240 - OpenAI Just Changed Everything! Sora 2 Breakdown with Elizabeth Knopf

0:00Wait, Claude can do it now? Claude can do it now. Oh my gosh, this is insane. It's you standing in front of a whiteboard and you're talking about the importance of automation, but this is not you. Everyone's kind of freaking out about it. If this nails this, this would be killing part of the job I used to have to do.

0:18Okay, Liz, I've been giddy as a school boy. I couldn't wait to talk to you. And so I said, let's talk early this week. And so we're talking this week because some massive changes have happened. OpenAI has just kind of flooded the zone. They've released Sora 2, which is this new image video generating software. It's insane. They've also released ChatGPT 5, and they've released this thing called Pulse. So the first thing that I want to do, the first thing I want to do is talk about Sora 2, because I do think the technology is insane, but the way that they've rolled it out is even more insane. And so I'm going to start by sharing my screen.

0:54Are you ready? Yeah, I'm ready. I'm watching this video, and this is Liz. she's in front of a whiteboard and she's talking about how to automate yourself. The crazy thing is, Liz, is this you? Did you film this? I guess me, the digital clone of me. This is insane. It's you standing in front of a whiteboard and you're talking about the importance of automation, but this is not you. Right. Everyone's kind of freaking out about it. It's invite only, which we'll talk about in a second, but let me just show you kind of the way that it works. You come in here, you go to Sora, and what you do is you create a cameo.

1:28So if you see this, this is me. This was my cameo. This is what I recorded. And I recorded some numbers and I like messed around with my facial expressions. And then from that, I was able to generate some of these videos. Like this is me at a sporting event saying how excited I am. And then cheering like that looks like me off like five seconds of video. Yeah. It's a little heavier than I actually am. But look at that. I'm at a soccer game and I'm cheering because they just scored a goal. I have another one. This is my favorite one. I'm, I'm looking into the camera and apparently I'm saying, is Sora going to kill TikTok and content creators?

2:03I even have one with my best friend, Chris. We're sitting here talking with each other. The prompt is Chris and Nick are hanging out with each other, but subtly trying to one up each other and said, our whole conversation is us trying to one up each other. And then the cool part is if you make your cameo, like your profile public, other people can make remixes of your cameo. So for example, Liz has generated a bunch of cameos of her and I podcasting together. Look at this. We've never actually, we've never been in the same room, but this is us. I thought we were doing that, uh, that matrix scene last week.

2:39You forgot about that clearly. Yeah. I forgot about us filming it. This is like, it's insane. It's insane. I don't, I don't know what else to say about it. Liz, the video quality is incredible. The speed with which you can actually generate something is nuts. I'm kind of speechless. What do you think? Yeah. I mean, I think you sort of hit on a few things that people always underappreciate. So number one, just you have to basically take that video for like less than 30 seconds. It's not even a cameo is like not even 30 seconds of a setup process. And it's like, you've done this previously with HeyGen, which is another tool that is the AI clones.

3:18They were sort of the first ones to really build that out. But the fact that from a consumer angle, the adoption is going to be so much greater because you don't have to spend much time in the setup power class. Well, the thing with Hey Jen, I don't know if you messed with Hey Jen. I think that you had. Yeah. I uploaded like hours of my podcast video. And so it took hours of my podcast video and it, it's got a pretty good avatar, but it's still a little bit off. This, this is not off. And I recorded for five seconds. Yes. Now, if you're going to go and create a cameo, you know, pro tip, it actually says, if you click on the tips to speak for the full five seconds, when you do it, this window pops up and it says, read these numbers.

4:02And so some people just read, I don't, did you read just the numbers or did you say something more? I did the numbers. My audio got screwed up initially. So I had to go back and redo it. So, so don't just read the numbers, like actually say something else, like have something else scripted to say. And then on top of that, make facial expressions. So like smiles, angry, surprise, sad. So I did that in this like five second window. And I feel like it did a really good job of capturing, at least in one of the videos, what I look like and what I sound like in five seconds. Hey Jen takes hours. This was like a five second, I don't even know what exercise for me to actually get a really good clone of myself.

