This Week in Startups - OpenAI DevDay!: demoing GPT-4 Turbo, "GPT Store" potential, and more with Sunny Madra | E1841

Episode Date: November 7, 2023

This Week in Startups is brought to you by… LinkedIn Jobs. A business is only as strong as its people, and every hire matters. Go to LinkedIn.com/TWIST to post your first job for free. Terms and conditions apply. Coda is the all-in-one doc for teams. And they introduced an AI-powered assistant to take the BUSY out of the WORK! Get started for free at http://coda.io/twist IntouchCX. Want to build a loyal customer base for your startup? Unlock the power of innovative AI and automated support solutions from IntouchCX to deliver fast, personalized support and enhance your customers’ experience. Schedule your consultation today at http://intouchcx.com/twist * Today’s show: Sunny joins Jason to discuss OpenAI’s DevDay announcements (1:33). Then, Sunny demos an AI-driven chatbot creator (32:49), an AI email-first Chief of Staff (39:12), and much more! * Time stamps: (0:00) Sunny Madra joins Jason to break down OpenAI DevDay! (1:33) OpenAI DevDay announcements: GPT-4 Turbo, expanded context windows, Custom GPTs, and more (12:57) LinkedIn Jobs - Post your first job for free at https://linkedin.com/twist (14:28) How OpenAI's "GPT Store" could change the way we consume the internet (22:02) GPT-4 Turbo, speed, cost reduction, expanded context window issues (26:16) Coda - Get started for free at https://coda.io/twist (32:49) DEMO: Droxy.ai: a custom chatbot platform (37:54) InTouchCX - Schedule a free consultation at http://intouchcx.com/twist (39:12) DEMO: Mindy.com: an AI-powered email Chief of Staff (45:43) DEMO: Brave Nightly's webpage summarizer (50:24) DEMO: Zapier's AI-powered Zaps * Check out what happened at OpenAI DevDay: https://devday.openai.com/ Check out Droxy Ai: https://www.droxy.ai/ Check out Mindy: https://mindy.com/ Follow Sunny: https://twitter.com/sundeep * Read LAUNCH Fund 4 Deal Memo: https://www.launch.co/four Apply for Funding: https://www.launch.co/apply Buy ANGEL: https://www.angelthebook.com Great 2023 interviews: Steve Huffman, Brian Chesky, Aaron Levie, Sophia Amoruso, Reid Hoffman, Frank Slootman, Billy McFarland Check out Jason’s suite of newsletters: https://substack.com/@calacanis * Follow Jason: Twitter: https://twitter.com/jason Instagram: https://www.instagram.com/jason LinkedIn: https://www.linkedin.com/in/jasoncalacanis * Follow TWiST: Substack: https://twistartups.substack.com Twitter: https://twitter.com/TWiStartups YouTube: https://www.youtube.com/thisweekin * Subscribe to the Founder University Podcast: https://www.founder.university/podcast

Open AI is a major target. If you say we're going to allow, we're going to pay for your legal expenses, under what circumstances I wonder, because if people go in there and explicitly steal, then they could have a thousand drive-by, you can have 100 drive-by attorneys, launch 100 lawsuits each,

and just drowned open AI in lawsuits based on the training data. So this, yeah, I don't think that they've figured out some defense here. This is a very dangerous, thing to do. It's literally like waving a red flag in front of bulls. Bulls being drive-by attorneys who are looking for a target. Be careful. I don't know who came up with this idea, but this sounds dangerous. This weekend startups is brought to you by LinkedIn Jobs. A business is only as strong as its people. And every hire matters. Post your first job for free at

LinkedIn.com slash twist. Hoda is the all-in-one doc for teams. And they just introduced an AI powered assistant to take the busy out of the work. Get started for free at coda.io slash twist and in touch CX. Want to build a loyal customer base for your startup? Unlock the power of innovative AI and automated support solutions from in touch CX to deliver fast, personalized support and enhance your customer's experience. Schedule your consultation today at in touch CX.com slash twist. All right, everybody, welcome back to this week in startups. It's Madra Mondays.

This is this week in AI with my guy, Sunny Mandra. We are back. We are back. Exciting day. Big exciting day. Tell everybody why. Exciting weekend. Well, there's exciting weekend, too.

Yeah, everything was exciting. Yeah, I know. Well, let's start with a day and we'll go from there. So Open AI Dev Day, first one. You know, what's interesting is we're just, just literally little over 11 months in, right? Because this thing launched at the end of last November. And so at the beginning of this November,

it's like the first kind of developer conference. And, you know, Open AI came out swinging. They have launched a bunch of cost reductions to their existing models, capabilities to fine-tune GPT4, which is important for folks that are building certain production applications and use case-specific applications. And then a bunch of new features, one or two of which we'll try to demo here.

And we can talk about GPTs as well. We weren't quick enough to demo that. But that's sort of like an evolution of the plugins. And I know, you know, we've talked about that. We've demoed them before as well. The plugins were terrible. They were broken. And now it seems this new apps framework

that Sam Altman demo today, was really, it looked like super compelling because you didn't need to be a developer to release something in the app store. I thought that was the most fascinating thing that civilians. And monetization. And monetization. So you could build. So let's start there.

Because I felt this was the strongest part of the presentation, just in terms of something that I could see becoming a hockey stick. So let's go right to that. Okay. Well, so all of the GPTs, let's say, are going to be kind of forms of assistance, right? So let's kind of start right there. So what you'll see here is, and I've kind of created two, just for fun really quickly, I created one around the executive order and one around the all-end pod.

And so the executive order one, and you can see here, it's pretty straightforward to create these now. You basically give it a name, give it some instructions, pick the model. And the instructions you said here is you are an assistant that is an expert on the new executive order from the White House on AI. Yeah. You pick the model chat GPT4. Well, this is the latest one, the 1106.

This is GPD4 turbo that has the 120K context window. Explain to people what that context window does. Yeah. So that is effectively the amount of tokens that you can give it and it can come back out as a combination. And so... Which in plain English, an example would be a PDF of a book or a trans. transcript of a call or a bunch of spreadsheets,

etc., a database. So it's the, when people say tokens, in a way, what they're saying is the size of the attachment is one way to think about it. And it's not like correct, but I think conceptually you can use it, which is you can think about a token as a word, right? And so when you have like $128,000, like a typical page is 500 words,

so you can kind of do your master. A book is 60 or 70,000 words. So this is like two book, one long book or maybe two small books. Yeah. So you could take the book Angel, you could take the book, the power law. Yeah. And you could say, hey, make me cliff notes. And the whole concept of like cliff notes can be done for any product.

