Idea to App in Minutes - Developing Fully Functional Apps using AI | Madhusudhan Konda

This talk focuses on the concept of 'wiping coding', a streamlined approach to quickly building MVPs (Minimum Viable Products) using AI capabilities. The speaker, Madu Maduk, introduces various no-code tools and platforms such as Replit and Cursor, which allow developers to create applications without extensive coding. He emphasizes that through AI integration, the development process can be drastically reduced from months to mere minutes, showcasing practical examples like a simple expense claim application created over a weekend. He also highlights the importance of understanding user needs and feedback to enhance product development. Throughout the session, he shares his experiences from working at the European Bank for Reconstruction and Development and discusses the evolving nature of software development jobs driven by AI tools.

Hello everyone. Thank you so much for attending this last impromptu session. I really appreciate your patience and coming over here for this. Um as you can see I am going to be w coding with you. If you don't know what is woding you will know after this session. Okay. If you have been experimenting with wipe coding, I will explain to you the some of the trips tips

and tricks that you may want to experiment and go away and do the things at home. Especially being tomorrow being the Sunday, you have an opportunity to play out the sessions what we are what what I'm trying to do today. So the session is all about taking your concept and idea and building an MVP really very very quickly very very I would say probably in minutes but

it might take a couple of hours to a half a day right I'm just going to show you the applications which I developed and I also want you to get addicted to developing these applications as well it's really really addiction right I'll show you how it is going to be that done obviously you will be interested in who I am right who am Okay. So it's me,

my name is Madu Maduk. I work for a bank called EBRD, European Bank of Reconstruction and Development. Me and my colleague Simon and I think my other colleague who is Jenna who has just joined the bank. We have set up a booth just outside this organization I mean this particular track. So I'll come back to that in a second. So with with respect to myself, I am

a passionate technologist. I enjoy technology. I've been in the industry for probably 25 years I think more than 25 years and I'm a core Java developer, Python developer and recently I picked up AI and I am being I would say probably my hobby is writing books or teaching or training or upskilling myself one of these things. I know it's a boring hobby but unfortunately that's me. Um

there is this is my book the recent one which was published about uh if you haven't seen this is elastic searches in in action. This was um published in two years two years back before the chat GPT. So I did not use chat GPT to write this one. So I can guarantee you this is my 3 years of baby. It took me almost all my weekends and

week nights. So probably that's the reason why you can see the thickness of this one. Luckily, one of you guys are actually going to get this one tonight or tomorrow probably because um outside this this track there's a booth where you can just make sure that you leave your details you know enter into the draw and certainly my colleague Simon is going to post you if you

want I can sign it as well with my name but you know you are going to give win this one okay don't don't forget that with that I have been developing a lot of books but this these are the list of books so far. So with EBRD, this is my passion about technology. This is my first ever company I've been with this long. Ebrid is is so

so dear to me. I've been with EBD for over almost coming into seven years and most of the time I used usually it's you know probably 3 years I get compressent. I don't want to become compressent so I move around like this is the one company which I really enjoy and I'm staying here as well. The main thing with this is the tech at EBRD. So we

have our new MD Sash Chandos who just joined about two two years back. We are transforming we have been transforming in the IT in the business of developing such applications really good applications for delivering for for the bank. I mean this is not a typical bank where you can see about you know just like normal investment bank or a retail bank. It's the social bank. It's a

development bank where you can see the impact of what we do in the uh probably 77 or 80 countries outside the outside the um Europe or including Europe as well. We are modernizing our tech stack. We are actually very much into digital transformation working with ton of applications and hence we are actually trying to do a setup of EBRD office in I think it's mega park if

I'm not mistaken in here so we are actually recruiting actively recruiting lot of technical roles some of them are already live so please do come and talk to me or Jenny or Simon outside this booth we are pretty happy to take it over that's the link. Don't forget to go to the jobs.b.com and you should see you know search for location set set the location as bulgaria

and you should see the jobs which are live today with that. So that's the introduction that's out of the way. I wanted to straight away jump into some of the applications I've been developing. I think let me get that over there. I thought it's extended actually. No. I was extending my mirroring my thingy but looks like it's not. Give you two more seconds. Okay, let me mirror

my screen. Okay, so let me show you one quick application. My son is an 18year-old. He is into something called A levels, which is basically 10 plus two. You after your school, you have your A levels. He and his friends were struggling to actually they are using chat GPT a while but they are struggling to take a picture and snapping the picture and then asking the chat