4:41So that part was crazy. What do you think? What do you think of the rollout? I mean, again, I think they have just nailed product every step of the way. And again, people underappreciate this, but it is one of the basic, the things that make or break a company. The adoption of it is, I think it'd be dramatic. I was looking for it in the app store. It isn't necessarily like the top yet. And surprisingly, Gemini is number one, at least on my app store list. But what's interesting that they did is they pulled out some old tricks that always work, which is they put a velvet rope and had invitation codes.

5:20And because of the viral nature of the content, they had people putting it out. And of course, they used Sam Altman and some of their team basically rolled out some clips and said, if you want an invite code, comment. And then, of course, that then makes their posts viral. And then people get five invite codes to then share with their friends, so on and so forth. So like, they just nailed the rollout of this app. Because again, it's not just like adding it on as a feature within chat GPT. I mean, you can go there, but really what they're going for are downloads. Sam Altman is all over this thing.

5:53Look at that video of Sam here. We've got a video of Sam here. We've got just video after video of him. Apparently he's on stage. Wow. He It looks like he's a fitness instructor. Huge in going to this whole concept of like founder marketing and, you know, personal branding and stuff is because he used himself as a person in the content that then people can remix and make. And so that viral element around it just, again, is just making this explode. Yeah. And that, well, that's, what's crazy about it is you can't post videos on this from like your camera roll. Right. The videos that you're looking at on here are AI generated.

6:33And so you take whatever, a few minutes to create your cameo, your bio, whatever you want to call it, your avatar. And then you can prompt the system to actually make a video. So I could come here and I could say, all right, I want to make a video with me. I want to make a video with Liz. And I want us to be podcasting at the beach with Sam Altman. All right. And that is what is now going to create the video. I didn't upload a video of myself and have it augmented, which I'm sure that will come at some point. But all the videos that you see here are all AI generated. So it's a totally new form of social media consumption.

7:13You have Instagram, which is like picture. Then you have maybe Facebook, which was written at first, but it's written and picture. You've got Twitter, which is written. You got YouTube, which is video first. This is content that is all AI generated. It's not actually video of you. And that to me, that's very different. And so that in and of itself is a genius way to roll out this new technology by supercharging it with the social and organic element of it. Like you said, the invite only. And I can't believe Google missed on this. Yeah. Well, and I actually think the ones they're trying to compete with here is even in a way meta.

7:50I don't know if you remember a few months back, meta had sort of pulled out this, like you could do pictures and remix images and some like clips and things. And they were trying to have it as a feed with a social element to it. And these guys are doing that, which again, Gemini and Google haven't done at all. But Meta is only the one that sort of like played footsie with this concept of AI and having a social layer on top of it. And so again, no one's nailed this. And who knows if this will actually be a thing from a social angle or if people will just again, just take this stuff and put it on the existing social media platforms.

8:27platforms, but I know like at some point we're going to see an AI social media ad. So I don't know. Here's why I think it's genius. Like number one, you introduce an element of scarcity. Anytime you introduce an element of scarcity in marketing, like there's just this, oh, what's going on? We saw this with Google. That's how Google got started. I don't usually remember in 2003, four or five, like you had to get an invite from somebody in order to get a Gmail account. And it was like the cool thing. Facebook was the same way. You had to be on a college campus in order to get access to Facebook.

8:58You had to be a college student. You had to get an invite to it. So they've now introduced this element of scarcity, interesting in and of itself. But then you add the social element onto it. For the most part, many of these tools that come out, it's like, oh, there's agent mode. Oh, Claude now has Claude code. The vast majority of users might go on and use it for five seconds and think like, oh, that was pretty cool or neat. Now what they've done with this though is they've introduced novelty and social interaction which allows them to hopefully keep customers sticky like they're going to be using it a lot more i don't know if anyone's ever done something like this before where they they like introduce a new product but it's almost like the social platform is secondary to it i've seen where the social platform is primary and then maybe there's a new product that's secondary but the social platform being secondary is new to me.