You don't have to wait for Cliff Notes to hire some writers and spend $50,000 making it and then printing it. It's just available to anybody anytime. And it can be customized, right? You can give it some instructions to write it to you like a pirate or like a child or whichever version. bullet points, tables, whatever it happens to be. Exactly. So we did the executive order and here you can turn on, you know, so functions is something you can get into a different time,

but this is like giving the assistant capabilities to interact with like an outside service. And so basically, you know, you'd be writing code if you're doing this. Code interpreter is, you know, the one that you've even tried before, which is related to analyze documents. And retrieval is the new feature as well, where you can now give it some documents. So in this case, I've given it like the overall summary and then our law firm DLA, they wrote an executive summaries. Also I've kind of given it both of those documents. And once you do that, you effectively have-

And now, what do they call it? They call these apps? What are they calling this? I think they're going to call them GPTs, right? GPs, okay. GPs, right? And I think that's how they've denoted them. and basically, you know,

you're up and running here. And, you know, this is still in the preview and the assistance. Okay, so what you're showing now is the, because we always like to explain it for people listening, is what looks like the normal chat chpT window. So in a way, all of the sort of instructions you might give it, the setup,

the architecture of this dialogue you're about to have with the language model has been done ahead of time for you. Yes. So if somebody wanted to create, a Shakespeare language model or a Shakespeare GPT, let's call it, a Shakespeare app. I'm going to call them apps. If I wanted to create a Shakespeare app,

previously people have trained a model on all of Shakespeare's work. Here, you could just upload all of Shakespeare's work, maybe some analysis from some famous people. And you could call it the Shakespeare GPT, or the Shakespeare app. Yes. And then I wouldn't have to do all that work as the next person who came in.

Correct. And you'll be able to publish this and you'll be able to, you know, kind of go off to the races. And so, you know, here, and look, we've seen a bunch of startups that were doing this. And this is like, you know, these demo days are becoming quite like when Apple used to have these days, right? They used to launch features that would wipe out apps, right?

That were doing, you know, features that should have been part of the OS. And so I think a lot of, you know, folks are building these, hey, take a document and create, and we'll demo one as well, but I think they have some good capabilities, better capabilities than this so far. So, you know, I took those two orders, I dropped it in here. And it does give you like an initial summary. And look, this is hot off the presence. Like, it just kind of came live about an hour ago.

It doesn't fully summarize the document here. It's only done kind of like two sections and there's multiple sections of this document. Yeah. what's super brilliant about this is now you can take this app. And your app is the AI executive order app. The one I described was my Shakespeare app. Another one might be, you know, Gordon Ramsey's cooking app. And Gordon Ramsey can put it in every single video he's ever made, everything he's ever made,

and they can publish it, and then it will go to the app store. And Open AI is going to have the equivalent of an app store. And they're going to have the ability to let you charge for these. Did they mention how that works and then the share, the revenue share? They didn't get into a lot of details, but you know what was interesting? Rather than like a cut model, which is what we have in the app store, it felt like a more of a Spotify model, which would be usage based. And so, you know, the way Spotify works is when you upload music there, it's a, you know, you get paid as a, you know, like a function or representation of how many streams you have relative to everyone else. And so the short blurb that Sam gave was more aligned to that type of business model than a charge $1.99 per use.

And so maybe what they're trying to do is say, look, people already paying us $20 a month. We'll just give you a portion of that, like in Spotify, depending on how much people are using it. Got it. So they'll create a pool of money that goes to people who release these in their app store. And you don't subscribe to somebody's Shakespeare app. you, they would just, hey, you know, a thousand people are using this a day, so we're going to send you whatever, 500 bucks a month. Just a, you know, kind of a non-specific revenue share, which is probably a mistake. They couldn't send these and give those, but they should come up with a legitimate, um,

tied to performance based use case. Now, you got to be careful doing that because then there could be abuse in the system. But I think, you know, if paid users use your app, X number of times you get paid for X amount of usage could be pretty easy. For every search they do, you get a penny. So if there were, you know, a hundred Shakespeare app searches, you got a dollar.

And if there were a thousand, you got ten. If there were ten thousand, you got a hundred. Pretty straightforward, right? Yeah. And so, look, I think it's going to keep evolving, but this to me is a good evolution of plugins. You know, I think a little bit of a concession by them that the original, and it's good.

Like, they're moving quickly, right? the whole thing has only been around for 11 months. So they're conceding that that wasn't the right play. They're going to have a store for these. They're going to make it such that non-developers can create them as well. And I think there's going to be a lot of innovation in the short amount of time. It's excellent for Open AI as a platform because they're going to get a lot of mindshare. This is an interesting thing that I also saw, just from a relative standpoint and kind of

tying back into your investing world, there was only about 40, 50,000 people on the stream this morning. Interesting. And so if you compare that to sort of like the size and scale of an Apple dev day or something like that, right? It's quite small. So it's really, really early. Like, we're in the thick of it, but I think if you're building something here.

This is literally like the first time Steve Jobs shared the app store. People are like, huh, maybe. But to say, and by the way, Sam did a great job. So shout out to that moment. I texted him, congratulated him. later this month we're going to launch the GPT store

so that's coming and how that works who knows but I can see this be very interesting for people who are content creators in other words

remember there were some authors who sued people I think they might have sued Open AI for ingesting their books yeah well this gives them

an off ramp they could say you know what in the core model we're taking out Stephen King's works yep but there is a Stephen King

app that Stephen King gets 70% of the dollars in. And if you want to quote Stephen King work and you want to build stuff that's Stephen King related, please use his app, right? So in a way, this reminds me of what YouTube did around copyright, which is instead of taking down all these videos, they allowed people to claim ownership of that.

So imagine somebody has a photo library and Dolly is using it. Well, now you can say, you know what, if you want to make images, by the way, check this box and you can use the Getty images. And there's a Getty app or GPD as they call them. And they just call them apps. I don't know why they're calling them.

GPT's. I'm going to call them apps. But imagine if Getty just said, okay, yeah, we'll put a link to all of our stuff. And you can make images based on these apps and we're done. All right. Congratulations to the team at LinkedIn who just completed the march to a billion users. Can you imagine a billion users now on LinkedIn? Of course, we all knew it was going to happen.

And it's happening. But, you know, it's tough for startups out there right now. And finding great candidates, it's hard. It's always been hard. You want the elite candidates. Where are those elite candidates? Well, there are some part of those billion users over at LinkedIn. And it's your job as the founder of your startup to get the right team members, get the right people on the bus before you decide on going to your destination.

LinkedIn Jobs is going to help you find that next amazing hire, the one that's going to level up the barraiser, as they say. There's also those passive job seekers. You know the ones, they got a lot of job offers. But, you know, they check LinkedIn once in a while. So what you want to do is you want to post your roles to LinkedIn. You want to get active on LinkedIn sharing on your feed, doing some blog posts. And then you want to post a job as well. Then you're going to be 100% certain that you don't miss out on the most qualified candidates. Everybody's on LinkedIn. That's why LinkedIn jobs is going to help you find the qualified candidates that you want to talk to faster. So I want to give you your your first job listing for free. LinkedIn.com slash T-W-I-S-T. LinkedIn.com slash twist. That's LinkedIn.com slash twist to post your first job for free and watch the magic as you get the greatest talent ever to apply for your open positions. Terms and conditions do apply because they're giving you something for free.