GBT you know at chatg marking those particular program is not necessarily the right way so I went away I created a a powered learning application it's a very simple assistant all it's going to do is that it's going to snap students answer or question whatever it is is going to start explaining to you it's nothing but a wrapper around your GPT for model and how long it

took for me a weekend seriously a weekend in fact started with at 2 hours on a Saturday and then 2 hours on the weekend so it's really really straightforward you could use it really very well uh on your mobile it's it's a mobile first application but you know you can always use mobile as well as on the web web as well there's also in in in UK

there is individual savings accounts. Unfortunately, not many people know about them including my own son. That's a pain point I wanted to exploit or explore. What I did was I used Replate or I used one of the tools to develop a simple website. You can see that this is looks like a professional website but this is all white coded. I did not do even I did not

write a single line of code. Believe me, me being a Java developer, me being the Python developer, I did not write even a single line of code. Everything what you see here is coded by the AI, right? It's it's because I just took the idea and explore the idea and created an application. It has got the chat GP integration as well like GPT forum model integration. So

you could get answers from the advisor if you want. Okay. Now me from a a corporate sponsor because I think my EBRD usually they I I think about a month back I went to Jitex 25 in Berlin. So one of the things I need to do is that I need to come back and claim for the expenses right that's that's what you do actually when you when

you work for a company you have to come back and claim for your expenses. What I did was it was a manual boring process. I need to take a snap of each and every receipt and then create a Excel sheet add the information to that one convert from euro to GBP is really manual process but that's a painoint that's an opportunity that I could do something right

that's an idea I got so why can't I develop an application that does really very well from you know some simply simply taking a snapshot of a receipt and to the point to the point where it could send of the complete invoice that that's the idea. So this is the claimly is a very simple application. I created over the weekend again it can go away. It does

the job of snapping all your receipts. So if you have a receipt you start create a trip snap your receipts and it's ultimately it's going to create a excel file. It's going to create a report and once you finish a trip everything will be sent to your department or your to your expenses department. It's really easy and again I did not do even a single line of

code here. My all my examples here I wanted to show it to you is AI has become so integral into your thought process that an engineer if you have really if you put your thought process into it you can develop an MVP in really minute right let me show you that one. So these are all the examples I got. I if I show you the examples, you

will be you will be literally amazed. So for example, I got my addiction nowadays is this thing. You can see that how many I keep been keep on creating them. Almost all of them are very very niche applications, right? I could bring them up. I could productionize them. I could start marketing them. But that's not that's not what I'm doing at the moment. Okay, that's a different

discussion. So what how do we do that one? So the idea here is that you have an idea you want to market that idea. When I was developing this MVPs about 5 years back about 10 years back it used to take at least 3 months to take an idea and build a prototype. I don't know how many of you are aware of that but it used to

be taking that long 3 months to 6 months to one year sometimes and by the by the time you finish your 3 months period you already exhausted with that idea you may have moved on right I used to have that problem I used to say I started with really you know so much of enthusiasm about an idea I started spending 3 months 6 months by the time

it comes to the actual prototype it's all gone my enthusiasm is gone somebody has already built that particular idea. I'm completely, you know, moved on. Nothing else. But now I don't need to wait that long. I can certainly go away develop these applications within minutes. Okay. Now I'm just going to introduce to you one or two tools. But these are not the only tools you should be

able to uh you know experiment with. Right? I'm not um I'm not telling you to use only tools. I'm all I'm trying to tell you is that there is a newer wave of coding called wipe coding. There's a newer way of uh taking the applications and dissecting them in such a way that using your AI tools to make them market them to the to your client very

very quickly. That's the idea here. So let's go to let me introduce you to first replet.com. Look at that. So this is a I'm not sure how many of you are aware of replet anyone idea cool couple of them couple of you so this is a very straightforward interface an application a full stack application right so when I say fullstack application what I mean to say is

that I could take away the application which is built on this particular uh uh this particular platform home. I can deploy that to the world. I could integrate it with authentication. I could have a database behind the scenes and everything in all in one platform. In fact, not just replet, there are other applications coming up similar to like this similar to replet as well. You can have

your own application as well. Okay. So, let's let's try it out. Actually, rather than me telling you, let's try it out how I can do that one. So, what do we have? So I think I have a small probably somewhere I don't know exactly but so let's take an example let's say okay let's build I mean I'm remember I'm not writing any code I'm just telling what