9:46I've never seen it. What about you? I think there are some apps. I'm trying to think what comes to mind. There's some that have like viral loops that create a social dynamic. But again, we're in sort of like just new territory because AI and the ability to create content so quickly is a new thing. So, I mean, there are again, these sort of social viral loops that exist in product launches and app launches where you just add in a social component, whether it's like, you know, sharing your stats for fitness or whatever. But again, nothing where we've seen this in the AI era. So. The last thing to like kind of close the loop on this is what's the use case for this?

10:29Like, like this image generation, this video generation, why would anyone use this? If you're a business owner? Yeah, we've, I mean, we've sort of touched on this, I think, every time with Nano Banana and some of the other ones. And every time there's a school new thing. At some point, it's going to be, again, ads, influencers, social media marketing, or just content creation for videos, for sharing like, here's my product, here are my features, benefits, doing sort of informational things. You could make your training internally more human, or I shouldn't say human-like, but sort of, because you probably know SOP or something.

11:07here's the point that i'm making is like we have to do sort of mental gymnastics to think of any other use case beyond marketing yeah in my opinion what what sam just did with this is he said i'm not worried about your use case of marketing i'm now turning the technology into a social platform like that inversion is a paradigm shift because otherwise you're like oh cool it's a cool technology 0.1 percent of people are actually going to use it now it's like oh it's like tiktok on my phone and I can go and just see what other people have created. You are now consuming a lot of this content that's being created.

11:43I think that kind of unlocks the capability of Sora's applications more than how can we use this in a marketing capacity? It's entertainment is what it is. And to me, that's like the genius of this launch and rollout. And yeah, I actually think that's a really interesting point because I'm curious from all the other launches for like image generation, I think that might be probably good adoption, but I am curious the adoption curves and like at the spike. And is there a major trough from a VO from Google, like, because it still has friction in the creation process. So. Well, yeah. And it's like VO with Google, if I wouldn't created something cool now, like I had, it's, it's novel.

12:23It's interesting. I might send it to a friend or two of mine, but then what do I do with it? It like lives on my photo reel on my phone. Like what am I actually doing with it beyond just, oh, this is cool. This kind of unlocks the usage of that. Other people can use my likeness to create things. And all of a sudden, a friend of mine creates a, whatever we're calling these reels. And it's like, oh, that's funny. That's interesting. Oh, I'm going to share that with this person. He's crowdsourcing usefulness. I don't even know how else to say it. It's just, it's crazy to me. One other thing I want to highlight, because people who haven't maybe used the product yet are probably thinking, oh my God, someone's going to use my likeness.

12:59They solved for that by enabling who you can share it with. So it could be your closest friends, just nobody and for your own creation purposes, or you can share it with the world and we'll see what happens. Yeah, that's a really good point. I should have shown kind of what that looks like in the settings, but you can actually choose who has access to your stuff and you can make it totally public. So anybody can use it or you can make it totally private. So only you can use it. if you upload an image and they, and it, they identify a person in it, they will not allow you to upload it. So I do think some of these platforms are taking more and more steps to make sure that they're, they're protecting user privacy, but man, yeah, this part was really cool.

13:38All right. So that's Sora go play with Sora. It's really cool. I wonder if this will now become a, another social media platform. I personally think that it will, I don't know that it overtakes any of the other platforms, but I think it's just going to become a new mode of creating art in much the same way that Pixar grew out of computer-generated images. I'm sure there will be something else that kind of grows out of AI-generated images. But okay, so we had some other updates. Claude released an update to Sonnet 4.1. It is now Sonnet 4.5. I don't know what the differences? What's going on? Okay.

14:20So actually, let's take a step back. Before they rolled out 4.5, there were some really cool updates that I think people aren't talking about as much and are pretty awesome. Number one, they came out with a Chrome extension. Like other things we've seen earlier in the year, we had OpenAI, had Operator, which went and basically worked in the browser. We then had more recently Comet by Perplexity roll out a whole browser. And then a company called the Browser Company, which got acquired, came out with a browser called Dia. And so we started seeing these browser wars take place. And then even Google with Gemini started sort of integrating more sort of sidebar usage that integrated with Chrome.

