Introduction

So what I'm gonna try and do over the next 20 minutes is, and I want this to be extremely engaging from all of you. This is not me just putting a pre-built demo in place. I want to actually live build an app. in the next 20 minutes.

And I'll be able to guide it a little bit to what is possible and what is not possible. But any ideas? Sorry?

It always comes up at like one, two, or three. Any other ideas?

Checking the weather. That could be another. Checking the weather, okay.

Creating a wedding speech. Interesting. One more idea. That is a bit more complex.

Sorry, what were you saying? Creating a recipe.

Brainstorming App Ideas

Okay, so we have a few ideas, let me see. We could go for, who wants to build the wedding speech creator?

Okay, recipe creator. Okay, I kind of almost want to do the dreams one, but it's a little bit, I'm like, yeah, I'm going to push that.

So, let's go for the recipe creator in this particular thing, right? So, I hope, yeah, okay, recipe.

You can all see this, right? Yeah, okay.

Gathering Requirements

What will the recipe creator do? What should it do? What are the functionalities? What do you want to do with this app?

This, I mean, we're live building an app, but basically this starting point is just a brainstorm, right? So you might have done this in companies all the time, like whatever the project you're working on, a kickoff, a go-to-market, an idea you're having and you're around a table, often in company, corporate context, there would be one person that is the scribe, or at least is tasked to make sure there are notes.

You might actually use post-its and then someone's job is to stay around and like note them down, figure out, create a document, right? What I'm going to do now is just dramatically more simple.

So I'm going to take a picture of our mind map and then I'm going to close this and I am going to go to ChatGPT. All I'm going to do is click upload. As you see, I'm just selecting the screenshot.

And I'm saying, imagine you're an experienced cooking teacher. Let's try and do that. I am trying to build a recipe builder app. attached is a brainstorm of the main features. Please distill this into a document for me.

Complete what I've missed. So what I'm doing here, as you will have seen, my handwriting was entirely terrible. But you can see that it is fully recognizing everything that I even put on my brainstorm here.

Input and Preferences

So user input, how skilled are you? How hungry are you? Ask what ingredients you have. Match recipes based on input. Effort and time investment is part of it. Occasion-based recipe suggestions. Suggested for special occasions, birthdays, anniversaries, and so on. User preferences, country.

Yeah, we mentioned country. So I guess it didn't quite make sense that I put that under use. And so it detected that. And so what it actually looked for here is user preferences. So it's not only recognized everything we said, it's actually reclassifying it in a way that makes more sense than I classified it.

Just a starting point.

Now, I'm going to say now, Actually, let's do it like this.

Market Research and Positioning

Now imagine you're an experienced market researcher. Give me a search query that provides you with competitive data so we can figure out positioning of this app.

Very simple. So we're building a recipe builder. I don't want it to be a recipe builder that everyone else is building or already exists.

So let's figure out what goes through here. Interesting.

So it is doing a Google query, which makes sense because that would be the thing. But I actually want to use perplexity in this case. So provide me with a single query, no or.

Okay. Now I'm gonna go to perplexity.

How many of you know about perplexity already? Not quite a third of the room, so definitely one you want to know about.

So Perplexity is a search engine, but it uses large language models under the hood. So it's like a chat bot, but that actually searches the internet and then gives you sources for everything that it answers.

So you can actually look at where do the answers come from versus just blindly trusting what ChatGPT does. So here, I'm going to hit Search. And it's real-time searching, going through different AI-powered recipe builder apps that have ingredient matching and step-by-step guides.

Okay. Now you can see here that it's got some sources. So it went through 12 different apps.

I can actually look at everything that it used and that it used to inform its results. I can go to each one of these if I wanted to. So I can look at the initial source.

But in this case, I am just going to take this entire answer. Actually, I can just copy this. I'm going to go back to ChatGPT.

And I'm going to say this is the result of the... Research. Now give me a spin on this app that makes it stand out.

Chef Expert. There we go. All in one AI recipe nutrition wellness companion.