One of the other things they launched today is copyright protection. Did you see that? Explain what that was. I was confused by that. Yeah. Yeah. Explain that. So, like, let me try to pull up the article.

code. Maybe Nick, you can out pull it up. It's called Copyright Shield. Yeah, Copyright Shield. Open AI will pay the cost incurred if users face legal claims around copyright and infringent on both chat GPT, Enterprise and the API. Yeah. So that means if you build something and someone comes after you saying, oh, hey, inside our, you know, in your model, and you know, you've talked about this a lot, right?

Yeah. It looks like my work is in here. They're going to help protect you. So they're really standing up to their ecosystem, you know, for their ecosystem that's going be building these things here. This seems like also though, waving a red flag at a bunch of bulls,

which are the lawyers. They have an unintended consequence. Open AI is a major target. If you say, we're going to pay for your legal expenses, under what circumstances, I wonder, because if people go in there

and explicitly steal, then they could have a thousand drive-by, you have a hundred drive-by attorneys launch 100 lawsuits, and just drowned open AI in lawsuits based on the training data. So this, yeah, I don't think that they've figured out some defense here. This is a very dangerous thing to do.

It's literally like waving a red flag in front of balls. Bulls being drive by attorneys who are looking for a target. Be careful. I don't know who came up with this idea, but this sounds dangerous. Or maybe it's not because they're confident that they have no copyrighted work in there. Oh, okay. Sure. Maybe you think they started pulling stuff out?

Yeah, and so they, like, I wonder if they did that. Do you think? And they use all AI to make sure that they don't have anything in there. And they're so confident that come after us because we can show you that none of our training material has copyrighted work. Yeah, I wonder if that's the case because there was Reddit, Twitter, Cora, Facebook. A number of people were like concerned about these open crawls on the internet. Yeah. And, you know, one of the things about these open crawls of the internet is if you're using one of those.

open crawls and we talked about it, right? Is it even called open crawl? I think it's, that might even be one. Common crawl. If you're using common crawl, it's in Common Crawls terms of service, it's your responsibility to then know the terms of service of who they crawled and what you're allowed to do with it. So any LM or LLM, you know, based on that common crawl,

it's up to that group to then go make sure that it's not stolen material. For example, if someone, There were a ton of websites that used to take my tweets or other people's tweets and republish them to their website and do like some analysis on them. Yeah. So if you were to do like, you know, this like Twitter counter or whatever and they had scraped and published that data, they broke the law doing that. Twitter never went after them. Didn't know it wasn't a big of target. It was offshore.

Then you scraped it. Now you've trained your model on stolen content, which is like you getting, buying a stolen bicycle. It's still a stolen bicycle. You just have to give it back. Yeah. Yeah. So this is going to be super complicated. It's kind of a bold move.

Maybe they're doing it for optics. Maybe they just, you know, they know that they could burn 100 million on defending lawsuits and not worry about it because they have so much cash. You know, it's a $90 billion company. Maybe they got so much money. Or imagine like, because, you know, they've invested in Harvey. That's like a legal AI. Imagine they fight the lawsuits with an AI. So they know they can outguns.

the drive-by's. Exactly. It's just pretty interesting. So this, I think, is going to change everything because this is like an app store that anybody can build apps for. Yes. This is going to be explosive in terms of its growth. I think there are going to be a bunch of surplus.

There's a bunch of surplus cognitive power in the world. When you're around on the weekend and you're like, what should I do tonight? That means there's surplus cognitive power just sitting out there. So there's a billion. Cash poor is another way to put it as well. Exactly. So you might have a billion people who are sitting there on a Friday night with nothing to do or they're unemployed on a Friday during the day.

And they go, you know what? That's a good place to make money. I should start building stuff there. And what that does is now you've got everybody finding and crawling and finding all the edge cases of interesting use cases, which is exactly what happened for Apple. Apple figured out that people wanted a flashlight. people wanted a calculator people wanted a stock app a weather app uh they wanted messaging apps that had all kinds of cool features in it they wanted an app for images that would let you put filters on it

and as you mentioned earlier every Steve Jobs was very in attuned to this so he would give people like three or four years to exploit the flashlight app and then like here five of the flashlight app they'd be like you know what we're going to put that in the control center yeah you're five of disappearing messages or emojis and messages. Okay, yeah, we're going to incorporate that into iMessage. Yeah. And so, or even groups, you know, like all of those features eventually made it into iMessage where now iMessage feels in some ways very similar to WhatsApp.

Exactly. Yeah. So we're at the start of it. I kind of agree with you 100%. I think there's going to be a lot of momentum in this and we're going to see a lot of innovation. It's really going to change the way we consume the Internet. I think that's the other fear because,

or maybe opportunity, depending on which side of the fence you're on here, instead of consuming the internet through the end sites, you'll find the bot to interact with, and the bots will have all different levels of capabilities depending on the, like apps, depending on the sophistication of the developer. And you'd be like, wow, Sunny,

I found this really good travel bot, and it's great for A, B, and C, you know, and it does everything you need. And so I think we're going to see, you know, one of the things that I kind of predict, you know, put it as a tweet out there and I say prediction, was that we're going to really start seeing a different way in that the Internet's consumed going forward. Yeah.

This could change a lot of things where, yeah. The same way apps took people away from Google Search. Now, Google Search has kept growing because overall use of the Internet keeps growing and people keep getting higher speed connections and more people get online. And they have more. They paid a little $18 billion a year to suck down that Apple traffic. But the App Store did, you know, take people away. When I search for restaurants, I use the Yelp app.

And then maybe, you know, sometimes I'll fall back to Google, local, into a Google search. But they did take away 90% of me searching for restaurants. That's but one example. And so the question is, will chat GPT apps or GPTs, OpenAI, GPD chat, whatever. There's so many names for this company. Will those actually intercept traffic or not? That's going to be interesting question. And then could they be published as apps?

The answer is probably yes. And Sam made it pretty clear. We're going to pay people who build the most useful and most used GPs in our GPT store. We're excited to share more info soon. That says to me it's an experiment right now. They don't have a formal process. Also Spotify like, though. It feels like that's why I was going.

It's like Spotify like, you know, you got to put things in there that people use. That's how you're really going to get paid. Okay. There it is. Yeah. All right. Okay. So that's apps or GPTs.