15:02but now Claude has enabled the ability to work within your browser and do tasks on your behalf within the browser. How do I go and get it? Okay, so let me share my screen and I will show you. Funnily enough, I've been going away from ChatGPT. It's been really hard for me to go away from ChatGPT because it has so much history on me and I do feel like it's persistent knowledge is good. It actually does crawl all of my conversations, but I really use Claude or I use Grok for the most part for everything that I do. I mean, I'm with you. Hey, I don't know if you remember this, but when we started this podcast, we entered into a social contract.

15:42I would spend time, energy, and money producing this podcast, interviewing these individuals and giving you insights into how to build, buy, start, grow your business. And you would like, subscribe, and leave me five-star review. Now, out of that, we both get to talk to really cool people and hear really cool insights. We both get a ton of value. But I just want to help you keep your word. So would you do me a favor? Will you go leave a five-star review for me on Apple or Spotify? It would really help. And if you want, even share this with a friend. You know, I was joking with you. I didn't get an invite right away for this part.

16:18I use Claude more and I talk about Claude more on the interwebs. so okay so you do have to actually go to claude.ai slash chrome yep okay okay then meet claude for chrome um and then add to chrome i already have it so it's basically already in here wait wait wait wait wait wait how do you already so i'm on there yeah it says meet claude for chrome i do that and it says join the research preview join the waitlist so i may have so i have to get on a waitlist. Dang it. You're too cool. Yeah. Okay. That's awesome. So you have it. You have the beta. I have the beta. You are so cool. Gosh dang it. I didn't get it for chat.

17:01No, no, no, no. Keep sharing your screen. I want to see what it is. Okay. So, okay. So I am in right now, YouTube and I have this little clod icon up in my bookmarks bar or my, sorry, extension bar. And once I clicked it, it pulled up this little sidebar here. So now I'm going to say, what is this YouTube video about? No way. And so the methodology too that it does for analysis of your browser page is it takes screenshots. So it's not actually scraping the data. It's taking screenshots of the whole page and then doing an analysis. So here it basically gives a summary, explains what it is, video details.

17:49I think it just pulled from stuff I have it in there too. Okay. So let's actually do now something with this. So let's go to economics. Okay. So for, if you're listening, Liz has a webpage pulled up and then on the sidebar, there's the cloud extension within Chrome and she's asking a questions based on what she's got on, on her screen. What's interesting is I do this all the time, but I take the screenshot and then I'll copy and paste it into Claude or ChatGPT and say, hey, I was trying to do this thing. This is the output that happened. What am I doing wrong? So it's not, I'd like to have more real-time ability to query it and not have to switch between windows.

18:33And effectively, that's what you're saying it does. It's because it's working off of the screenshots and it's embedded now in the Chrome browser. it can real time kind of give you the feedback that you otherwise would have to copy and paste for. Okay. So what we're going to do right now, just as a quick demo of this is I'm going to ask it to provide a summary of this video and then to input it into this little Google sheet I just created quickly and to put in the data here. So this is simple, obvious use case, but I just want you to see how it works. So summarize this video with key insights and then add the details of this video to this Google sheet.

19:18And just put the link in, I put in the link and just a quick tip, make sure that it's you can edit it as editable. Yep. Otherwise, it'll get stuck. And then make sure to include the title of the video. What else did I put here? We'll do the title, channel, the channel name, the length under time, and then summary in the summary column. All right. So let's see how it does. So if it's not actually watching the video, what information is it pulling? It's going to pull the transcript from the page because it can still pull the transcript from YouTube video. So it's going to be able to do that. And so if I wanted to like research, I could use this, I could go find 10 YouTube videos and like run them all simultaneously to populate into a consolidated spreadsheet.

20:18So it's actually watching the video right now. So I don't know if you could hear it a second ago, but it's watching a video. So I don't know if it's listening or not because I had to mute it so we could talk if it doesn't. And so it's taking screenshots. And I think what it's also going to do is grab the transcript somewhere. So before when I've done this, it grabbed the transcript. But yeah, just as you were saying, like you can use this for research when you're not around. I would say you still have to interact with it and it's not going to just do the work completely unsupervised, but it does a pretty good job to do a full task and workflow.

