AI coding platforms like Blitzy can generate complete, production-ready full-stack applications from a single prompt by first ingesting and understanding the entire codebase to create detailed technical specifications, then producing hundreds of thousands to millions of lines of code including backend APIs, real-time messaging systems, authentication, and frontend interfaces, fundamentally shifting the developer's role from code creation to code management and refinement.
Deep Dive
Prerequisite Knowledge
- No data available.
Where to go next
- No data available.
Deep Dive
I tried building Whatsapp in one prompt
Added:All right. So, today we're building WhatsApp [music] with full backend, realtime messaging front end, and we're doing it all [music] from a single prompt. Or at least that's what I hope will happen. So, [music] how are we going to do this? Well, we're going to be using a platform called Blitzy. When they reached out to me, I did what [music] you are probably doing right now, which is basically roll my eyes and think that, well, okay, another AI coding platform tool that's going to be revolutionary, that's going to do whatever. But I decided to look into it and they really are doing something quite different. So first of all, they've managed to do one thing that I've not seen any other AI company do yet. Blitzy [music] understands your entire enterprise code base. It spends up to a couple of days just [music] ingesting and understanding your codebase. It then produces what they call a detailed technical spec.
>> [music] >> essentially a 200page document that essentially maps out every dependency, every architecture decision, [music] every flow in your app. And only after it's done that is when it actually starts building and doing stuff. And when it does build, we're not talking about small little changes. We're talking about hundreds of thousands of lines of code from one single prompt and in some cases up to like 3 million. So yeah, they really are doing some things quite differently and that's why I decided to accept their offer, but only if I was allowed to really put it to the test and they agreed. So that's what we're going to be doing in this video.
So I wanted to throw something quite serious at this to do a proper test. So, not just a to-do app or a landing page, but we're talking about a full WhatsApp clone with realtime messaging, user authentication, chat rooms, message history, [music] the works from one single prompt. And personally, I genuinely doubt that this will be possible, but the team over at have uh assured me or they seem very confident that it will be able to do this. [music] As I'm recording this right now, I have not sent the prompt yet. So, I genuinely don't know [music] if this is going to work or not. Okay, so here we are inside Blitzy. And the first thing that you do is you create a new project [music] and give it a name. I'll call this one WhatsApp clone. All right, so the next thing that we need to do is we need to write a prompt for this WhatsApp clone.
And the way that I did it was actually worked with a member of the Blitzy team to help me write this prompt. It was also my first time using the Blitzy platform, so we didn't want to limit ourselves to my ability to write a good prompt, which would not have been good to be honest. Now, it's not very hard to learn how to write a good prompt for Blitzy, so you'll be able to learn that in a short amount of time. It was just that we wanted to give ourselves the best opportunity possible for this video. And here's what the final prompt actually ended up looking like. We specified the text stack to be React on the front end, Node.js JS on the back end, websockets for the back end, a database for messaging persistence. I outlined the core features, user registration, login with authentication, realtime messaging, blah blah blah. We also connected it to a WhatsApp Figma file that we created, which is basically the WhatsApp screens so that we actually have something to connect it to in Blitzy [music] that it can look at and reference for how to create the app. But that's basically it. So now the prompt is in and um I guess the only thing to do is [music] to press send or create plan is what you actually press here. Um and then um I guess [music] we just wait and see what happens.
Keep on walking.
[music] >> [music] >> So, it's been a total of I started running this on Thursday last week and it is Wednesday today. I think it finished sometime this evening or the last night I mean it's been running a while which is interesting and gets me kind of excited or curious to see what it's even done.
So project guide [music] project over cal is a production grade horizontal scalable WhatsApp clone web application um full mon [music] reposed backend API full front end application [music] database layer I'm not the most technical I can get some stuff done but I'm not the most technical like this seems very very detailed [music] remaining gaps the 32 remaining hours of work fall into verification [music] enter okay so basically just testing the the product I think environment setup okay this is actually really good cuz I don't know how to run this like I think it's using [music] cuz it would have created a backend server like it doesn't actually use a backend server that already exists like [music] something like Firebase base, superbase, stuff like that. Uh it probably could if I would have uh prompted it to, but what it's done is basically built everything from scratch.
So that means that we're going to have to use Docker, which I've never used in [music] my life. Uh so just running this will be a little bit of a challenge for me personally. [music] Um so we'll see.
Okay, so I [music] might have to install some things. Docker All right. So, this is going to be a little [music] bit interesting. Here's the GitHub repository like the we basically just merged it. It was an empty git repository to start with and it created all this and basically it has one [music] single command bootstrap.
So, git clone you clone this repository and then cp [music] blah blah blah docker compose up and that is all on first boot [music] the system automatically. one starts a postgrql 16 and whatever this is radius containers [music] waits for database readiness runs all Prisma migrations to create a schema this is the one that I thought was interesting so it seeds the database with the deterministic demo data so users conversations messages and valid cipher text encryption key bundles and then it starts an express API server nextjs front end bull mq worker backup server I don't even whatever I haven't been doing like I use flutter most of the time and [music] very simple stuff compared to this as you can see here I don't know what [music] this even means but it should be just the one command then to get the app [music] up and running so then access point front end we can go to the front end and this should [music] be okay wait that's actually really good just seeing that but before we go there I I want to open all of them in different windows.
So, we're going to start with the API.
[music] Open like a new tab. Health check. I mean, there's a health check.
That's pretty cool. Uh, so this is health data status [music] healthy version blah blah blah. Up time 67. So, we can probably reload this. So, it says uptime 67. [music] If we reload that. Yeah. So, that changes. So, it's now 143.
And then this is the actual app. So, we go to localhost 3000 and then we get to this place where we can log in. I don't know [music] if we have to type in an actual number here. So, we're going to try to just do 499999.
Done.
Okay. And [music] then we get into this page. We have a settings page. It actually I mean it looks good.
Hey there, I'm using WhatsApp. Okay, chats.
We can clear [music] chats. We can archive chat. We can delete.
I mean, it's taken [music] the the Figma files and it's basically just replicated the entire design here. Um, [music] where everything is clickable, everything works the way that you would expect it to work. You can see like all of this stuff like it's gotten [music] so far in in this amount of time and it feels very complete. I mean I can change my username. I can change my number. I think I can change whatever it says here.
Um the way that it's set it up so that you just have to run this and then everything will work. one of the main things that I was worried about because I knew that it was going to use a lot of different um software to get this done.