What else? And they did talk about also lowering the cost of the API. Yeah. So that's kind of tough to demo. But what I'll demo is alongside that they have this new model. It's in preview as GPT4-1106, right? That's today's date. And this is GPT4.

And if you ever used it, it can be slow. So I'm going to just do this here. Tell me about the theory of relativity. And if I click it here in the playground, you'll see how quick it. It spits stuff out now. So it's running about as fast as GPT-35 Turbo was now. which I think is going to be really useful because the capability of the GPT4 model is significantly, you know, better than GPT3-5. And so getting the cost down, I think, something like 7X and then getting the speed up is going to create really good experiences.

You know, one of the things you and I have talked about is that back and forth speed is still a little bit lacking. So now they're just cutting away at that. And I think there's one more iteration from it being like real time where you won't even realize that you're even talking to something that's processing it on the other side. Yeah, I felt like this typing and the haptics where it's like typing out your answer and you're kind of following along with it. It was like exciting and, you know, you know, was kind of enthralling a little bit for that first year. And now it's just annoying. Yeah. Like I like, I appreciate Bard when I just type it in and barred just like snap, boom, here it is.

Just give me the whole thing. It's not the whole answer. And then, you know, with the talking to your assistant like the her kind of modality, It's just too slow. This whole CB radio thing where it, you have to say your request over and then it does it over. I don't like that. I like to be able to stop it and say, whoa, whoa, whoa, whoa, whoa. I don't want seven restaurants.

I just want the one best one. Okay, sorry. Sorry, boss. But the rate of innovation is incredible, Jake. How, like, you know, for the longest time, the measure that we used was like Moore's Law, right? And, you know, Moore's Law was like every 18 months, we'll get, you know, twice as many transistors or something like. that. Yeah. And so if you look at the speed of GPT4 in less than a year, right, these are multiple, we're not talking like it's 2x faster, right? You're talking like 7x faster. So I think

the speed at which this is happening now, it's going to really give us some of these experiences that people are looking for these kind of human-like, almost instantaneous, you know, sub-a-hundred millisecond kind of things, right? Or maybe 250 milliseconds. Well, and here we are. Yeah. So what else did they launch today? I think that those are the major announcements. I think like some new APIs, the assistant APIs, the GPTs, and then the Privacy Shield,

those are the major announcements, right? The token window, which I'll just say is the attachment window. Correct, correct. You know, just so I can translate this for people. These are apps. I'm going to rebrand all their stuff. Okay, go do it, do it.

They're not tokens, it's attachments. That's it. Let's stop with. the context window. Context window is too confusing. Attachments. Attachments go up to 120aK. So just like Gmail, remember

when Gmail had like the 5 megabit limit and then it went to 10? And then they were like, okay, do whatever you want. We don't care. So that was huge because Claude 2 was up to like 100K in tokens. Now they leapfrog them. So this means they're paying it.

This says to me, open areas, paying attention to the competitive set. I just wanted to one up it. This also costs a lot of money. I got to think when you're uploading a book where you upload an episode, you, I think you upload a transcript of the All In Podcasts to, exactly. They're like, you know, you can then put the whole transcript up and then start asking questions about the episode. Pretty cool. Yeah. Yeah, I'll just pull that back up real quick.

Yeah, so I was doing in the background we were talking, but I uploaded the whole transcript here, asked for quick summary. And so, look, you know, the other thing is there's a lot of tools that we're doing this now. Now you can kind of build your own bot here. That can be an all-end bot. You can create a pipeline from Descript or whatever tool that you guys use there to, to transcribe these things. And it's all, you know, you'll have a bot instantaneously that you can chat with about episodes. Incredible. All right. Let me tell you about how I'm able to manage thousands of applicants and hundreds of attendees for Founder University. Every time we do a new cohort, which is quarterly, we have thousands of people apply. We track them all. Encoda.

Coda is a dock on steroids. And then when they get accepted, we have three, four, five hundred founders inside of Founder University. And every week, I ask them to send me an update on how their startup is doing. And then we can look at all that in Coda. Then we send automated reminders to those folks. And we track their week over week growth with these beautiful charts that Coda produces for us. Let me just tell you, I was going to build all this with a developer. But my team, they built an app in Coda in days and saved me, I think 100 large.

They also just announced Cota 4.0. This includes Coda AI. Basically, they've made a work assistant that knows your company. And then Coda AI will actually perform tasks for you. But here's the best part. It's free for doc makers. You can make all kinds of great documents inside of Coda. You can build databases and great workflows. This is the future, honestly, if you want a platform that will empower your team to work, organize and collaborate together at lightning speed and speed matters. I always say this velocity.

Well, get started with Coda today with a free trial. Head over to coda.io. That's Codda.coma. dot I-i-o slash twist and get started for free. I am obsessed with Coda. I love Coda. So go to coda. And get on Cota today for free. I think that covers off all the major announcements from opening I day, which was really fun. The next one I... And the cost went down just so we're clear. clear, if you blend it at all, it's about 2.75 times cheaper than the GPT4 model, according to our notes. Yeah.

So faster and cheaper and bigger context windows. So you get a lot of. Yeah. So they're working on cost reduction because they want more people to use it and they put speed on the back burner for a little bit. Yeah. They want it to cost two to three times less money to do stuff. One cent per 1,000 input tokens, three cents per 1,000 output tokens.

So pretty great. And the knowledge cut off. for GPT4 turbo is up to April 2023, I see. So this is interesting that these are the variables now. I bet you we'll be sitting here in two years,

three years. These will no longer be the variables that we talk about. Cost won't even come. Yeah. Next year, we won't be talking about cost, just like nobody talks about

storage cost anymore. That was like an issue for three or four. That's an issue for maybe a decade. People talked about online storage costs. It's not going to be even discussed. and I think the attachments you know context window okay context window

we're not going to even talk about that because it would be like it'd be like talking about like sharing like human like you know at 128,000 or 828K it's already beyond most human like you know you can put books in at this point yeah but if you uploaded a movie

if you uploaded Blade Runner the entire film and you wanted a frame by frame analysis kind of situation sure that would be a lot bigger than the current context window so I would think. Like if you said I wanted to analyze each frame. So, yeah, you know, it's just, this is exactly like that with storage and transit.

There was a time where the idea of sending somebody an entire movie or even an HD video that was more than a minute was you needed to have something like, what was Kim.com's company? Oh, yeah. Share file or something. No, it was something for sharing files that he did, you know, like back in the day, you couldn't share. large files he had to use like you know these third party attachment things where you would upload it would take an hour then you'd send somebody to the link they would download

it Kim.com had a service that he got busted for yeah mega upload there you go yeah mega I think it was like for short yeah and then those things became unnecessary that's what's going to happen here with all this stuff but I'm really interested in the apps I thought that was the biggest announcement for me today

is the apps all right so should we rate that should we rate this stuff Oh, yeah. I rate the apps. I'm going to rate the apps. I'm going to rate the app announcement as and the functionality that you displayed it, a B plus. Okay. That is a B plus out of the game.

