combining their rich interactive visuals with the power to chat. Let's start with Corsera. Let's say I don't spend enough time thinking about machine learning at work and I want to learn more. I can ask the Corsera app in chat to help me learn more about this. I can say Corsera, can you teach me something? — See, and this is where it's interesting. — And if you don't ask Corsera, — is it going to recommend Corsera and what does it recommend? — Yeah. Well, they have a learn feature on chatbt that I've been playing around with. — You played around with um what's the other new feature they've got where it's like daily news articles. — No, I haven't checked. — Yeah. — Oh, on pulse. — Yeah. No, I haven't got that on my one yet, I don't think. — It can support anything that you can render on the web, like this video showing here. — This is interesting because we've had quite a few questions from clients even on the sporting stuff to be able to pull up footage — like put up game footage. So you can talk about obviously what's happening in a game and then pull it up. — That's the interace is always tricky. — Wait, so he logged into Corsera there. He has a course that he already has access to. — Yeah. — And now he's using Chat GBT2. So Corsera the app is allowing the material in that to be used as context. So it's asking questions about it. Can you explain more about what they're saying right now? So it's like outsourcing the AI feature of course instead of course having to build their own in AI chip. — So they're building the platform essentially where you can go one centralized place. — But why would I what's the advantage to Corsera in this more — distribution — I guess. — No but they still have to go and sign up unless you type in hey I need a course on this and it's going to recommend Corsera because they have an app on it. — It's like why would — Yeah. I mean, that's going to be — they probably want them to be on platform. Well, they've already got an account, so they're happy with that. They've already paid for the course, but then they're going off platform to consume it. Seems like a weird way to — And so what you what are you saying? It's like why are they keeping them on? Why would — call Sarah go to all the effort of building out an app, you're going to be giving all of the course material as context, like giving all of your course material to OpenAI. — I mean, they could say, "Oh, we're not going to train on the information, but it's can you explain more? " or it's obviously going to have all the knowledge of the — of the video there. — How's it pulling that information out if it's in video format? — It's either using a uh like Google Gemini is really good. If you go on a Gemini, you can upload like a 10 20 minute video and then ask your questions over the video and using their like — Sure. — I guess it's cuz there's this is multimodal and it's able to like actually ask questions over the video content. — Um — or it's just doing a transcript, but I'd say it's taking the visuals in as well. So that's pretty. So I guess that they're using the power of um OpenAI's models for that. — But it just seems weird that you would — as Corsera put it out. — Yeah. — Well, it's so what's the advantage of any business essentially going to open a I mean if you look at it through the lens of Open AI — is trying to I mean this is the Apple App Store moment, right? which is essentially what this is and why essentially your apps loaded onto app store is because obviously Apple's the go-to device that people are going to access its distribution and they're betting on the same thing being the case for open AI. — It's feeling a bit GPTish to me. — It's feeling GPish, — you know, as well like yeah, cool. But yeah, what's the edge here? I'm super excited about how learning with Chad GBT, one of our top use cases, is continuing to evangel. And with app and the ass SDK, you can unlock richer educational experiences for users around the world. So here, chatbt responded and explained that the instructor is to talk about data preparation steps before training machine learning model. And then it breaks it down into book turning for me. I don't need to explain what I'm seeing in the video. Chat sees it right away. Users also love being creative in chatb here. I have a conversation where I've been brainstorming some ideas to help my younger siblings start walking business. We've gone back and forth a few times and now it's time to make this into a reality. I'm going to ask Canva to turn that into a poster. I can say Canva, can you make me a poster with the Walk This Wag game? I want it to be colorful. — This one makes a bit more sense, — right? And I prefer — like [ __ ] they're still paying a subscription, — you know. Yeah, — I don't care how they get access to it necessarily. — Cool. Send that off. And now in the background, Canva is generating posters based on the context for my conversation. Canva is great at creating assets like this. And now you can kick it off directly from chat. Whether you're making professional marketing assets for OpenAI or just a fun demo for dev day, — it's smart because now — OpenAI doesn't need to like stress about having like the best everything because you get the like spiky best like designers on Canva and like each different platform or app have their own strengths. — As you can see, this is a live — could be an opt and try DFDs for