These NEW Lovable AI Tricks Will Change the Way You Build!

Description

👉🏼Join the Early AI-dopters Community: https://bit.ly/3ZMWJIb
📅 Book a Meeting with Our Team: https://bit.ly/3Ml5AKW
🌐 Visit My Agency Website: https://bit.ly/4cD9jhG

In this video, I showcase the latest Lovable.dev features that let you build and iterate on your applications directly from your browser.

We’ll explore:

1.Generating a Full App from a simple URL—no local environment needed.
2.Using the Labs Feature to chat with a Large Language Model without altering your actual prototype.
3.Seamlessly Importing Figma Files (using auto-layout) into Lovable.dev to kick-start your design.
4.Leveraging 21st.dev to instantly import React components with prompt-engineering shortcuts.
5.Creating a “Car Prompt Engineering Agent” that helps generate prompt ideas for your projects on the fly.

This guide will help you:
✅ Understand how Lovable.dev’s newest features can speed up your workflow
✅ Implement 21st.dev’s prompt hacks for component importing
✅ Combine design and AI capabilities to build robust prototypes
✅ Streamline your prompt engineering process with an on-the-go AI agent

Whether you’re an AI-first innovator or just exploring these new possibilities, this hands-on demonstration will give you the insight and confidence to harness Lovable.dev for faster and smarter app development.

👋 About Me
I’m Mark, owner of Prompt Advisers, helping businesses streamline workflows with AI. With the right strategy, you can fully harness the power of AI to scale, automate, and grow your business efficiently.

#AIForBusiness #GenerativeAI #AIImplementation #AIWorkflow #BusinessAutomation #AIAdoption #FutureOfWork #TechTrends #ArtificialIntelligence #aiautomation

TIMESTAMPS ⏳

0:00 - Introduction and overview of video content
0:06 - Explanation of features and hack introduction
0:13 - Mini prompt engineer app creation preview
0:19 - Video tips and tricks teaser
0:25 - Feature 1: Creating applications directly in the browser
0:30 - Demonstrating browser-based app creation with prompts
0:41 - Explaining how quick demos can be created
0:50 - Example of generating a to-do list app
1:06 - Using "Generate.new" URL for direct app creation
1:26 - Overview of demo output and further customization options
1:55 - Utility of browser-based automation for quick prototypes
2:01 - Feature 2: Activating chat modes in Labs settings
2:15 - Enabling LLM chat mode without code changes
2:33 - Difference between "default" and "chat-only" modes
2:52 - Example of conversational guidance without altering code
3:30 - Chat interaction for app improvement suggestions
4:04 - Transitioning from chat guidance to action steps
4:29 - Feature 3: Integrating Figma designs into Lovable
4:40 - Leveraging Figma designs for live app development
5:12 - Explanation of Figma and Auto Layout requirements
6:04 - Verifying Figma designs for compatibility
7:16 - Setting up Builder.io plugin in Figma
8:32 - Using Builder.io to export designs into Lovable
9:16 - Importing and refining Figma designs in Lovable
10:04 - Editing imported designs directly in Lovable
11:01 - Finding free Figma templates for import
11:49 - Checking Auto Layout compatibility in templates
13:10 - Testing templates for import suitability
14:13 - Importing functional designs from Figma to Lovable
15:26 - Introducing 21st.deev for advanced UI components
16:08 - Example of importing futuristic designs
16:45 - Customizing UI elements and buttons using 21st.deev
19:01 - Testing and iterating on imported designs
19:50 - Setting up a spinning globe design in Lovable
21:37 - Creating functional email submission workflows
22:40 - Customizing imported designs with iterative improvements
24:43 - Building a prompt engineering app structure
26:17 - Addressing errors and refining app layout
28:35 - Creating task-specific cards and navigation flows
30:17 - Ensuring proper title and card layout
31:05 - Connecting Lovable with Superbase database
32:13 - Enabling OpenAI integration for app functionality
33:33 - Defining and implementing app sections
36:06 - Testing task-to-prompt functionality
37:26 - Iterating to fix formatting issues
37:57 - Testing prompt simulation functionality
39:53 - Troubleshooting errors in conversation simulator
41:12 - Verifying context retention in conversation threads
42:08 - Storing prompts and chats in Superbase database
43:00 - Resolving authentication issues in Lovable
43:34 - Viewing prompt history and database integration
43:52 - Unlocking role-level security in Superbase
44:03 - Final thoughts and recap of the video
44:16 - Call-to-action: Like, share, and subscribe

Summary

Revolutionary AI Web Development Features in Lovable.dev

In this comprehensive tutorial, Mark from Prompt Advisers showcases groundbreaking features in Lovable.dev that transform how developers can build and iterate on web applications. The video demonstrates how AI-powered tools can dramatically speed up the web development process without requiring extensive coding knowledge.