I'm going to go B minus. And because actually, there's some nuances of what we've done here. So in both cases, you know, in the one case I uploaded the White House executive order, in the other case, I uploaded the entire transcript of the last all-in pod. It wasn't able to analyze the whole thing. And so this functionality just came live and maybe they're throttling it. I've been getting some errors and timeouts. But from the engineering standpoint, it's meant to work.

It's super easy. But it wasn't doing the whole thing. So we have to call that out. On a vision, I feel like it's almost all there. So B plus, if they had the pricing and everything all sorted, that I'd probably go like A plus. But it's kind of like the app store in the early days, which is like, who knows exactly what the rules are going to be.

But I give that a B plus for me, B minus for Sundeepe. And then for the chat GPT4 turbo, I don't know how to rate that. That one's harder. I think let's wait for people to integrate that in. I think that's more of like a developer kind of feature. Nick, anything else we had to rate there that we didn't. No, it was good. I thought

what Sunny was referring to is the GPT4 new developer model in the playground was only grocking, if I'll use that word, the first like 15 minutes of the transcript or maybe 10,000 tokens or so. Whereas

Claude too, which I use for almost everything that needs longer context window, more tokens in the context windows, we ran the same transcript in Claude and gave a very similar prompt. And it gave basically a perfect summary of the entire episode, right? They even joke, like, they even had the, all the way at the end, the host closed the show by joking about David Sacks, unreliable RSVP habits, right? Almost perfect, um, I mean, it's nuts.

Yeah, it's pretty amazing. Claude is nuts. Chad is not. Chbd wasn't there yet. So I'm still giving Claude the edge in this specific. Okay. All right. So you would go with, uh, okay.

So for the token window. Yeah. Okay. I got it. Yeah. All right. Great. Uh, let's go on to our next demo.

Okay. Congratulations to the opening I team. Yeah. So the next one, you know, for these guys, I actually used this last week for a little project. And that's why I had some of the context ready. It's called Droxie AI. And what they have is like a platform that allows you to create chatbots, very similar to what we were just, you know, showing off there.

But I want to maybe lead into this one where you can give some guidance to teams like this who are building things that are getting, you know, a bit maybe cannibalized from the platform. So in their case, you know, you can create a chat bot. You can upload your content. You can do your configuration on which model you want to use. And, you know, what your behavior is very, very similar to what you saw there. You can customize what the chatbot looks like. You can integrate.

You can do integrations. Look at the conversations. And so here, you know, once this is done, you can click off to it. And similar here is like, you know, you can chat with it regarding the executive order, right? So you can say, you know, give me a summary. So the same Shakespeare app or the same executive order app can be built here with proxy. So you ask me the original question, how can you differentiate? Very simple.

If you allow white labeling and you allow these things to exist outside of your domain name, then that's something that chat GPT is not going to do, right? They're going to have all of their apps or GPTs in their store. It's going to be part of their flow. Sometimes you want to have these things on your own domain name. you want to control them and you want to monetize them, how you want to monetize them.

So it's kind of the equivalent of if you were on Patreon, right, or substack, and you have to exist in their ecosystem on their domain names, typically, you can do custom domains on substack,

but essentially a substack is you're on substack for better or worse. But then there are things like Ghost, which is a back end, and then there is like Stripe for memberships, and there's other membership software, that kind of disappears into the background

and lets you do these things on your own without having the rule set and living in the ecosystem of another player. And that allows people to do things that they wouldn't be able to do when they're inside of that. So I kind of like those, right? Yeah.

And so that's the right path to go. Yeah, so during the Dev Day, they showed a demonstration where there was like an assistant that was integrated into like a third party site, like WonderList or something, but it's a very dev-heavy integration. I think what you can do here is what you talked about is make it such that a non-developer, low code, can create one of these things for your customer service area, whatever it happens to be for your business.

Yeah, put a domain name on it. Yeah, exactly. Just slap a domain or hit export and make an app out of it. So you could take all this code and release it as Swift or whatever programming language you're using and then you could then compile it and make an app out of it, right? So those are the ways to kind of free people from being caught up in somebody else's ecosystem, which is always dangerous. Like Sam Harris had a Patreon, but he didn't want to be beholden to Patreon and have them take his money away because he watched them do it to somebody. And the person they did it to was somebody doing something abhorrent, right?

But he's like, okay, they took the person down doing something abhorrent, right? let's just say they were like doing hate speech. So he was like, okay, well, at some point they may come for me for doing something controversial. Therefore, I don't want to have that attack vector. And I want to own this relationship with each of these people. So I'm going to move off the platform. Just like some people will, we use as an example, Wistia. It's a great video hosting service for businesses.

And we can white label it and use it for our videos instead of using YouTube. So we're just independent of that, right? some business users will want to use that, right? Very simple. Yeah, and I think it's a good idea. So, like, I really liked it. I used it. You know, in fact, I used it for some of our documentation as well.

So, and, you know, kind of continue down that path to create something that allows you to white label it, put it out there, make the workflow very simple, allow you to attach a domain to it, all the good stuff. Pick a monetization model. Like, you're going to be beholden to the monetization model that Open AI has, just like you are, beholden to Patreon. Here, you could tell people sell it one time, sell it per query, sell it for a minute, whatever you choose, whatever your jam is, subscription, etc. So, yeah, I would just listen

to your customers and keep building features that chat GP doesn't get to. There's a long tell of features that people would want. And Enterprise and White Label are amongst them. Great. All right. I'm going to keep rocking and rolling. I'm going to give them a B on their service right now. Yeah, I'm there with you. I think they've done a good job. It works really well, super simple, easy to get set up.

They just need to add sort of the kind of the distribution loops into the application. All right. Both B's good stuff. Listen, a lot of times when you're building a startup, you have to take on better funded incumbents in your market. It's simple as that, right? This is David and Goliath. This makes the startup journey really hard. But one of the proven ways to beat an incumbent is by building an insanely loyal customer base.

We all know this. You want to have those advocates. You want to have those promoters out there working for you. and one way to build loyalty is by giving your users a wonderfully customized experience. You might not have the time to manually analyze every interaction you're doing, which means you're missing out on key insights. But InTouch CX is going to help you solve this problem. They can help you build automated customer experience solutions

so you can personalize every interaction. InTouch CX is going to help you deliver tailored experiences through voice, email, and most importantly, what all the millennials and the Gen Zs love, chat support. So improve response times and increase productivity by 30% or more. This is going to help you drive greater satisfaction and loyalty with your top users, the advocates, the promoters, make every part of the customer journey a personal one with InTouchCX. Get started with an automation expert at intouchcx.com slash twist. That's in touchcx.com slash twist.