I want to write I'm just my idea is to start a a recipe plan okay just let's build a I'm a vegetarian so let's build app okay that's all I'm not I'm not exploring anything at all but ideally What I normally do is that I take the small piece of text that is a prompt as you know I go to deepseec.com or chatjp.com I'll expand that you

can do that one but here I'm not going to explain to you but you know you understand the idea right you can start with this small prompt as well so go ahead and then ask it to do a a plan so all is trying to do the moment is that there's an agent behind it the agent is going to take away your prompt it understands the intent

of the user is going to take away and it sells that okay Mr. cond wants an app for so and so so and so things so for that I'm going to go and create a plan because obviously if you think about it right if I ask you to develop something if you if I ask you to bake a cake what do you do you just don't go

and start doing everything right but you have a plan you you know you have to go go to get d get some x and y and zed from the market right you have a plan that's exactly what this guy is trying to do for me right so you can see that there is a plan that has been created created based on the plan I can say yeah

this looks perfectly fine right it tells you that the initial version is kind of okay to me it tells me that in a you know app being built with react and express remember I did not mention the stack I did not say that you know built in streamllet or built in react or built in any particular language or any particular stack I went still I just said

that just go and do this job for you for me I don't care I really don't worry about don't ask me what I want but you know just go and use your own brains and then come back to me with an answer. So it has managed to do that one. If you want you can add more features as well but for this particular you know let's let's

stop there. So I can I can looks like it's it's already you can see on the preview on the right hand side well I did not give anything at all but it's managed to do that. So it looks like everything is fine. I can go make the changes but in the interest of time I don't need to worry at all. Let's say I just wanted to be

I'm happy with this. I'm approving the plan. It's now trying to build that build that application. Right? So if you see on the left hand side there is an agent which is working for you. This is the agent AI agent which has got the capability of taking your application or building your application from scratch. That's exactly what it's trying to do there. Now everyone must have I

don't know how many of you been experimenting with AI. This is the something this is something which is going to be future for you pretty much in the next one or two years if you think about it. These things are going to be coming into us or is going to be we are going to be exploring with them very very much. I think probably if I'm not

mistaken the software industry or the most of the disruptions are going to be around the engineering communities right is is certainly going to be coming to that nature. So you can see that actually it's managed to do that and still doing the options. So that's perfectly fine. Okay, I'm I'm happy with that. Let it do because it might take a minute or two. That's fine. Now I'll

leave that there. I just want to jump onto something else called a cursor tool or IDE. I'm not sure if anybody is familiar with uh coding assistants. Anyone? Cool. I I see that actually some of you are familiar with familiar with coding assistance. But the difference between the coding assistant and this guy what I seen I shown you is you need to understand the code a little

bit on the other side but that also there is a trick for you not to worry about that. Let's go go through that. So I have a cursor here. Okay you can see that I just opened the cursor I simply let me open a new one. Cursor is if if you don't know what cursor is, cursor is an IDE based on Visual Studio for code. It has

been integrated with an AI agent inside. So it's basically it comes already packaged with AI integration AI um agent inside it. Okay, you can change the agent the way you want it. Like for example, you can pick up the anthropics agents or you can pick up the open AI. You can it's up to you whatever you want, right? But let's say let's let's rather than me talking

much about it, let me create a new application. For example, I'll just call it as chatter. I just going to open So if you see that there's nothing in there other than just a creating a folder. All I have to do is that on the right hand side there is a chat client, right? Or the chat interface. If you think about it, do I need to increase

my font? I think that's okay. If you look at it, right, that's what the you have multiple agents already working for you or at least the cursor is allowing you to work with. So in this instance, I could pick up the cloud force onet which is absolutely amazing by the way or you can also go for 4.1 GPT 41 which is also good but my preference is

obviously sonet because that's more uh coding side is really done very good. Okay. So I said that you know I want agent that's what I did. Shall I increase the font? I think probably I'll increase a bit. Is that is that any better? Okay. Okay. So you can see that I am saying to the cursor pick up the agent. In this case that's my agent. I also

said that you know instead of asking for other type of models or going for auto because auto means like it can choose its own agent. I chose the cloud force sonet. That's what I wanted. Perfectly fine. Now this is exactly what I'm trying to do there. I'm just going to come and prompt it here. Right? That's that's what uh we managed to say it's a chatter. So

let's say let's build I mean again I'm not though we are coders we are not doing anything coding. Right? So we just say let's build a chat application using open air. Right? That's what I'm do you want to specify a particular stack in this case let's say we use streamllet I don't know if anyone of you are aware of streamlit streamllet is a very lightweight python based

UI or you know streaml is very straightforward to do that job for you streamlit for this purpose okay that's it nothing more than that right and you click the button obviously I have logged out I've been logged out so hopefully I should be all Right. I'm not I don't remember which password or which username I used, but let's see. Okay, luckily it managed. Okay, so let me

go back and do the job of um you know cursor doing the job. So this is again the agent but in a coding assistant. So similar to this coding assistant there's another one called GitHub copilot right there's others actually windsurf is another one client is another one there are multiple coding assistants which are coming up if you haven't been experimenting with this coding assistant and if you

are a developer I think you are missing a lot I'm not selling this product because I'm not I have not even close to any of this right but my my point here is that these coding assistants are are actually going to be helping you a lot you may have an issue of saying that okay are they going to be writing bad code etc forget about that right