So ChefExpert isn't just a recipe generator, it's an AI powered kitchen and wellness assistant that focuses on both cooking and holistic wellbeing. Okay, there we go. Combining wellbeing and cooking is not quite a positioning that has been used yet.

Now, this would take quite a while to build, to be very clear.

Developing the Minimum Viable Product (MVP)

So what I'm going to do is that now imagine you're an experienced product manager. Please provide me with an MVP spec for this product. Make it as minimum as possible.

Okay, the MVP goal, deliver an AI-powered recipe builder that personalizes a recipe based on preferences and available ingredients while offering step-by-step cooking guidance. MVP focuses on providing intuitive, easy-to-use experience that attracts early adopters, okay.

Now, I wonder how it would do the recipe building. So I imagine the recipe building itself probably should use, OpenAI's API. I'm just going to use that because otherwise we'd have to build an algorithm that somehow figures out which recipes go into an actual thing.

So, describe this in five sentences. And I'm gonna say, do not focus on the tech stack. So I'm gonna try and get this distilled in very crisp definition. It's not going to be perfect, but here we go.

ChefExpels MVP will leverage OpenAI as API to generate personalized recipes based on user inputs, like available ingredients, strategy preference, and so on.

Using Replit Agent

Now, how many of you have heard of Replit Agent? Just one, two, three. Okay, there we go.

Replit Agent, so Replit, who have a product that they call This is a coding platform. It used to be a coding environment for engineers where you basically could have all-in-one and you could both code and deploy what you were doing.

Now you can use this platform with natural language. So with the Replit Agent, I am just pasting the definition of the MVP that we just created, right?

So Chef Expert. I'm going to hit Start Building. This is going to take a little bit. So I'll walk you through quickly what's going to happen.

Basically what it's doing is trying to understand the specs. It's going to come back to me and suggest what features I might want to include or not, or if I want to expand. And then I'm going to let it run in the background, because it's actually going to start writing the code for this app for me.

So here you go. Okay, I've created a plan. Personalized recipe generator.

Don't need to look at the rest. So would you like any of these additional features? Implement user authentication, add recipe rating and reviews, integrate a grocery delivery service.

Wow. I mean, that would be great, but I'm not going to select that here because I'm going to be stuck in a loop if I do that right now. It's going to take more than the next 10 minutes.

So I'm going to hit approve. And what you'll see now, Let's see. Let's see if it starts. There we go.

It's now live building the app for me. And what it's doing, it's not only building the app across multiple files. It is going to go through its own checks. It's going to then start the app.

It's going to create screenshots of the app to figure out. It's then going to compare the screenshot it takes of the app it builds against the specs that I gave it at the start in order to figure out is anything missing. And it will get in a loop until it realizes that it has built the thing that I asked it to build in the first place.

Creating a Brief and Presentation

But we, whilst that is happening, are going to go back to ChatGPT. Wait. What did I miss? I think I might have deleted the app before.

So let's see. This is a different window. So I'm just going to go back here.

Recipe builder features. I am going to take this and I'm now going to create a new chat. Actually, I can use the same chat, I think, because I can just switch the model. Here we go.

How many of you have used Canvas with ChatGPT already? Not that many.

So what I want to do now is I am part of a restaurant company. I want, write me a brief to pitch this idea to my exec team.

Now you can see ChatGPD has understood what this is about. Actually, it's actually very, Very short, make it more complete.

So it's understood that this is now a document that I might want to use. And so it's used this canvas mode, which basically gives you the document view on the right. But it also means that you can start to interact with each element here.

So I'm excited to share an idea for the new project Chef Expert. I'm going to take this first paragraph. I'm going to say, make this much more extravagant.

It's going to look at that particular paragraph and it's now just going to rewrite that particular bit in the style that I'm now asking it. This would be how I would go back and forth with a document that I'd write now, where until you're kind of done with a particular section, you think it's covering everything, you can just go back and forth and then you can export this in the way that you want. Go back into Word, do whatever you want with it.

Launching Notebook LM

I am going to take this brief And now, I mean, it was mentioned before, but how many of you have tried Notebook LM?