The tutorial explores four game-changing features that allow anyone to create sophisticated web applications directly from their browser. First, Mark demonstrates how to generate complete applications using just a URL with the Generate.new feature - no local environment needed. By simply typing a prompt into the browser, users can instantly create functional prototypes like task management apps.

Next, he explores the Labs Feature, which enables users to chat with an AI assistant about their project without altering the actual prototype code. This allows developers to get guidance and suggestions while maintaining their current build. The third feature showcases how to seamlessly import Figma designs with auto-layout into Lovable.dev, bridging the gap between design and development.

The most impressive hack introduced is leveraging 21st.dev to import polished, modern UI components using prompt-engineering shortcuts. Mark demonstrates how this allows developers to create stunning interfaces with futuristic elements like animated globes and sleek buttons without writing complex code.

The video culminates in a practical demonstration where Mark builds a sophisticated prompt engineering application that helps generate and test AI prompts. This real-world example shows how these tools can be combined with Superbase database integration and OpenAI connectivity to create functional, professional applications in minutes rather than days.

Throughout the tutorial, Mark provides valuable tips for troubleshooting common issues and optimizing workflows. Whether you're an AI enthusiast, product manager, or developer looking to streamline your process, this video offers actionable insights for leveraging Lovable.dev's newest features to build robust, visually appealing web applications with unprecedented speed and efficiency.