companies integration. — Oh, this should be basic enough. HTML — Corsera video as well as chatd kind of explaining what it's done for us. But we can explore another modality in apps SDK which is full screen. I click an asset and the app requests full screen. We're able to focus on a specific asset and from here I can see it in more detail. I can ask capture to request changes, maybe visual tweaks just like with our image generation experience. But since we're in San Francisco and it's dev day, let's ask Canva to convert this into a pitch deck. I can say, "Canva, can you please make this poster into a pitch deck? We're trying to raise a seed round for dog walking. — Yeah, I wonder what the monetization is structure is for these. Is Canva given a cut over to Is Openai taking a cut on these apps coming through? Where's the purchase taking place? It's obviously taking place natively inside of Canva. — Yeah. — So, how are they clipping the ticket on that? — Um, — I mean, depends if they're like referring people. — Yeah. Like what kind of deals do they have off the back end? I don't not sure of the monetization. I wonder what — search it up. — Damn, man. This is going to like — the amount of creativity people are going to be have when using chat GBT. app to say, "Please show me some homes for sale there. " — Oh, he's chaining them together. Now he's using the Zillow one. — He's talking to Zillow to fetch the latest housing data and we'll get a — map embedded in chat. — We'll explore how the full screen experience there goes. — So, we have our map, gorgeous loading state, and boom. — Wow, look at that. It also looks like our slide decks are done. So, we'll go back to those in a moment. But this map is a little hard to see in the inline view. So, I can click a specific home and open it. — Holy smokes, it's a full like web browser. — Most of the Zillow experience embedded int. — So, you can basically have a c like a embedded version of your software, your web app — and chat GPT. Bro, they're going to just suck all the eyeballs onto their platform, man. — Of course. — CHP will talk to Zillow again. And because the app is in full screen, it can now. — Yeah, we'll share. So, they don't have any details on monetization yet. I'd be really interested to see what they do that way. — Well, if there's like a like an apps gallery. — Well, this was the whole reason with plugin uh plugins, right? And GPTs. It's just the monetization to it. — Um, — cool. Let's zoom in and find a specific homie. Maybe — this literally like replaces — the browser. — The browser like it knows what I'm looking at. I can — I mean another thing is data. Who's getting that data? Is Zillow getting that data? — What the search data and stuff? Yeah, I must be feeding back to them. Yeah. — Or what's the limitations on Yeah. third party developers — at disposal like search. So it's able to give more information about this home. That UI is getting a bit busy though, — like layers on layers. — Let's check back in on the slide decks. So, if I hop back over into this conversation, see Canva has given us a few options here. Wipe the look of this blue. So, if we open that up, we now see the slides in full screen and I can see all the beautiful slides that Canva has generated for me. When I'm ready, — [ __ ] that's pretty crazy. — So, there it is. The magic of apps from chatbt. Conversations that you manage the intelligence of chatbt with your favorite products resulting in truly novel experiences. — It's literally AIOS. It's like an AI operating system. — Can't wait to see what you do. — Voice mode and payments on top of that. — Talking to Chad GBT. I swap Siri out for JT GBT. I — just a bunch of people at the same time. So, you did very well. once it can make payments for you and stuff. — The apps that you saw in the demos along with a few more from these launch partners, they'll be available in chatb today and this is just the beginning. We're going to roll out more apps from partners in the weeks ahead. — For developers, the access available in preview to start building with today. — Our goal is to get in your hands early, hear your feedback, build it together with you, — and then later this year developers will be able to submit apps for review and publication. — Okay, so they're on a slow roll out. Typical. — Typical. — Yeah. — Interesting one will be when you get the voice mode and it's like has a visual supporting so you can just kind of talk to it. It might be like up on your thing. — Yeah. — And it's updating the like — showing you things on screen. — So apps and we go everyone loves it.
will pass it off to Christina. — Today I want to show you how agent kit helps developers agents faster than ever before. This the great thing is if you're an agency, you're like — just ride whatever. — We can get into this in a bit, but this is perfect for agencies. Absolutely perfect. — You navigate your day and put you to the sessions that are most relevant to you. We're open AI. We need to have AI in our yday website. So that's what we're going to build together. An agent powered by agent hit deployed right here inside this pipe. And to make this interesting, I'm going to give myself 8 minutes to build and ship an agent right here in front of you. And I'm going to start the clock now to keep me honest. So I'm starting in the workflow builder in the open air platform. And instead of starting with code, we can