Okay, so another thing I'm going to launch in the background is Notebook LM. I am going to first paste the text of the brief. I'm just going to give it to Notebook LM, and then I'm going to go back to Perplexity, and I am going to take...

I'm going to take the two sources that were at the top, nothing more. I'm just going to copy paste the link. Let me see here.

I'm going to paste this link in here. And then the second one is also getting pasted in here. And then I'm going to hit generate.

And we're going to wait another bit. I'm going to finish on that in a second. Whilst that is loading, I'm going to go back to what we have here.

And instead of the brief, I actually want to outline this in a presentation that I might have for my exec team later. So now provide me with a presentation outline to present this in a PowerPoint. In this case, it's not using the canvas, so I can actually remove that.

I am going to take this. And again, I'm just copy-pasting this into one more tool. And I know this is a whirlwind tour of a whole bunch of apps.

How many of you have used Gamma before? I mean, I'm introducing you to a bunch of apps tonight. So here, Gamma is a presentation tool.

I can hit Create. You have just a paste text. I'm not doing anything but pasting the text I got from ChatGPT.

I'm going to hit continue. I want my text to be brief on the slide. So I'm one more like I want to be able to talk to a slide rather than having everything on the slide.

So I'm going to select that. The audience here, potential investors, venture capitalists, and tech industry leaders, fine. I'm going to use US English.

And for images, I'm just going to leave this as it is. And I saw here that we have 12 slides that ChatGPT suggested, so I'm going to make 12 cards. I'm going to hit Continue.

I can select a theme. I'm going to make it a little bit colorful. Hit Generate.

And then we will see the presentation that comes out. It is now live building a presentation for me based on the presentation plan that ChatGPT gave me. In parallel, it is building the images that go along with each of these slides.

And you can see it'll get to 12 slides soon. It's now at slide number seven. All of this is editable.

All of this is exportable to PowerPoint if I wanted to. And I can do anything I want with it. I selected a template here that was colorful.

I actually have a Mindstone template as well, where you can just upload a presentation from your company and basically it'll automatically figure out what the right colors are and things like that. And so you can apply your own color scheme to it. But let's go see here.

ChefExpel, AI-powered culinary revolution. Introducing ChefXpert, AI-powered, luxurious, and intuitive. The market opportunity, 43% increase in personalized meal solutions market.

The health consciousness, 68% of consumers prioritize healthy eating. I guess all of this came from the perplexity search, right? Because we actually launched that before.

AI adoption, AI in food tech projected to reach $29 billion by 2026. Wow, I did not know that. Maybe we should build this thing.

Providing Background Material

Now, My exec team has no idea of this market. So what I really want is I want them to have some background material that they can go through.

And I know that they do not really look at presentations. So I'm going to flip back. to notebook.lm.

This is, I was really hoping it would be ready at this point.

Live Demo of the App

Let's flip back actually in the meantime to the Repl.Agent as we go through. I knew this was gonna happen because it's asking me for the OpenAI key. So the Repl.Agent has gotten to the point, I told it we were gonna use OpenAI to inform the recipe that was gonna get built.

So it needs an OpenAI key, which I had not actually planned to go and create in this demo to be clear. But I can go and do that, so let me quickly create one, API keys, open AI, create a new key, demo recipe builder.

I'm doing this on purpose extremely fast, so apologies for those that are not technical, but I am literally just logging in, creating key, copying key, pasting key, right? There's nothing technical to it, I'm just doing it really fast. So now the ReptileAgent is going to continue.

It's got the OpenAI key. I can flip back to notebook. And here we have some background material on the app we're building.

Let me switch the audio feed to what we have here. Or is it all just hype? It is interesting to see this technology, AI, trying to come into the kitchen, which has always been such a personal and experience-driven space.

Yeah, exactly. And we've got some really interesting articles here that dive into all of that. We're going to be looking at some of the big names out there like Chef GPT, Dishkin, even Plant Jammer.

We'll even be looking a bit at the future of all of this. Where is it all going? Well, before we get ahead of ourselves and start talking about flying cars in the kitchen, we should probably talk about what these apps actually do.

