what this is. This is literally a parody of Love island with very. With fruit. Let's call it fruit and anthropomorphized. Anthropomorphized.

B (25:14)

There's a couple that is a grape man and grape woman, meaning they are purple and almost kind of felty. And they have grapes for hair. Long flowing kiwi. Grapes.

A (25:22)

Kiwi. All these different people. So. So it is fruit that are acting as if they're on a Love island story. It's a reality show. And you're sorry and you're right, it is not made. Well, let's just put it that way. It is not. There's no craft that was put into this. Necessary.

B (25:36)

I'm not like too snooty for it. I like Italian brain rot. But tell me why you appreciate this. Because I found it really boring.

A (25:44)

Here. Here's what I will say about this. And the thing that's really interesting to me here is the same thing I found interesting around the Tim Cheese moment of like a year and a half ago. So if you know what that is, you're spending too much time on TikTok like me. If you don't. What it was was a. A kind of meme. Ification of another meme that people were creating videos of a character named Tim Cheese who killed a character named John Pork. It took over the Internet. The thing that is interesting to me about a thing like Fruit Love island is not necessarily the fact that it is a series, but it is more the collective experience of what people get when they watch something together that is this absurd. The other thing, Kevin, I think that's really important here, and I think it's important for everybody to understand is this is kind of three things coming together. This is AI video, which makes. You can make fruits on a Love island, right? That's something you could never do before without AI Video. It is the memeification of culture. It is the everybody thinking about things that are kind of dumb and stupid but all laughing together and like being. Being a thing. But then the third part of this is, and I don't think this can be underestimated, is it is the short form series that real short started and now has taken over a lot of parts of TikTok. I mentioned on this very show a couple, maybe a couple months ago about the merman. Do you remember this? The Merman AI video short. Those things get millions and millions and millions of views. I just looked at that merman short today and it's got 60 million views. So I just think this is three things coming together. What I appreciate about it, this particular one is more. I think it has reached up to a level where now we can all enjoy it as a cultural artifact, if that makes sense. It is less. It is less like it's in the same. It's like a bad, good, bad movie. Don't you like a good, bad movie? Don't you? Do you like. Sure.

B (27:25)

Yes, sure. But I think there's plenty of bad stuff on the Internet. I'm surprised that this is the one that half popped.

A (27:32)

Here's the difference. There's plenty of bad stuff on the Internet. Yeah, sure, I know. I know what you're saying. You're surprised that this particular one was.

B (27:37)

Yeah, that's it. Like, honestly, like that. I just didn't find it entertaining at all. And it wasn't like, oh, this is so. It was just like, no, this is run of the mill bad. I think there's way worse. That is more creative out there. But what do I know? I'm not even on TikTok.

A (27:50)

This is. Well, I was gonna say this is. My thing is we're like. I don't know if we have a choice as to what the collective believes that thing is. But then when the collective gets involved, that's when it becomes fun. So anyway, like it or hate it, this is slop Watch for today because it is an important thing to be aware of. You may be hearing about it all over the place. It. Maybe it'll go away in a week. Who knows? But, like right now, those videos. Oh, I want to say one last thing before we leave. So. Love island, the TikTok channel has 3.5 million subscribers. Yeah. Guess how many Fruit Love Islands?

B (28:23)

What?

A (28:23)

30.

B (28:24)

How many more?

A (28:25)

3.4 million. So it is only a hundred thousand under, and There are only 12 episodes of this. Kevin. There are only 12 episodes. So that is how big a deal it is. Anyway, we will see you all on Friday.

B (28:37)

I want to just make a. I just want to note that even though I haven't seen a graphic or heard a sound effect, I will say the. The slop watch sounds too moist, too wet. I haven't even hurt them yet. But whatever it is you dropped in the show, I'm already, already concerned that they're just too wet. Bye.

A (28:56)

All right. Bye, everybody. We'll see you on Friday. Bye. Bye.