actually wire nodes up visually. — I mean, that already looks better than — than half the [ __ ] solutions that are out there. Yeah. — For example, tools like MCP guard rail loop and other module nodes — today. human in the loop as well — flow that uses two specialized agents. The first will be a sessions agent which will return information about the schedule and the second will be a more generic dev day information agent. So I'm starting off with a categorizing agent. — So funny that remember when this used to be like swarms like agents. — Yeah. — I wonder if they've integrated much of that — know behavior based on that classifier. Next, I'll create a session agent. Here, I'll drag and drop um an agent node. I'll call this session agent. I'll give it the context um about kind of grabbing information about a session. And then I can add in various tools here to — Okay, so file search. Yeah. — A do with all the information about — I mean this nothing like that revolutionary, is it? — And attach it. So, this agent now have all the information needed to answer my questions. Um, but showing the schedule should also be fun and visually interesting, not just plain text. So, I'll also create a widget for them. I'll head over to our widget builder. Um, here I could create a widget from scratch. — We could create it just by far out — gallery — and we use them — purchase widgets. — Haha. So, they don't purchase — in the chat — and they don't buy widgets. [ __ ] Can you just create a new one here? — Yeah, new widget. — One on boarding session. um in Golden Gate Park. So, we can simply download this and then head back over to our uh — I don't see how she's going to bring this home in a few minutes. — So, outputs it to a widget. — I don't think I clicked download. So, let me go back and actually click the button download. Head over and attach it as output format for the sessions agent that we just created. Drop that in. We can preview it to make sure we added in the right widget and everything looks ready to go. So this person agent is now done. Um — where's that display agent? — Display inside the agent. — Yeah, but it depends where they're calling it from. — Like there needs to be an input. — Once again, — it must be interacting with it through what's the interface that the user's coming in through — just to make it really on brand with the day. Um we'll adding a file once again. So, we have a file with all of the information about the day. Call this step day. Attach it. This agent is ready to go with — I'm sorry, babe. This is not a great demo. — It looks like I have a couple more minutes. So, let's add in some additional security with one of the pre-built arills. So, what is the most important thing in building? — I mean, I suppose it doesn't seem that I'll turn — revolutionary, but it's a good base for them to build here. So they're making pass. Yes, they got some easy just like quality of life things. Simple. Yeah. — Making a lot easier for you to run a lot of these — an additional agent to handle pieces when um — when it fail. Okay. So it's just a simple pass fail. — So again, I'll make it speak in the style of froze to stay consistent and I'll remind it that um it cannot open questions but contain sensitive information and remove kind of the context. Um great. So I think this workflow is ready to go. I can also configure um the output to determine what shows up to the end user. In this case, I can also turn off um file search sources if that is kind of more internal. Um and I think that's it. Let's test it out. — So if I ask — this directly from um our agent builder. So here I can ask what session to attend to learn about building agents. And I can see this message moving its way through that workflow we just created. checking guard rail, categorizing intent, pulling information from the file of sessions that I just added in, um, finding the right session, using the widget that I added, um, and determining, you know, orchestrating agents at scale LM15 with James and Rohan is like the best session for me to go to learn more about this. Um, and then I see a couple rivets because this is actually Fro talking to me and riveting at me. So, okay, I think this agent looks good. — Yeah, Ste's all a bit all over the place. specialized agents using tools. We added in guardrails. We customized them using some widgets and then we also tapped out the work — agents SDK on the right. We also have codeex score in case I want to run this in my own environment in my own servers. But you can see this is quite a bit of code to write and so I'm just going to stick with using the workflow ID that we just created and then head over to my site. So here in my dub day site, I'm first going to create a chat session using um the workflow that we just created. I'll simply drop in. — Okay, simple. — Here we go. — The chat react component using that client secret that we just created in our own server. — Where's she snagging all these copy paste from? Or is she like command zing it or something? — Really themed. In this case, it's going to be called ask froge. It's going to continue to rivet in the placeholder. Um and it'll have some fro specific colors and starter box. So, where's she programming the — It's going to be on a website. — Yeah, but where's she selecting the brand and color pallets and everything? — Ah, she's obviously done it all before. — Yeah, she's just like