Next one. All right. I think you're going to like this one, J. Kel. So this one's called Mindy. It has no interface. It's email. And so this one, I'm going to jump into my email. So when you get started with it, basically this is like the first email that you get.

And it says, hey, my name is Mindy, glad to work with you. And just reply to this email so that you agree to terms of service and let's get started. And then basically you get going. And so you can get a virtual. It's an AI chief of staff, it says. Yes, exactly. going to sort your email? Doesn't sort your email.

And so, like, the examples we're going to look at is it can do research on a topic for you. It can analyze documents. It can organize meetings. And so I'll just show you here because it's not instantaneous. So it takes about, you know, in my case, look like 20 minutes. So I did the same thing. You know, we're on this summary of this executive order. So I had a PDF of the executive order, same one that I put in DROXC, same one that I put

into chat GPT assistance API. And basically said, hey, I have this PDF. executive work of the White House, can you summarize it for me? And basically it sends me an email back of a summary of it. So inside your workflow, and I know, you know, Jake, how you've been talking about getting people

to just use these things in their existing workflows. Now it's here. And so... In Gmail or in your email. In your email. So this is like a chat assistant, a co-pilot for Gen Xers and boomers

who live in their email boxes. I love it. Yeah. Yeah. Well, what's great about is you can see them, you could see C, Mindy. Yes, as well.

And so I did another one, which, you know, pretty cool, I said, Hey, Mindy, could you have a calendar right for a weekly

poker game this Thursday? It did. It sent me the calendar invite for the poker game, which is right here. And, you know, sort of like,

well done. I got to give them credit on this. And the last one, you know, again, you don't have to even give it an attachment.

This is one of their examples. And I wanted to summarize this Supreme Court opinion and just gave a link to it. It did the exact same thing. It gave me a quick summary of it. So this is really low barrier to entry using assistance for a lot of different workflows and use cases. Love it.

Yeah. I give this a B. It's neat. What I like about it is the ability to CC it and have it suggest things. Yeah. So I like the idea that Mindy is CCed on these communications. And then she might say to me in Slack or whatever, hey, you didn't get back to this person. Yeah.

Hey, put this on my to do list. So this idea of an integrated in my email box, I wanted to really sort through stuff. So already you have Gmail and superhuman putting things into tabs. What I wanted to do is show me a bunch of stuff and say, I think this is not important. Am I correct? And you say, yeah, it's not important.

Never show it to me again. I think this is a founder pitching you to invest in your company. What should I do with it? And I say, oh, you should forward that to this email. So I have an email for our database team. Okay. And I wanted to, I want to say, yes, that's correct. In the future, I want you to automatically email any stuff from a founder to this email

and let this person know. And if it's a biotech company, I want you to respond and say, we don't invest in biotech. So I could see this assistant learning and doing reinforcement learning. And then I say, well, that's my wife. That's my family member. That's my brother. This person works for me at launch. This person's an LP.

And if it started to learn who the different people were and how important they are, so it just asked me, hey, I think this person is really important. On a scale of 1 to 10, oh yeah, that's a 10, that's my wife. Oh, yeah, that's a 10. That's, you know, the president of my organization. Oh, that's a 10. That's a CEO of my most important portfolio company. Then it could start to learn over time. And then this is where the talking interface, I would love on a wide-treed monitor for a just to go, boom.

Yeah. Here's 10 emails. Let's go through them real quick. number one, this, number two, what do you want to do with this? Number three, what do you want to do with this? Number four, what do you want to do with this? Can you imagine if on my thing, it just brought up 10 emails at a time and said, these three emails look like their startups pitching you.

Would you like me to send them to Andre? And I would look at them and say, yes. Boom, done, dispatch. Okay, these two look like they are press releases. Would you like, what would you like me to do with them? I would be like, delete them, delete any email. from that person before and put them on my spam filter. Yeah.

Boom. You know, like, because that would be like I'm training and what to do in real time. Yeah. And then it could just show me. This email came in. Yeah. Imagine you're sitting there. And as, as email comes in, it goes, an email from a founder who wants you an invest in this

company came in. I'm sending it to Andre in five, four, three. You know, like when you open ways, it gives you that countdown clock because it thinks you're going to your office or whatever. and it's just like going to your office in five, four, like that, boom, and it would just dispatch. Wait, wait a second. Let me see that one. Yeah, you know what?

I'll respond to that one. Yeah. You get the idea. Yeah, I mean, what's interesting is what you're talking about is a little bit of a mix of an end service but also an email client. And what these guys right now are they're an end service, right? So they'd have to get CCed on every message coming in and then process it on their side. it's almost interesting to see what could happen if you took some of the ideas you're talking about

and moved it into the email client. Yeah, shout out to the Mindy team. Yeah, shout out to the Mindy team, email me. Like an email assistant that sorted emails a much better idea than just talking to one, like creating a gateway. I would study how people are drowning an email and how they route them and how they dispatch emails

and be the email router or sorter part of Chief of Staff. I think that's a really good kind of, you know, a direction for them to go. in there, I think it can be really powerful. Well, I mean, and superhumans done some stuff, but what superhumans done is really around responding and the email, composing emails, right? Yep. But I really want the higher level stuff, which is sorting through them.

Sonny, you got a grade? Yeah, to me, this is a B plus. I'm giving it B minus in this one, actually. Okay, all right. I feel it's very simple as presented, but I love the potential. So, you know, keep impressing us here, but great job. Great start to the Mindy.com and great domain name too. Mindy.com.

I love it. Love it. Exactly. Great. Okay. Next one. We're going to keep, we're going through a lot of these today.

Cook with oil here. Yeah. I like it. Okay. So Brave browser, you know, most devs. Oh, my favorite browser.

I love Brice. I use Brick. Excellent. Okay. So Dave, in their nightly build, you have to get their nightly build. Okay.

You know, you download the browser. You do your regular thing. Oh, you know, I'm trying to learn about the Queens Gamit. I'm going to go to the Wikipedia article. you know what, like screw it.

I can't be bothered reading it. You just go right here, you click on this thing called Leo. You say summarize this page and boom. Beautiful. Right? Yeah. And you know,

like I could be, you know, in tech meme today and, you know, I can't be bothered to read about what happened at Open AI Day. But like, you know, I really trust.

Just give me the short summary of that page. Yeah. Look at that. Fascinating. Fantastic. Right. Yeah. So this is a great feature.

And then I think it should do this for bookmarks. So I would say this summary should summarize and bookmark and put into my knowledge base. Okay. Which is to say, what I would actually like this to do is I would like Brave to take every page I browse and organize them, right? Okay. So instead of doing this, I want Brave. And this would be a paid feature I'd pay 10 bucks a month for. Okay.