I've been experimenting with most of these models from I think last 2 years and the the I can only say this one the quality of these agents doing their job from zero to till now is only exponentially growing is really really exponentially growing okay some of the sometimes I find the code that is written or the way the these agents debug I can't do it myself. Sometimes

I sometimes I do think that I didn't even think about that particular problem in this in this area. So how did this guy got it? So that's what it is. So your job is not to worry about how these agents are going to take away your your job. I mean your day-to-day jobs instead use them to make your make yourself much more productive. Right? That's that's the

that's the thing which I wanted to say to you. Now look at that. This guy has gone away has created me. I mean while I'm talking to you it's managed to create a lot of things. You can see if you go on from the top it says that okay I help you build a chat application perfectly fine. It created the you know uh the basically the the

the pretty much you can on the right hand side left hand side you can see the code. I didn't write any code. It it wrote this one on the replete. I did not need to see the code but I'll show you the code in a second as well. So it just went away. Okay, done the job including it created me a readme file as well, right? I

mean that's that's awesome, right? I mean I don't need to worry at all. That's good. And it has gone away and then said that these are the libraries. These are the dependencies for your application, Mr. Cond. So I'm going to add them to you to you don't need to worry at all. So everything has been added to me. I'm looks like everything seems to be perfectly fine.

So it said that it's already been ready. How long it took me? A minute, right? So that's fine. I could you know obviously if it doesn't work I can ask it to make sure that it goes away and does a job right it tells me you know like as a dumb developer it tells me all you got to do is that you go and install the dependencies

which is just normal dependencies for an application for a Python based application and then start the application right that's that's that's pretty much I have to do nothing more than that so let's suppose I have been happy with this one I just simply accept all so that means whatever the application files that has created it has managed to give it to me very very simple right I

mean very straightforward so my job now is to test it out right so I think because it's a python based application obviously what you need to do is that you need to create a a what do you call a virtual environment right I'm just going to create the virtual environment and once it's going to be created. I'm just going to jump into that. What is it's is

a it's a little bit about the uh uh virtual environments on the Python side. So don't need to worry too much about it. It's just a normal thing. Be activate that should be fine. So now I'm my I mean you can always ask the agent to do that job for you as well. But in this case I I have been a bit lazy did not ask the

agent but you can always ask it. Okay. So I'm happy with this one. So let's let's first it's asked me to do the job of you know running the installs I mean requirements. So so I'll just let me say I can ask that to run as well. Okay I can just say go and run it. You can see that it's is running for me. I don't need

to do at all. And then you can always ask once it's done I can ask the streamlet to come and start doing the job. While this is doing let's jump on to the stream. I'm hoping the streamllet is somewhere there. Yes. Okay. Oh, initially I had a green one but looks like the screen has changed a bit. Oh, look at that. Looks nice actually, isn't it? Okay,

so it has created this application for me from scratch from completely from ground up. And if you look at the initial version, it says it's been completed. So that's good. I can, you know, it tells me that there's a checkpoint has been made. The checkpoint is basically to say if you want to if you keep on iterating at some point of time if you want to go

back to a previous version that's a checkpoint. You can always go back to the previous version so that you don't need to worry about it cuz sometimes these agents are a little bit tricky to work as well. They do a quite an obvious mistake. Sometimes they screw up your code. So you got to be making sure that you go back to your previous version as well. So

looks like everything is fine. It did mention to me to say that Mr. I have built your vegetarian uh recipes app. Are you happy with this one? Right, I can say yes. But let's let's try it out. Right. So on the top, remember I said to you this is a full stack development environment. Right? So if you see if you look at it, right? There's a QR

code. So you can always access that QR code and check my vegetarian app or I could go to my development version of it. So let's see actually how it looks as a development version. Right? So that's the application which has been developed. Right? You can see that it's got a I mean this is I don't know if you are anyone is here a front-end developer but I'm

pretty pretty sure that you won't be able to develop this in a minute or 2 minutes right I'm not challenge you challenging you but that's that's the reality right so looks like it's good right it's not bad at all it also given me the trending this week I don't know how it managed to do that I didn't even give any information but look at the footer as

well it it managed because a website kind of a thing application it managed to say Mr. Kund needs a footer as well. So I'm going to create a footer as well. It's kind of I did not give you you seen that right one single prompt is what I gave right. It went and done the job everything automatically. So looks very very good. Now you may say that

well that's okay. you if you want to deploy this. By the way, if I if you want to deploy this one, you can simply go over here and ask go to your deployments tab and you can say set up my deploy deployment and how many machines you want, two machines, one machine, 10 machines is up to you, right? But this is all on replete uh infrastructure, right?