Transcript

0:00 in this video I'm going to show you 0:01 three brand new features in lovable. 0:04 deev and one major hack that's going to 0:06 take your applications from looking 0:08 decent to looking modern and futuristic 0:11 and then we're going to apply this hack 0:13 and build our own mini prompt engineer 0:15 that can create prompts for you and 0:17 simulate their actual outputs this video 0:19 is going to be full of tips and tricks 0:21 so buckle up and let's get right into it 0:23 all right so the very first feature I 0:25 want to walk through is one where you 0:27 can actually create and initiate your 0:29 first application 0:30 right from within the browser without 0:32 even being on The Lovable platform 0:34 itself so imagine you're in a meeting 0:35 and you're discussing building some form 0:37 of application and you want to create a 0:39 demo really quickly and you want to kind 0:41 of look like you're paying attention to 0:43 the meeting while you have a tab open 0:45 typing a very basic prompt in your 0:46 browser so we can do that exactly now by 0:49 just opening a tab here going to write 0:53 generate. new slash and then you write 0:56 your prompt just in one Big Blob no 0:58 spaces obviously because it's a URL so 1:00 you can write something like 1:02 q1 task 1:06 prioritization to do list 1:10 minimalistic design it's a bit weird 1:12 writing in one large blob but you'll get 1:14 used to it and you just click enter and 1:17 then this will go to The Lovable screen 1:19 initially and then it will actually load 1:21 lovable which will actually understand 1:23 this blob of text as the actual command 1:26 and it will put together a small demo of 1:29 what you actually asked for for and 1:30 you'll see here it was able to 1:31 understand our request create basically 1:34 a to-do list app pretty much that just 1:36 has the ability to prioritize certain 1:38 tasks so obviously we can go a lot 1:40 deeper here we can make this more pretty 1:42 more complex we can add authentication 1:45 we can add storage of the tasks using 1:47 superbase but as a starter again in that 1:50 specific situation where you want 1:51 something quick and ready to go this is 1:54 a cool hack to actually use your browser 1:55 to create this automation all right so 1:57 feature number two if you go to lovable 2:01 now and you go to your account settings 2:04 and you go 2:06 to settings here you're going to see 2:09 this option called Labs so if you turn 2:12 Labs chat modes onto on you'll be able 2:15 to now have a chat directly with our 2:18 large language model behind the scenes 2:20 that doesn't necessarily change your 2:22 interface or code whereas usually with 2:24 things like lovable or bolt or wind surf 2:27 when you ask for some form of command 2:29 usually changes the underlying code no 2:31 matter what you do but with chat modes 2:33 set to chat only you can ask a general 2:36 question like am I going deep enough in 2:38 my design do you have any 2:40 recommendations for how I should 2:41 actually structure my application so you 2:43 can have a direct back and forth with an 2:46 llm itself without necessarily changing 2:48 anything in your underlying application 2:50 infrastructure which is super helpful if 2:52 you're actually looking for guidance you 2:54 don't want to have to constantly go to 2:56 chat jbt or Claud to Think Through Your 2:59 Design or what you want to build next in 3:01 your actual project so once you have it 3:02 toggled on we can actually go into 3:05 lovable let's go back into the project 3:07 we just made and then you'll see this 3:10 little icon at the bottom right hand 3:12 side that says default and you'll see 3:14 here the default is just the typical 3:16 experience where you say some basic 3:18 commands and it usually does some form 3:20 of structural change to the application 3:23 but if you click on chat only you can 3:24 see here it says chat without making 3:27 edits to your project so if you click on 3:30 chat only I can ask it now a general 3:33 question like what do you think I should 3:37 do to make this app more unique and 3:42 usable by my team and now it should be 3:46 able to just answer that question 3:47 conceptually without necessarily having 3:49 to do a Next Step so now it says let me 3:51 suggest some features that could make 3:52 this task management app more team 3:54 oriented and unique and then it goes 3:56 through a series of recommendations and 3:58 at the very end it's says would you like 4:00 me to elaborate on anything we can start 4:02 by implementing them one at a time and 4:04 then you can go back to default mode 4:07 assuming you're inspired and you know 4:08 what you want to do next and maybe you 4:10 say you know what let's implement the 4:12 task assignment feature so even though 4:14 you're not directly changing The Code by 4:16 talking chat only you can use chat only 4:18 to get some inspiration or get some 4:21 guidance and then build on that guidance 4:23 from the conversation thread and say you 4:25 know what let's Implement number one all 4:27 right so that was feature number two and 4:29 now feature number number three is going 4:30 to be a doozy it's a bit unexpected 4:32 because most platforms don't have 4:34 anything like this just yet but now you 4:37 can use a figma as an inspiration to 4:40 actually build it directly in lovable so 4:42 you can have a design in figma and if 4:45 you're not familiar with figma I'll show 4:46 you what it looks like in a second but 4:48 imagine you have a design there you can 4:50 now hook it up take the underlying 4:52 design and then import it into lovable 4:55 to start building on existing designs 4:57 which I think for a lot of product 4:59 managers product designers and 5:01 especially entrepreneurs and business 5:03 owners this now enables you to take 5:05 static designs and literally bring them 5:08 to life if you're not familiar with what 5:09 figma is if we go to the website let's 5:12 go just figma over 5:15 here you're going to see that I'm going 5:17 to be logged into my personal account 5:19 but it's pretty much somewhere where you 5:20 can go and take a look and create 5:23 designs so before typically these apps 5:26 like lovable. deev existed in the old 5:29 world you would