Every single page, instead of having a history, build a summary. of the page, give me a picture of the page, and build me a knowledge base and categorize it. So then I can say, I would, you know, I surfed a page. Okay. I remember there was a story about Corvettes being converted into EVs. And it would be like, boom. Yes. Here's the story you read on this date.

Yeah. And here are other stories related to a boom. That would be a killer feature, right? So this idea of like your history, your browser history becoming, an AI, a language model into itself. That could be killing. Integrated into everywhere you've served. And then we always have those moments like, oh, man, where was that?

Then you're searching like crazy and you can't quite find it. And it was the one particular article that you got through to some link. And then suggest things to me based upon what I've done. You see I'm going and looking at NICs. You see me looking at recipes for breakfast foods. Okay, you know, I'm into, you know, I dig my brunch. You know I dig my next. Give me my next thing.

Love it. So I give this a solid B. Okay. This is something we all want built into a browser. And it shows incredible potential of, you know, browser plus AI. I think they should keep pulling the string and see where this leads them. Yeah. And you can see here from the, you know, monotization.

And you can get different tasks. You can up, you know, kind of upsell into more advanced models and conversations about the page. So not just summary. Yeah. But like conversations, generation and. And so this is a great place for them to put those features in. So it's a great monetization, as you would say. Lots of monetization here.

Lots of monetization. All right. Yeah. So, okay. Yeah. You didn't give a grade. What do you give them? I give these guys a B2.

I think this is great. It's a good start. You know, I use kind of brave, not as my main browser because we're, you know, I use Google Workspaces for everything. But I think this, for my browsing browser, I may move all the activity over here now. So I'm very excited. You can do it with one click, by the way. It moves everything over

because it's based on chromium. Yeah. And so, but the other thing that happens with the brave browser is because they get rid of all the fakaka, you know, all the cookies and ads and popovers.

It's incredible in terms of the speed to load pages. And also, when you have pages that are just too heavy that have all kinds of crazy pops, it doesn't take over your browser. Now, I know that's only like one out of 50 pages

you might go to, but it really does speed up. your browsing. And when you're, also I would check out the Brave mobile browser. Not to make this a commercial for Brave,

but shout out to my friends at Brave. Because when you're on mobile and you're on a really crummy connection, like all that stuff is like 50% of the page is stuff you don't need. Oh, nowadays it's really bad. It's really bad.

Some people are like, you go to the New York Post or drudge or some of these pages. It's just like, I don't know how many cookies are going on it. like, first the cookie box and then you've got to sometimes an add and the X is like too small. Yeah, I got X things out.

It's just, it's too much. So when you're using it on mobile, you really can see how dramatically faster it is. Okay. Oh, I'm going to give that a try. Yeah, it's pretty great. Okay. And let's go to our last one,

which is, I know you're a big fan of these folks as well. Zapier makes you happier. I love Zapier. Exactly. And so, you know, I didn't do the end-to-end demo here as well because the team doesn't sometimes like it when I integrate things into our Slack. So I don't always get the full freedom. But here you can see how straightforward it is.

And I can say, when I get an email from Michael Gert and send me a Slack message. And you and I were having a kind of a group chat conversation this weekend about like, hey, when certain emails come in, you should put it into a Slack channel so you can talk about it. So they're just making this super. I was speaking specifically about expenses. I wouldn't say whose discussion this was. But I have a channel for expenses. So when producer Nick or Presh or somebody on the team expenses something,

I and the accountants or whatever see it in real time from, I don't know which card it is, but whatever card we use, it zips it over into Slack and I see it. Now, what's interesting is I'll see things and I'm like, oh, we're paying for that. I didn't know that. I might want to use that.

Or what is this thing? I don't even know what that's about. So I can just reply to the Twitter threat, to the Slack thread. What is this? then imagine fraud. Imagine somebody were stealing money

or using the corporate card in the wrong way. You could be like, hey, who went to Miller and Lux and a Warriors game? And they'd be like, that was you,

Jay Callan. I'd be like, okay, that's a lot. But if it was somebody else, I'd be like, what's going on here? You know?

Yeah, I haven't been a Warriors game in three months. What's going on here? I know. Let's go. Yeah. But so, you know,

the idea, you know, look, Zapier had always done a good job of making these things pretty easy. I think it's now at the lowest barrier. You just come in here, type in what you want. It then creates it for you. And then you just kind of go through their usual builder to connect it to all your services to make it happen. Right. And so they couldn't have made it easier.

I think they've done an incredible job. And there's no reason you should not be using the service now. Yeah. So when you previously did this, you had to do if this, then, that, bullions, and you had to like do each thing. It was a little bit, I don't want to say it's complicated, but I think, you know, the bottom 50% of your company, you're depending on how tech savvy they are, like at a small business, I would say the bottom 50% would struggle with this. The top 20% would do it easily and 30% would take a, you know, 15 minutes to figure it out. They'd have to watch a video, right?

This is just going to make automation a lot simpler. Yep. If you can do it in natural language. So this is another example of making things easier for. civilians to be really good at technical tasks. Yes. So it falls into the whole thing we did with, you know, the code interpreter, which is another terrible name.

What was doing the branding at opening eye? Just call me. I'll tell you what the call things. They shouldn't call it code interpreter. They should call it data scientist. That's it. Well, they changed the name like twice. They called a code interpreter.

Then it was advanced data analytics. And then today they called a code interpreter again. Just call it analytics. Data analyst. I like data analyst or data, you know, something like that. But anyway, the point is making things easier to use. So if everybody could be a good writer today with AI, that's great success. Everybody could become good at scripting and doing Zapier.

That's amazing. If everybody could do basic U.S. design, if everybody could do basic music composure, the entire society will go from being neophytes to good. So, like, you no longer have to suck. Like, 80% of people suck at everything. Now everybody's going to be good at everything. And then there'll just be some people who are elite. But we wouldn't have this like elite and suck.

Yeah. Just everybody will be okay to good. Sort of the mobile phone did that for everyone too. Like, I don't know if you're, you know, once everyone got on the smartphones, well, look, once everyone got on smartphones, we had that same, like, you know, people used to struggle with directions, right? Oh, yeah. Right.

Think about like, and even when you had on. online maps, you still have to go to, you know, MapQuest and print it out and do that. Yes. Like now... It was an equalizer. Exactly. But now have you run into any person that like can't get directions right? Like, you don't even think about it anymore.

Yeah, the idea that you would be lost. It's an edge case being lost. Like if you're lost, it's because you're in a third world country or... Even then, right? Like... Yeah, it would be even hard because you would have the GPS location or your phone died. That's actually the reason, yeah. The fear of the phone dying.

Yeah. All right. I give this, gosh, it's either a B plus or an A minus for me because it's a new. The vision is great. I'm going to go ahead and go B plus. I'm going to go B plus. I'm at a minus. You are seeing.