So you can simply say because for me it's a development version. I would say probably I could also edit this one, but I'm not worried about it. I'm happy with this one. So do you want to change this name let's say rather than monda let's say I want that one nobody has taken that one so deploy so that now whatever the application I have developed now is

available for the world not just the development version what you see the URL which is a gibberish URL here not that one is now trying to deploy my application into the cloud into its own infrastructure right that's actually an amazing thing remember if you are from older generation you know how painful is an application to be taken from this you know an idea into hands of a

user right if you think about it this is a MVP right this is your vegetarian class is your MVP you are taking that you're giving this URL once it's been uploaded that URL is is is you can give it to anyone to right go and do the testing right that's that's what you can you can go to your I'm sorry VC this is the app I built

it over the weekend. I mean you don't tell them because they won't give money otherwise you just said this is my app right and it this does the take away this particular pain point are you going to give the money for me right that's the MVP that's exactly what you did right you did not spend 3 months for that you spent few minutes probably or 3 minutes

rather right you can see there is a URL you can if you open that URL that should be available for everyone so you can go to that URL it's available you can also let's say you have you've been really happy with this one Right? You said that oh this after a few iterations. So let's say I want to say here um it it manages to give me

everything what I wanted. But let's say you wanted um it it's asking you can you browse recipes and asking me okay I'm happy with this one. I can say what are the do I need to increase the font? Let me increase the font. What other features I should build? Because let's say I'm very dumb developer. I don't know anything at all. I'm trying to build this one,

right? It's it it tells me because based on the context what it has got, it says that this is a vegetarian delight app. So, I'm going to build something else, right? Oh, it's going to say there are it seems actually, by the way, look at that. There is a TypeScript error. I didn't know that. Right? It managed to find out that and it's actually going to fix

that by himself by itself. Right? That's clever. Right? That's really clever. And he's going to go and do the job of, you know, doing all those things. Now while this is doing let me show you one quick thing on the if you go onto this particular uh files folder you see there is a there's a code that has been generated right so this is exactly the same

code what we sent similar to the cursor or the coding assistance what has been done now if I'm if I were you or if you if you want to try it out on your cursor you can uh go to the three dots yes and you can download that as a zip folder folder right once that particular folder once that particular that means actually that code is completely

into your uh environment it's not telling you reply is not telling you to say you know it's only in my environment right so the claimly for example the one which I shown you in the beginning of it the one which is somewhere here I don't remember but that guy right this one initially I developed everything on on on replet right but there is a after that I

have to because you you want database and because you want to have deployment into your own environments you don't want that to be you know like locked to replete I want that to be outside replet so all I did was you go over here right you download your files once you're here you use your uh your cursor to open a new folder and obviously go to that

repository and that should if I have it here somewhere I should have it here. That's fine. Sorry about that. So, I should have one of them uh you know the the vegetarian device as you can see there. I once I unzip it I my my application is already available there. Right. Pretty good. Right. And now I I brought all the things what replica has done for me

into cursor. You may be asking me a question why would Mr. cond is going to be jumping from one environment to another one. It's just that I don't want to be locking myself into an environment like replet. Instead I want all the code base with me. I may require something which I don't want to be experimenting on on replet right that's the reason but if you're happy

with replete make sure that your MP stays there you go to your VC get your money and then you know build a build a big team and start asking them to develop this one from end to end it's up to you how you wanted to market that one right so got 13 more minutes so that's that's what you should do now let's come back very quickly onto

this guy so this seems to be all the everything seems to be perfectly fine in hopefully this works out. Let's let's run this application, right? Obviously, it did not work out. So, it tells me that there are no secrets because at the end of the day, OpenAI requires you know your you are expected to have an open AI key. But look at the the way here it

has created it has managed to it should be using 3.4.0 to zero but you know it just managed to it also has given me in the mechanisms to change the temperatures which is really good or temperature of the models and also I can just tell that to you increase or decrease the tokens which is really good and so on and so forth right I'll show you one

application which I just done exactly like that using streamllet using cursor that's the product owner AI assistant right so this particular application if you think about it right I have A you know here here is let me increase the font size for this. Oh yeah. Okay. If you look at it right, this is a a PO that is a product owner. If you are product owner, so

you know actually you had to write some stories, user stories for your development team, for your analyst to start, you know, exploring them, right? So in this instance, I created this simple very very widecoded application end to end. Okay, I didn't do anything at all. All you're going to do is that you can say in this instance, I said I need to build a set of user