20:54It's kind of like operator. Exactly. It just works a little bit better. Again, none of these are perfect yet, but I would say like any demos you saw with operator, there was just, again, a lot of friction. I just feel like some of that friction again has been removed or reduced here. Well, because it's like embedded in the browser. The thing with operator is like it was kind of almost remoting in and controlling the browser. Exactly. This is a Chrome extension, so it's embedded within the browser. It's different. Yeah. And this sort of goes along with that theme of like the first people to bring about this concept was Dia, which was this company that got acquired.

21:31And they had sort of like a browser bar, but they also had their own browser. And then you had Gemini, which obviously from Chrome. And then you had perplexity. And so they're all sort of taking notes on each other for like interfaces. But again, these guys I think have just sort of nailed like you don't have to download a whole new browser. you can really like do things a lot more smoothly without law friction. What are the updates with cloud code? Okay. So there are any updates. Yes. So the other, not with cloud code, but a couple other updates besides just the extension while we're letting it go.

22:06So file formats, I'm a nerd. And so I, I get excited when I can generate a PowerPoint or an Excel spreadsheet. And so far the ones that I think have done an okay job or have been the best at it had been on the PowerPoint side, Gamma. And then on a CSV has actually been Manus AI. No one had really nailed it well, but actually Claude now can do both. And I think they can do a pretty good job. I still think Manus has maybe a leg up on the CSV and research, but for the most part, Claude is now able to do PDF format. Wait, Claude can do it now? Claude can do it now. so i oh my gosh i i was just doing this two weeks ago it was freaking frustrating because i was i built this it was an artifact within claude and the output was it needed to be in in csv format and it was like here you go all you have to do is copy it from here and now paste it into a csv and it like the formatting always sucked yep i think i showed this to you i think I complained about it, but now it actually does it.

23:11Yeah. I have so many artifacts. I wish Claude did a better job of like organizing your artifacts. Yeah. I have to, I agree with that. I have a bunch of financial statements in here. So based on the project details, create a financial analysis for this company, but change

23:36and provide this book in a PowerPoint for an investor presentation. Pitch deck. And CSV summarizing and aggregating the Gmail. We'll see how this works. It's a little bit more complex. I haven't tried it with this, so we'll see how it goes. If this nails this, this was like a bunch of grunt work I used to always have to do because I would get companies running me just PDFs and never the Excel files. And so I'd literally have to go and reformat everything and then build it all into this deck. So this actually would be killing part of the job I used to have to do way back when. All right. While we're waiting for that to go, let's just check in on...

24:27Oh, yeah. The Chrome extension. There it is. Oh, my gosh. So it added and we needed to see it fully in progress. But this is an interview with Justin Moore, author of Sponsored Magnet and a couple of points. So I can tweak the prompt to make it a better summary, but it completed the interview. Now, did you have to have that window open or that tab open in order for it to populate? Or could you have just given it the link to that Google Doc and it would have still populated for that to populate? Did you have to have the Google Doc open the whole time, that tab open? Or could you have just given it the link?

25:04Just given it the link, it would open it up. So if I had... Okay, it would still be open though. Okay. Yeah. That makes sense. Yeah. So it does need an interface with it. Like it's actually you in the browser. So that's also really helpful for scraping data, even though, unfortunately, Claude always is very ethical. And I've tried to scrape websites that you're not supposed to scrape, and it won't. So that is unfortunate, but it can do it. Darn you, Claude. Yeah. Stop being so ethical. It can't do it if there's no terms of service that it's aware of. Okay. So let's see how this is going. All right.

25:42We're still working. Oh, it's still thinking. Well, this is also, I'm not surprised because cloud 4.5, which is one of the new updates is it can have extended thinking for a very long time. Meaning if you have it in cloud code or actually I think even in just the browser here, it can do it for like just hours. I've seen as much as 30 hours, 20 hours, things like that. So it can actually process a long time. Yeah. Yeah. I haven't yet. That's a really long time. Exactly. I haven't yet seen it like that, but all right, let's see what's happening. It looks like it's doing something. This is not quite, I don't know if it's creating, I think it's creating an executive summary and then I'm wondering if it's going to then put it in the format.