The one thing I was going to hold back, like, you know, the one thing that held people back is like that if then else kind of, you know, scenario or I have FFT, right? And it's just simplified now. You don't have to do it. You specifically tell it, hey, when I get an email from this person, send me a Slack message. and it walks you through the configuration, and that's going to really make things a lot more productive for folks.

You can use it outside of work use cases as well, and I'm very excited by it, the simplicity of it. Fantastic. And you know that Zappir has all kinds of really interesting things. Like they have Typeform, LinkedIn, Instagram. Yeah. You can do all kinds of crazy things. Yeah, I'll actually just go back and let me just pull that back up.

I might have lost it here. Oh, yeah, here it is. I might have lost here. Producer Nick, you use Zapier at all or no? Do you have any Zappi set up right now, producer now? No. So why did you go ahead and do this for the show? Here's some of the examples.

Well, yeah, see if you can do some of these, Nick, for the show where when, I think we, I think I set this up long ago, when it, when we publish a YouTube video, it goes to a Slack room. Like, that's set up, right? Oh, that is set up. Yeah. Okay. I thought you meant, do we have any in terms of, like, booking the show, not with, like,

No, what I would like to do is when the show publishes to automatically do the tweet, do a LinkedIn, you know, whatever, do an Instagram. I don't know what's possible of like just automatically when it goes up on YouTube, have an automated, published, colon, boom. Because then if we forget, it's okay. We just at least got that out, right? Yeah, the problem is when you're uploading to YouTube, you have to go through, I mean, sorry, when you're uploading to Twitter, you have to actually upload the file manually. It doesn't do it via Zapier. I would still do that. But I'm saying, like, have Zapier.

at least when the YouTube video goes up, do a tweet, do a LinkedIn. It doesn't have to be the full thing. It's just like a little quick signal that it's up and running. You can still do the fuller one. Twitter not currently supported, yeah. I think that was when...

Oh, is Twitter. Sorry, Twitter is not supported. Yes, you got telegram. Yeah, so they are trying to figure out my understanding with the Twitter situation. Remember, Elon said, you have to pay in order to have access to the API. I think Zapier maybe hasn't paid or doesn't want to pay.

I don't know what the situation is. I think like the men price is quite a lot. It's a good, you know. 50 grand a month or something. Something like, yeah, they don't have the... Does it do it to LinkedIn?

Because I think the last time I tried it was LinkedIn, it wasn't working there either. Some of the social networks were getting annoyed that people were doing too much of this automated posting. So I think that it was frowned upon by some folks. But yeah, look, you're saying, When This Weekend Startups publishes on YouTube, send a post on to the This Weekend Startups

LinkedIn account. Yeah, I just was copying and pacing, so I got the text wrong. Yeah, it looks like it does LinkedIn. Yeah, yeah, so it's great. We each have to connect it. Great. Awesome. All right, yeah, I'm going to go B Plus.

Pretty great. Okay. Awesome. All right. Well, what a great day. And then next week we'll do GROC because I got access to GROC. You got, did you get access yet or no? Yeah, I just got it when we were sitting here.

Yeah, so I just got it. It's pretty solid at first. glance. Yeah. I would say it feels like it's maybe six months behind. Something to that kind of thing. What's crazy is, look, they did it in like, you know, less than 90 days, which... So that's impressive.

Yeah. Yes. And it has the entire Twitter corpus. So I think what my gut's telling me, because I did some searches, okay. I think for real time, it's going to be the best real time ever made. Okay. Because you'll be able to do something like, uh, tell me about, the

Lakers game tonight and it'll be like yes you know LeBron had an rolled his ankle

left at this time you know Anthony Davis had 40 points and this person turned the ball over seven times so you'll get like that

in real time okay to be really interesting and then also it let you have multiple chat windows open at once yeah I saw that

which was kind of cool and then you can also see the branch yeah explain to people what the branch thing conceptually that is with an example please well um yeah we'll do it like live next week but i think the you know the concept is you start a chat and then you want to basically keep that main chat

going and you want to take you know you want to branch off that chat because you have like sort of another idea so let's just say you're talking about um you know this week's basketball game and it's it's going you know you're having some back and forth but then you're like oh i want to you know talk a little bit more about LeBron because I'm curious about X, Y, and Z, you can kind of jump out from there into a chat related to LeBron. That's the concept that they're going after, which I think is a unique UI concept. They have some definitely really different UI concepts, which I think is good for this chat era. It's also a little bit rebellious, so I think they're going for a little bit cheeky. And so I had it on rebellious mode. It was kind of being funny the whole time.

Yeah, which was fun. You know, like, I think it's going to have a playful nature to it. And listen, Elon entering the race with, you know, all these other players is only going to up everybody's game. And he's going to push people to go even fast. I mean, we're already going super fast, right? But now I think I'm very, I think overall for the ecosystem, it's good to show that in less than, you know, kind of 90 days. And you can only spend so much money in that amount of time as well that they were able to pull this off. And that's going to, you know, force people to reset. Like, think about what you've got to be thinking if you're at Google right now.

They haven't even released, you know, sort of their latest and greatest yet. So it's going to push everyone on for you. I think Sundar could take a note, which is if you saw Sam out there today with Satya, Nadella, you need to get out there and start having the leads of every single division, Google flights, Gmail, docs, YouTube, every leader on January 15th needs to get out there and do the do their demo of what they accomplished and if you didn't accomplish anything

then maybe somebody else needs to run YouTube maybe somebody else needs to run Gmail or somebody else needs to run Chrome like the fact that Brave is releasing stuff and Chrome isn't yeah tells you like even the email thing like that should be in Gmail right no you know like kind of

maybe they should just go buy Mindy or something but like that's the I mean they just need to keep up and I think this is where like it's a great organizational case study. How paranoid Sundar is about this is going to speak volumes.

He should be super paranoid and I would be lighting a fire on every division head under their feet. Let's go. We're going to do a demo. Every single Google product is going to demo their AI stuff on January 15th.

Good luck. 100 days and counting. I would just tell everybody Like 100 days from now, we're doing this. You're going to have to demo what's new on YouTube. You know? That's like that's Jan 6 pretty much. Yeah, Jan 15th, you like I said.

That's 100 days from now. 100 days, right, whatever it is. Like, it's, let's see, it's November, December, January, February, wait, December, January, February, February, yeah, like March 1st, we're going out there for South by Southwest. It's all. Let's go. Rock and roll.

The end. Okay. There's been another amazing episode of This Week and Startow's with Sandeep from definitive intelligence. If your company needs help with AI, he's your guy. Yep. He's Sundeepe on Twitter, right? You have the first name club?

There you go. Yes, I'm the first name club. All right. Follow at Sundeepe, and we'll see you all next time on this week in startups.