assembling a website, — but it's all code based. — Oh, it's that one. Yeah. — Approach top of the site. Let's try it out. So, what something to — send to learn about building ages? And again, this is running through the exact same workflow we just created. Checking for guard rails, categorizing the message, pulling from tools. — I don't know if you can how customizable that is — and then again deciding orchestrating — um — but look like what happens if you send another message now? — Yeah. — Yeah, I get what you're saying. Can you chain them together essentially? — Yeah. Well, you must be able to, but it's like glorified mini chat, you know, it's like what mini chat could have done on this thing directly in the um in the visual builder and also deploy these changes directly to my site without making any code changes at all. This includes adding new tools um adding new widget. — That's pretty dope. — Adding new guardrail — just like it on there and yeah, test it around with it then publish it. So, in just a few minutes, we've designed an agent workflow visually. We added in some tools and widgets. We previewed it. We deployed it. We tested it. And now you all can use it. This is actually live now in your Devday site. Um, you can tap your badge and you should be able to see it and use it and um find the sessions that are about Scrio. So, we're looking forward to using it and meeting Fro and also seeing all the new experiences that you'll now be able to build using ignit. Thanks. And back to Jan. — Yeah. Pause that. — It's mini chat. — It's like a literally cuz the mini chat is the same thing. You have the entry point. — Yeah. — And it goes back around and loops. — Yeah. — Um it's probably what Mini should have done a long time ago. But I have do have to give them credit for stitching together all of these little pieces like the file search, the code interpreter. — Um allowing you to easily cuz it's essentially it's just like sticking a bunch of little uh like chat GPT nodes together on something like make. com, but it also is allowing tool calling which you could technically still do previously. — Um but it's just in a nicer way of doing it. But what I'm already thinking versus like is this N810 the N810 killer or like uh I mean they've obviously been um ahead of the charge here. You've got voice flow as well um with a similar product but the triggers is probably the if they start adding in triggers here like you've got in Zapia you've got make. com you've got nad — and they have the trigger aspect as well because right now it's just start and that's only from a chatbased input. So once you can start building these and um that are based off triggers like on all these other automation platforms then you've really got a trouble on your hands for these guys. But the tricky thing the bullcase v8 still is going to be around them being model agnostic and not being these guys obviously can't have like they have to rely on their own tech for everything right. So I mean a lot of the time we find with morning we just stick within one ecosystem. if these guys are the best. Say some of these new projects we're taking on. Um I could already see us being able to do if we can do um natural language to SQL query and like chat to database through this because we were talking about multi- aent solutions for that. — Y — then you probably wouldn't need to GPT5 is already really good. do you need to go outside the ecosystem and a lot of those little quality of life features like guard rails um that personal personally identifiable information — um those quality of life things that are really going to help speed up the dev process but — I definitely want to get in there and actually hook it up to some real tools and real APIs rather than just like oh a little file search — um cuz it's the access and hooking into your systems that really matters and they haven't really shown that. How big do you think the white labeling of it is in terms of being able to customize that UI? — Um, it looks like they had a live like a library on there. — Um, where you can basically scab other people's work, which is I guess is really big. They're handling all of the little tricky bits like the output format from the agent like, hey, you know what? You know how this widget works. — Someone's created it or I've created it. And they didn't really show the creation process for the widget, but maybe that's like just vibe coding up this dope widget. — But that's what it looked like. And there's a chat that said create your widget here. You can say I needed to have an image and then a title and then — which is where the unlock start right and I think that's been the big question about how you going to build the interface into these chat models and that's the start of it and if you go you saw even in the widgets they have products in there this is the start of being able to shop within the chat interface — um yeah it'll be interesting let's see where he takes this — one of the most exciting things happening with AI is that we're entering a new era which that changes how software gets written Y — anyone with an idea can build apps for themselves, their families, or their communities. — Listen to this. — Data streamers are helping people reconnect with memories using chatbt, image generation, and Sora. At ASU, med students need a better way to practice the kinds of difficult human conversations they'll have as doctors. So, they built a virtual patient app with our models where they can try, fail, and get better before they step into a real — what? Like, you're going to die. That's ready RNS. — RL. — That's ready RNS. That is — training. — Educate nurses before they go into exams. — I hit it again, — bro. — Where you have a live discussion with art and sculptures with our real time API. — Wow. Revolutionary — history becomes a conversation. — Yeah. No, low key could have liked that on some of those turkey spots. Hey, I was having this like shitty little audio