So, I didn't know if I was even going to be able today to run this or if I was going to be able have to like contact some people that know better than me.
It's all here. Like, it's documented everything and literally just ran one command and everything's up and running.
So, the only question is how do we find other users? Like, should they have shown up here? I'm not really sure. So, I'm going to actually have to ask the team at Blitzy because I think that there might be something that I'm doing wrong because looking at it, it seems very complete to be honest. Uh, all right. So, I'm back again and uh now what I've found out is that there is actually users and C data available in the app. It's just that it's in the back end. was existing in the database but is existing via I think it was let let me have a look cuz I they sent me a message the C data is present in the codebase and would need to be converted to phone number login instead of email the back end is set up to handle either but UI only works with phone number login instead. So yeah that is essentially the the situation right now. All right. So, I've now gotten a response from uh the Blit, which means that they've also written a refine prompt. So, we're going to send this and then hopefully after that, we'll have the C data in place.
So, let's have a look.
All right. So, here is the prompt. Uh apply five targeted bug fixes to the WhatsApp clone. Uh, one fix login page to use email plus password form.
Uh, fix the data or credentials message content and image media URLs. I don't know. We have a lot of stuff here. So, it's six pages of a prompt. And uh, I guess we'll just copy this. All right.
And then we just paste the prompt. So, we'll paste that in here. And then we'll let it run for a few days and come back to it. See what happens if it does in fact have the SID data if it's worked out how to do the things that we need.
And uh yeah, I'll be back in just a couple days and we'll see what the result is.
All right, we are now back and the code has uh finished running or Blitzy has finished running and so this is the review or like the the project [music] guide that it has created afterwards.
Um, this project delivered five targeted bug fixes. N director blah blah blah to [music] the WhatsApp clone monorapo ensuring the docker demo stack operates without errors.
Fixes address key accompliment [music] login page replace phone number keypad UI fully replaced with [music] email plus password form post blah blah blah demo credential hint. But we're just going to merge this.
And because what we're doing now is not working on a project [music] that is super important in the project. The actual thing that we're doing is we're just testing [music] Blitzy to see what it can do.
All right. [music] And and now we're running [music] it again.
And this means that we might be able to do this.
Okay. So now we have this demo credentials.
Sign in now. Here we go. This looks better. We can actually see some of the chats.
Um, [music] so if we press that, we get into the chat and we can see the messages that have been sent, which are nonsensical, but that doesn't really matter. [music] And then we can type in hello world and send it. It has now been sent. Um, and so yeah, I guess now it actually does work. It updates the the notification too to say like it says that we have two unread messages here.
So if we press that we then get into the chat and then when we get out of it the indication that we haven't seen the message is is gone. On the settings page we can check our profile. We can see our number is this.
What I'm thinking that we can do can we open up another window? Okay. Yeah we can. And now let's try what happens if we just use another thing here. So holdengmail.com and then one 2 3 4 5 6 qh exclamation mark. Okay. So now we can't create a a an account. But that's pretty simple to add. I think if it's able to do all of this stuff, it would be pretty simple to add. What I wanted to do was I wanted to check if this if we can actually send messages between two users, which is now not something that we can do quite yet in this. But I mean, it's still pretty all of the components are there. I want to be honest about this stuff and say that it definitely isn't completely finished [music] and it's not something that's not going to require a developer to take a look at and actually [music] kind of guide it through until it gets to a finished product. But what we have is we have the architectures fully [music] done. We have a tech spec that's very like indepth. We have like there's so much stuff that's just been done for us.
[music] The core features are already there. Basically, what that does is it shifts the role of the developer from being someone who actually creates these [music] things to being more of a managerial role where it's like you're looking at it, you're reviewing it, [music] you're polishing it, and then you're actually able to shift it. And I think that this works similar to how [music] I believe the indust industrial revolution worked uh [music] where it didn't replace what was happening like it did probably in a lot of cases [music] but the main result of the of it or the outcome of it is larger production [music] like you could produce a lot more. So that's how I kind of view all these AI tools in general.
It allows you [music] to produce a lot more and a lot faster. And that's why I don't think it's going to replace developers because [music] you still need a developer to be able to look at this stuff and understand it and know how to even prompt it because there's so much stuff that you don't know [music] when you're not a developer. I have not tried not being a developer [music] and trying to build stuff because I am a developer and I know a lot of stuff like I know [music] I think I'm kind of the perfect level when it comes to using AI for this sort of stuff because I know enough about the details of things that [music] I could actually build it on my own. But I'm not advanced at it. I haven't worked as a backend or full stack developer for years and years [music] in different companies learning and developing my skills to be like super efficient at it, [music] but I know enough to be able to actually do it and get it done.
[music] And that means that using these AI tools, you can become so much quicker with completing [music] things when you have that sort of base knowledge. And it's not hard to get the base knowledge that I have, but even looking at it when you [music] see the prompts that the Blixie team wrote for me that helped out, like they have way more technical skills than I have. [music] So what they're able to create in terms of the prompts is just so much more complete and so much more full, if that makes sense, where they're able to predict some of the problems. They're able to even explain the structure, the architecture, how it should be done.
And with something like Blitzy that allows you to actually create something that's very full and very complete in the end. Whereas using things like clawed code which I've used you basically end up with a problem where it's doesn't it's not able to keep the entire code base in mind. So what it will do a lot of times it will forget that it's written a file back in like I don't know 3 weeks ago that it could now reuse for this new feature that you're building. It increases the possibility that will introduce some like bugs and things like that in the codebase because it just adds new features and adding a new feature has the potential of increasing or introducing bugs. So, is Blitzy the most advanced AI platform out there? As far as I can tell, and what I've seen here, it does seem like it genuinely is in a league of its own.
Like, it's doing things that I haven't seen anything else do. Mainly to me, it is the deep context that it has, the fact that it actually looks through your code base, the existing code base, and just the sheer scale of what it can actually produce in one single [music] go, like from one single prompt. Nothing else that I've used really comes close to it for a project of this size. So, I highly recommend checking it [music] out. There will be a link in the description. And they do have a free tier so you can actually play around with it, test it out without fully committing to the software. [music] And um yeah, I hope you enjoyed this video and uh thank you for watching and I hope I'll see you in the next [music] one.
>> [music]
Related Videos