stories for my develop dev conference app. So I want I'm trying to build a dev conference app for example. I want to create user stories and you you want to pick up the way whatever you want. You want to create an Apic or you want to create a user stories is up to you. If you have a mechanism or if you think that your organization has got

a set restricted um product ownership related standards you could upload them to say don't use everything what you know but just use make sure that this is what you should do right and you can generate that one is goes away. This is using actually open AI. It goes away. It generates it and it gives you the user stories as you expect it to do that. But this

is again white coded right. This is from zero. I took about half a sorry half an hour to bring it up. So these stories you can see the user stories. So as as a user it tells me exactly as a so and so as a conference attendee for example you want to register that's exactly what you want to do so that I can access the event schedule

this is as a so and so and what you want to do these are the user stories that has been created now imagine if I integrate this one with confluence with the Jira tracker or with any of them I could straight away get this get these stories directly into your buckets into your um next step which is basically for development for for the engineers etc. Right? That's

that's the idea there. I want to show you one I got 10 minutes so I just want to show you um one particular application which I did. I'm not sure if anyone is so by the way I'm jumping from replet is done uh AI coding which is cursor is done. Now I'm introducing a third one which is the N8. N8 is an automation tool. I'm not sure

if anyone is familiar here. Good. So, look at this one. So, I'm glad actually someone some some of you are actually aware of this one. So, I am currently building a an engineering manager, right? So, the idea here is really really simple. What I mean to say here is if I I mean um as an engineering manager I can give a required uh instructions to my manager

to my engineering uh master to say go and build this one go and build a user authentication story for my feature as a feature. What does it do? Right? If you think about it right, this guy is actually going to use the open AI but look at there are three tools it has. It has got the requirement extraction agent tool. It has got the enterprise architect tool.

It also has got the software engineering tool. So that means your requirement is going to be passed on to the extraction the requirements agent tool. So that agent is has got additional agents behind the scenes. I'm going to show you what those agents are. Okay. So for example, where is my requirements? So if I if I show you the requirements agent, that's the requirements agent. So from

here when the engineering manager has kicked in, it says to the I mean whatever the requirement you put in here, it goes to the engineering manager. Engineering manager is going to say this is a requirement. First I need to do is that it has got his own brain. What first I need to do is that I need to break that requirement into multiple tasks or multiple user

required stories. Right? It is going to go to this extra the requirements agent. And what does requirements agent do? Requirements requirements agent is going to use product analyst that is a product owner. It also uses the user researcher to have that user designs. It also has got some domain knowledge expert or or theme. Right? So these are all the individualized agents which are actually making up to

be an extraction requirements agent right. So this is a super agent which has got other three or four agents below it as tools working together. But this agent is again is going to be given back to the the engineering manager. So engineering manager uses that guy right. And if I I think I should be able to run an example. So I think let's say the same one.

Okay. So you can see that actually it's kind of goes one to the another one. Now once the extraction of the uh user stories has been done, the manager, the engineering manager in this case, the agent in this case is going to go to the architect to say these are my user stories. Go away and build an architecture or a build a design for this one. Right?

Just imagine how much is actually doing behind the scenes. I think that may not work because I'm using the MCP tool behind it. MCP server has not been started. So probably might not work. Just ignore the ignore the working of it. Now engineering architect or enterprise architect his job is that it's got the user requirements. It's got the functional requirements everything has been given to it. So

he that that agent is actually now going to be starting to create a architecture for your application. Think of it as complete architecture for your application. Right? And then it goes back to the engineering manager to say that I am done my job. Here is the architecture. Right? The engineering manager now comes back to the software engineer to say that these are the architectures. These are the

user stories. Now go and build the system for me. Right? How how intelligent is that? Now software engineering manager or software engineer agent is then going to go away and start creating the required tools. I mean sorry required code whatever your whatever your stack is is going to start doing that one. I also have three more agents which is the DevOps agent, the testing agent and um

monitoring agents, right? I haven't integrated them yet but that's the idea here. So if you if you uh if you think about it right so this is the the requirements agent that has done everything for me. It created me the the BDD functional specification. It has said that okay this is the you know the idea behind it. So you can see that each one of them is

a story. Of course, I have to expand a little bit more on, you know, writing the story and so on and so forth, but you get the idea, right? The idea here is really simple. And a simple prompt that has been given by the user is taken by multi- aents, right? It's a multi- aent system, right? And developed into a working prototype. How how it was not

imaginative in our world few years back, right? That's the beauty of it. I mean do not look into AI as something which is a disruptive but think of it as something which is actually making you 100 times better at what you're trying to do or what you will do. Okay. So this is if I I mean I haven't as I said to you I haven't fully uh