I mean, at their core, they're using AI to analyze tons of recipes, looking for patterns and connections in the ingredients, right? So like a digital sous chef. But instead of tasting for seasoning, it's all about the data.

Yeah, exactly. And they use all of that analysis to generate recipe ideas that there are some genuinely complex processes happening behind the scenes. It's not enough for these apps to just recognize chicken or broccoli.

They need to actually understand how those ingredients interact. You pick one and the app, because it's synced up to your smart kitchen, automatically starts preheating the oven. no more so what we're listening here is we took these two first sources we got from perplexity right and we combined it with the brief that we had created from for our own app just put it into this tool and it's created a podcast that is now diving into the research that we had from those two sources i could have added dramatically more information i could have added industry reports the first time i used this tool and this is only about three weeks old

First time I used it, I took two years' worth of future of work industry reports, threw it at the tool, and I got a 20-minute podcast that was referencing across these different reports and giving me an overview of how the future of work predictions have changed in the last two years. So a pretty dramatic change. And here, look at this.

In the background, the RepliAgent has finished here. So is the Flask application running and accessible? Can you see the homepage with the options register and login?

Yes. Okay, so let's figure out if this works. If I hit register,

Dietary preferences. I am a vegetarian. Cooking level.

I can't even write the word. Okay, wait. Did I actually got my... I think I might have gotten something wrong on the...

I'm going to do that again. Oh, wait, no, my username was Joshua. It was actually looking for my, that was why.

There we go. My, I mean, okay, wait, I mean, this, I'm actually interested. It might have actually worked.

Wait. Available ingredients. What ingredients do we have available?

Carrots. Chicken. Tomatoes.

What else? Curry. Eggs.

Sorry? One more. Make it weird.

Chocolate. Okay. Hunger level, I'm really hungry.

There we go. Generate recipe. Oh, OK.

Let's see. Getting an internal server error when trying to generate the recipe. It would have worked if I had selected light, right?

And there we go. So I'm going to give this one iteration go as I answer questions, because sometimes it might take two or three tries when you get there. But you could see that we did build an app that got there that is able to have a registration system that got my ingredients, and then it's just that last step that it didn't finish with.

It has figured out, by the way, where the error is coming from. Now, I know that because I'm an engineer by background, but I don't know if it's going to get the change right, but I do know it's in the right place.

Q&A and Final Thoughts

Any questions at the moment? Yes. How do you choose from all the noise?

Well, we run the biggest AI community in the world, so that is partially how we do that. There's also a really good WhatsApp group from all the speakers we've had from all over the world that do that as well. So that's partially where we get it from.

Is there any kind of risk? So for example, I work at a corporate, I have a big client, I have to make lots of presentations every single day, but I'm handling their sensitive data. So is there any concerns with me doing that?

So there are two things. One is different companies have entirely different policies on this. So there is the policy answer, and there is the what is the contract answer.

The contracts with, if you have a commercial license, sorry, if you have a paid license of Gamma, it doesn't train on your data. But that doesn't mean that your company allows you to use it. Now, Copilot does the same thing, actually.

Copilot in general, decent, but where it's actually useful is it does the same thing for PowerPoint. So if you have a Copilot license for PowerPoint, you could do the same thing using Copilot for PowerPoint.

And there, maybe you can go around the policy, or maybe that's allowed within the company. I'm going to try one more time here. So we had tomatoes.

We had rice, bread, chicken, curry. camembert, chocolate. I know.

Well, I mean, it's in my fridge, right? So it doesn't mean that I'm eating it, but it's in my fridge. Okay, let's see.

Oh, interesting. Wait, it's actually, it did save it last time. I didn't see that.

This was what we had, but I guess it just came back. I'm going to try again. I'm light in this case.

So I clicked generate and let's figure out if it comes back. It didn't because it stripped that out of the MVP. Yeah.

I'm just going to try one last thing, which is what happens if I click on this. Okay.

So I'm going to stop there with the demo. And we've got Frank who's going to take us away.

But you can see how much we just got done in, well, we didn't end up doing 25 minutes. But the hope, it gives you an idea of just how much is possible today.