Expanding Stikbot thumbnails
leopoldshorts
2K views•2023-09-24

Digital Discrimination: Cognitive Bias in Machine Learning
redmonktechevents2974
4K views•2019-12-18

Evolutionary Approach to Clustering by Ujjwal Maulik
ICTStalks
279 views•2019-06-26

Rose Yu "Learning from Large-Scale Spatiotemporal Data"
networkscienceinstitute
2K views•2019-03-04

Stanford Seminar - Generalization through Task Representations with Foundation Models
stanfordonline
4K views•2025-07-14

Satellite-Based Wheat Yield Forecasting using GEE & Transformer Neural Network
gisrsinstitute
634 views•2025-06-15

Paradigm Shifts in Data Processing for the Generative AI Era: Robert Nishihara of Anyscale & Ray.io
GradientFlow
2K views•2025-01-02

How to Build Your Own GenAI-Based Knowledge Management System
2150GmbH
360 views•2025-06-03
Trending

Playstation NO DISC/NO BUY Fight Is Over...
DavidJaffeGames
4K views•2026-07-23

Americans Confused in Australia for 17 Minutes Straight
IWrocker
17K views•2026-07-23

Bitcoin Social Interest: Dozens of us Left
benjaminjcowen
12K views•2026-07-23

Tesla Profits Plunge & SpaceX Stock Continues Fall
TheJohnJohnstonLounge
6K views•2026-07-23