guide. Yeah. — Chat GBT stuff. I think that's pretty big. — Yeah. — Overall, um they're basically replacing uh the need to use a browser for a lot of stuff. It's just centralizing all your like instant usage. They've got search, they've got like writing, they've got coding, they've got image generation, and now they're just going to bring all your favorite apps into one place. Um so I think it'll be interesting to see who steps up and starts using that. — I think the Canva use case seems great already. Um, and if there's more of those, then you can see people using a lot more. — No, I agree. I think at the end of the day, OpenAI time and time again has shown that they're ideological and not practical and most of the time and they ship things whether it's plugins or GPTs and they ship a vision and then they take usually 6 or 12 months really to be able to execute on it. — Yeah. — Or sometimes they can't uh as in the case is plugins. Uh, but I think it's the whole shift from hey, we're a model provider into the platform and it's the AI operating system essentially. Mhm. — It is the Apple's moment of, hey, this is the app store. Very interested to get in and see how smooth that interface is. And that's the first portion of it. — I think they're paving the way for the voice stuff is really the play here. Once you've got either on your computer screen or on your phone, say you've just got like on a little stand, then you can just talk to you put it in voice mode. And then I'm saying, "Okay, yeah, I've got this idea for a um for a slideshow or like business idea I have. — Um I want it to be about this and this. I'm thinking maybe like this kind of pallet. Can you cook up some designs? And it goes and sort of just showing you in real time. — Yeah. — Canva generating those. And you get to see it and say, I like this, like that, don't like that. And it goes back and is able to send those messages for you. — Yeah. — I think that's what I' I'd really be excited about. And I can't seem to think like I guess there's the like the Zillow use case they had there. Um the Corsera one didn't seem too strong um in my opinion. But overall, I reckon these guys know exactly where they're going with that. Like every time and time again, people are saying that OpenAI's got phenomenal product people. So I wouldn't doubt them after what they've done so far. So on the consumer side, I think they've nailed that and they're ahead of the game as usual. It's interesting thinking about like what happens to Claude really long term like the as a consumer competitor to Chat GBT now. They're already kind of falling behind on that stuff. They've been really strong on like the writing use cases and they've got their artifacts and they've gone really deep down into the coding side of course via the APIs but not necessarily having as much extra functionality. I think they cook there. Uh the agent kit, the boards of Zapia, NAT, make are all going to be having a firm sit down and uh and think about this and figuring out where — I mean this is obviously them trying to get deeper onto that actual implementation of these multi- aent systems. Like we've been watching very closely this trend of increasing headcount of digital employees and these like digital like AI workforces. Um, you're seeing startups like Relevance AI get really deep into combining these agents into work forces. And by the looks of things, it is very dev focused. Not surprisingly, it's a dev day. But, um, it's funny thinking back to 2023 when um, we first started talking about this and how like no code AI automation was the way to people get in. While yes, this is going to have a lot of uh more like code based features that you can strap in, like building extra APIs and tools to connect into it, at the end of the day, it's a no code builder. I think we can take a bit of a victory lap on that one, which is nice to see. The question is in my mind is whether they want to take this really take the battle to Zapia and make an NAN. Um, and it looks like the integrations between these agents and getting them to work together is uh easier than something like in 18 already, but how advanced they're going to get with these like teams all working together with different triggers and entry points and stuff like that. So, it's good initial case and I think yeah, like you said, the interface and those chat widgets is a big step in the right direction cuz that's been missing in the market for a long time. So, I think it's a pretty bold statement on the on that front. it's kind of wiped out like the whole lang chain um and what they were building with their multi- aent frameworks and stuff like that. I'd say they're just biting close on the heels of what lang chain's been building for a while and they've taken elements maybe of voice flow and the way that they've laid out the