finished development on this one. Just imagine if I do this one right it's it's going to be given to the enterprises for example and also imagine if it has been deployed on your organization with existing code base they can these agents can start looking at your code or looking at your application start monitoring the CPUs start monitoring the user bugs and it's actually going to go and

fix it as well you don't need to be part of that that you know the whole ecosystem right it kind of manages proactively to see if that application has done you know has gone away from I mean is is the CPU has gone off or you know there are some 404 errors or something like that and it goes oh you know Mr. Honda has cond's application has

been a problem because there's four or four application errors are happening or there's a DOS DOS effects right right attacks and those can be integrated or this can be taken care by these agents as well that's the NAT okay let's come back to um the final part you got 4 minutes um so here I asked it to by the way this is again a multi- aent system

right this is simply put if you think about it right you have an agent behind the scenes working for you but it has got multiple things that is pretty much what I did in nan is actually trying to do it for you behind the scenes that's what replet is trying to do here so if you see actually I asked it to say what are the features I

should add probably I don't know yeah sorry it has given me a ton of uh features says that recipe creation probably advanced search and discovery probably I mean you know there are smart features for example let's say uh let's build some Smart features. Okay, maybe that's what it is. Or social. Let's build some smart features. So, still remember no code, no touching of my any of my

code. Everything goes by the prompt, right? So, that's the beauty of here, right? And I can only tell you this one. Sometimes it irritates you as well because these agents are this go in the loop. So you go to have some tips, right? One of the tip is that try to go back to your code sometimes. Again, if you're not a coder, that's okay. Understand, ask it

to say where exactly this particular code has been written. It is actually going to give you the location of the code. Is going to tell you which part of the file that particular code has been present. You go and check it. You go and inspect it. If you are a developer, you know exactly where to do that one. And if you want, you can always change that,

right? You can always improve that. If you think that that's actually going through the loop of, you know, not it's called some some people call it's called of a loop of hell. Actually, it doesn't it doesn't do really very well, right? You can do that. Certainly, it will help you or you help it to make sure that your your application is up and running. At the end

of the day, your job is to make sure that the applications are working as expected. that that's that's the job for you, right? So you work with the agent is an assistant. It's a companion for you to make your idea to bring it to the to life, right? That's that's what you should do. So there is a mechanism to do that one. I'm I'm not going to

be introducing you to something called chat feature inside the repate or I'm not going to do anything at all. But I just wanted to introduce you to one last thing which is allowable. Okay. Exactly like um uh replete there is another application called lovable.dev. Another one is called bolt.dev. Bolt you can also run it locally locally as well. as you can see you can simply say a

particular query or you can give a prompt the application will be ready in few minutes time. Just imagine you used to take at least a a week or two weeks to create a a client's website earlier but now it can be done in probably few An application which is front end back end and also middleware used to take a week or two weeks to months and months

and months. Now you can do it actually in about I mean I wouldn't say everything will be done in minutes but I can always tell you that you can get that running up and running in just about half a day to one day. We've been developing agents or sorry we've been developing c um chat clients or chat bots in our organization. We have created this platform where

we could churn the applications or churn this chat boards pretty much half a day or one day you know that that quickly you can do that one that was never used to be the case earlier. Okay. So keep this in mind. These are the improvements. These are the things that you could certainly help you do that one. My last thing is that these are not disruptive tools.

If you think about these are disrupt absolutely they are disruptive tools but use them for your own benefit. These are amazingly intelligent. I mean they not they don't have the intelligence at the end of the day they are all models which are you know predictability unpredictability they but use them for your own benefit right become the 10x developer what you want or be that product owner who

creates that idea or takes this idea into market or into his hands within a weekend or within a few days time right I think that's the thing which I really really want to stress upon you upon you here potential of these models is only beginning. If I'm if I'm looking at it, I can only say I was I was I I was experimenting these models for the

last two two and a half years, right? Every time I see a new model, the potential is is exponentially increasing. It's it's mindboggling. I mean, how much the the the the stuff is actually happening between these models, right? And it's only going to go better and better. So don't be on the sideways. Come onto the mainstream. Adopt, adapt and make sure that you learn AI with an

intention that you want to get succeed in your life. Oh, that's exactly 45 minutes. Thank you so much, Maddu. Uh now it's time for the Q&A session. And regarding the uh extraordinary circumstances that we have um actually we can um give you the mic and you can just if you raise your hand you can ask a question. here you are the first question. >> Thank you for

the great session. Um, I have a question. I have about 50 of those projects started, but the moment I see they're going to work, I lose interest. >> Sorry, I didn't get that. >> Do you have any advice? >> Sorry, could you repeat that? I didn't >> Yeah, I'm saying uh I have about 20 such projects, small things that are doable with AI started. But the moment

I see it's going to work, I lose interest almost at that point. Do you have any advice how to close? make that too because I also you must have seen my projects right I got 60 or 70 projects only on replate about 100 projects on the cursor so exactly like you I create I take a painoint and I try to solve it right the problem you and

me have is that we don't talk to the customer you and me take the idea and then make it a solution for you and me but that's not the that's not what customer wants right that's not what a client wants that's what actually ordinary person wants, right? Thank you, Simon. >> Uh so, >> so can I can I smuggle a second followup? >> Yeah, sure. So, so

just to just to finish off the answer is answer is focus on what somebody wants. The claimly I know it's a problem for me. I developed because it's my my pain point. At the same time, I also interested in using AI to do these things, right? But if I really look at it that plainly or that particular expenses app there's plenty of the apps already in in

the market right so only my my if you are interested by all means experiment them but if you're actually looking into a painoint go and talk to the customer go and find out the real pain point and then come back and do it very quickly >> okay sorry u I think you sort of already agreed with what I was about to ask Do you think that this

uh the way that AI works right now in software development means that there's a huge premium on product and business skills even for technical founders essentially that the needle has shifted and now what matters is building the right thing and not really to know how to build it because to know how to build it is to just type a few prompts. Maybe I'll talk to you offline

on this one a little bit more in detail but it could become one month bands like you know there there is a choice but at the same time if you think about it right application needs a stronger scalability reliability you know all those enterprise versions right um you you know for a smaller applications one person can work out but if it's a bigger applications probably a team

of people should be around so I strongly wouldn't suggest that uh reducing your team or you know doing other side is not going to work out very well. Instead making sure that you know instead of doing it a feature for 3 months work on a feature which gives you returns in a one week and then develop 100 features in in a year. >> I'm sorry somewhere in

this someone asked a qu wanted to ask a question. Here you are. >> Sure. Thank you. So um I have a quick question. From problem to prompt what is your process? Sorry. >> From problem to prompt. >> Yeah. >> What is your process? >> Oh yeah. So, so I think I mentioned it earlier. I took I take a prompt or sorry I I have this idea right

idea is is expressed in a language like you know for me it's in English for you could be Bulgarian or French whatever it doesn't matter. So that is a natural language right now. Computers really wouldn't understand the natural language. Obviously motors are the ones which are now dissecting or deciphering them into a language that can do that. So but giving a simple prompt you know the the

prompt like you know like the one which has shown you is not is not necessarily the right way to ask these agents to work for you. So what I do normally is that I go on to chat GPT or I go on to deepsek or I go on to manus sometimes manus as well. I ask it to explore first of all before I jump in there. I

explore that particular problem and I want to build so and so critically evaluate this idea. So I start evaluating this idea with these models. Right? Once you know that idea has been evaluated, I ask it I'm going to develop this one in cursor or I'm going to develop this in N8 or whatever the tool. Can you create me a prompt? It's actually going to be a long

prompt for you. Right? of course take that prompt and then stick it up in in any of them and that works out very well but I still have a mechanism like for example in GitHub copilot I use some some instructions for every project to say that although I'm going to give you prompt but you should make sure that your project is created with the standards right I

always do that one the reason is that you may have created multiple projects but they're all exactly the standardized ones that's what I do but Always don't just go for oneliner. Explore, evaluate and come back and then fix it because that's where your ch chativity is going to help >> Thank you very much. No worries. >> Thank you. And we have time for I believe one more

question from the lady over here. >> I met you there. You eating your nice ice cream there. >> true. Uh so thank you for making all of us from engineers you raise us to managers with with your scalable application. Uh so the the direction of uh development I see here uh not only in tech technological perspective but also from the human side of the organization. How do

you see that? >> I think it's a really deep question. So probably I don't know whether I'll answer to you that or but rather than calling yourself as a manager I I'm termin engineer. So we used to be software engineers you call yourself as a product engineer. So you are combining the product and combining the engineering skills what you have and becoming a project engineer. Right. The

second question probably I'll talk to you offline because I think that's a human intervention. I think one of the things I always say is that you cannot get rid of the humans in this particular I mean humans in the loop is is the biggest ask here right we are not building the machine learning models we are not doing the predictive analysis but still I would strongly suggest

you to keep that embrace that idea I'm not expecting anybody to go away from that particular part of inter intervention but that is what I'm expecting >> thank you Matthew thank you so much for being with us today. >> No worries. Thank you so much again once again. [Applause]