builder there. You can see a bit of influence there perhaps. But yeah, it's the classic like sit around and wait, see what everyone else does, see what works, what doesn't, and steal all the best features and roll it out to your fire distribution. So, got to give it to [ __ ] I think we could try to roll this out to some of our client projects already over the next couple of weeks. But yeah, if you guys are watching this, let us know what you think down below. — Where are agencies positioned with this? What are you thinking as an AI agency owner? — Generally, this has been the curve for a long time for agencies. We knew if we just sat here like no one's [ __ ] their pants over this as an agency. You're like great metal tools like the curve has been towards easier. You remember when we started start of 2023 — and I was using text da Vinci 03 text completion model didn't have any chat history. You had to like jerryrig this whole chatbot around it. So it remember the chat history pre- chatgpt they fine tuned it into chat. It was like an absolute mess. But I remember building chat bots um very early chat bots based off that to try and replicate chatbt and how hard that was to get results for clients. And you just had to really be like chash was your worst enemy because it would it set the bar of the experience up here and you had to like really struggle just to meet the same benchmark as that — and over time it's just gotten so much easier. the chat Gvt API came out. Um you've got all the tool use, you had the assistance API came out. Similar event to this nuked all these startups that have been rigging basically the same thing there of adding in tool use and prompting and knowledge bases as well. This is just that next step and in that evolution and they've really made it easier for everyone. So now you've got this and the uh they've integrated the responses API into it. It's a great day for agencies uh who are selling AI services and — I think that's an important I think a lot of the time and we see it even inside the accelerator rise a lot of agencies start to panic when this stuff comes out because the barrier to entry obviously gets lower and — at the end of the day I think that is why you position yourself as an AI transformation partner and I think even our shift from development up into transformation partner is proving uh to be serving us extremely well. I think at the end of the day it goes from uh you know purely development focused uh which in this case has been you know it's going to be a hell of a lot easier to okay who's the transformation partner and who are the experts that are really going to help me identify where I can implement this technology that's going to take care of governance — and also companies most of the time aren't going to want to give everything over to an open AI they are going to want to stay agnostic — and they're going to be looking for those middle layer companies to be able to say hey these are the best models this is what you should be doing and they're looking at the experts. You know, as we always say, Facebook ads, you can go on there, any business can go on there and create a Facebook ad in 5 minutes. Why are there agencies out there that make millions and millions of dollars a year? Because they're experts in the domain and they get better results than the company's doing them themselves. — So, I think this is exciting for us, honestly. interesting point here is this trend towards like what I kind of call vibe automation where you've got um Zapia was the first one with that text like prompt to workflow um and now you've got NAT's dropped theirs. I think there's some other startups as well that have integrated it. Maybe make's got it too. But because it's all based on JSON, you can like write in, hey, I want a workflow that does this, this, and this. And then it's able to print out a workflow that's pretty much ready to start testing and iterating on. And that's a big trend we're seeing. So it's probably the master stroke of this would be having that kind of agentic editor where on the side you can say, well, it starts off with a text prompt. I think they've kind of got it in that initial interface that I saw where you could it looks kind of like chatbt where you could put in I'm building this and this and this. It needs to have these functionalities. Um and then it's going to go into kind of a lovable thing where you've got the side panel. It's got the workflow builder there. You're able to chat back and forth. You're able to run tests on it. — Um and so this is all great news for agencies and people looking to get into this as a beginner. The question is how you going to survive in a world where development might be quite commoditized. Being the agency who knows which shots to take when the cost of development drops so much like this and the difficulty there's so many different directions you can take it. Which use cases do you pick first. So being the agency who knows what to do and when is really going to be the edge you bring to client discussions. Big day for us and I think it's a big step in the right direction. So, I'm just glad I'm in this position and not those startups right now. You know, we might see some of you guys here in Sydney for the event we're hosting this weekend in the harbor on a boat. So, you guys will see that on the vlog channel when we drop that. But yeah, catch you guys later.