have a design team 5:31 create different wireframes of the 5:33 different screens of what your 5:34 application will look like and then once 5:36 you finalize that that's typically what 5:38 you'd give your development team as a 5:40 blueprint for what they should build in 5:42 terms of the SAS or the General 5:44 application and you can see here each 5:46 one of these wireframes is a separate 5:48 screen with individual components and 5:51 once you add them all together this is 5:53 where you get some form of semblance of 5:54 an app blueprint now to make this work 5:57 there are a few contingencies so step 5:59 one one is whatever figma design you're 6:01 trying to import into lovable has to 6:04 have Auto layout enabled so if you're 6:06 not sure what that is if you click on 6:09 any one of these little wireframe 6:10 screens here you'll see at the right 6:13 hand side there's the word layout so 6:15 you'll see here we have fixed dimensions 6:17 of how the layout is designed so it's 6:20 not necessarily going to Auto siiz to 6:22 any screen it's on and it's pretty much 6:25 fixed in the way it's actually 6:26 structured something like this wouldn't 6:29 really gel very well with lovable what 6:31 we need is for a specific component or 6:33 wireframe element to say Auto layout 6:36 Auto layout will basically reinforce the 6:39 fact that if this goes into a brand new 6:41 application or a brand new screen it 6:44 will automatically resize all the 6:46 components to fit it as well as possible 6:48 so if we go back to files right now and 6:51 we pull up another one of my designs I 6:53 have here and I hover over any one of 6:56 these so let's just take this screen or 6:58 wireframe okay so when I click on it now 7:00 you see on the right hand side here we 7:02 have Auto layout meaning it should be 7:04 able to resize and be adaptable in 7:07 different platforms that's step one so 7:10 that's checked off now step two is you 7:12 need to use figma's app store to install 7:16 something called a 7:17 builder. builder. will be the bridge 7:20 between figma and lovable so if you go 7:23 on figma plugin 7:27 store and you go here all right and if 7:30 we go to and write uh Builder IO you're 7:35 going to see this pop up which is 7:37 Builder IO AI powered figma to react 7:40 code and then you want to do open in and 7:43 then ideally whatever design you've 7:45 chosen and I'll show you how you can 7:46 actually uh get some designs for free 7:49 online that you can use and import into 7:51 your figma account once you've actually 7:53 used it in one design it should remember 7:55 in the plugins tab of your figma 7:58 instance so if you go into where we are 8:01 currently and you see these little block 8:03 icons these actions will allow you to 8:06 invoke different things one of those 8:08 things could be plugins so if you write 8:10 here 8:12 builder. you'll see matching actions 8:15 correspond to any plugins you have 8:17 installed so when you click on that 8:19 you'll get a popup and for the first 8:21 time for you you'll just have to click 8:23 through um it'll say a few different 8:25 instructions on what to do and then once 8:27 you get through there you should be able 8:29 to see export to code now you can go on 8:32 easy mode or precise I'm going to 8:34 explain the difference between easy and 8:37 precise when it comes to easy it's 8:39 trying to emulate some semblance of the 8:42 design you're exporting meaning there's 8:43 some margin of error it's not going to 8:45 look exactly like the design it's going 8:47 to be heavily skewed or very similar to 8:50 it with precise you're trying to push 8:52 the boundaries to make it as exact as 8:54 possible and although inherently 8:57 intuitively it might make sense to want 8:59 to do precise all the time with this 9:02 integration and lovable at least early 9:03 days this was just released it seems 9:05 that easy is typically the more easy way 9:08 to actually do this so once you're 9:10 toggled on to easy you can click on 9:12 export to code right here and then it 9:14 will take its time and go through a 9:16 couple different checks and then you 9:17 should get some lovable icon at the very 9:20 bottom of your screen that you can click 9:22 to then import that in your lovable 9:24 platform you can then get a popup like 9:26 this where it says choose your preferred 9:27 Vision which of these looks closer to 9:30 what you're looking for I think both are 9:31 good so let's click the first one here 9:34 and now you see this little shiny button 9:36 open in lovable so let's click on that 9:39 and then this will go through a couple 9:40 steps and then it will Ex export it 9:43 directly to the application so you can 9:45 see here it's a success so it seems like 9:47 this project is ready to import into 9:49 lovable and all we have to do is click 9:51 on launch project here and this will 9:54 initially look like it's doing nothing 9:56 but within a few seconds it'll start 9:57 importing that entire project into 10:00 lovable and there you go looks super 10:02 cool that we can actually import this 10:04 exact design from figma and we can now 10:07 start to add onto it so we can say can 10:11 we change the font of the panda 10:16 media at the very top in this case I'm 10:19 going to use the select little icon here 10:21 to click on Panda media so now it's 10:23 going to specifically Target that div or 10:26 that element to change the font so I'm 10:29 I'm just showing you this to show you 10:30 that you can actually not only bring it 10:31 to life but you can start to actually 10:34 edit things you can start to use the 10:36 normal functionality of lovable with the 10:38 head start of the design so you can see 10:40 here we made an Ever So subtle change to 10:42 the font and then I asked lovable to 10:45 double the size of the font a couple 10:47 times until it was big enough to really 10:49 note that it's different from the 10:50 original and then we can keep going and 10:52 say you know what let's remove this and 10:55 then change different components let's 10:57 make some things functional but you get 10:59 the general just there so the next thing 11:01 I want to show you is how do you 11:02 actually find a design online that you 11:04 can import start playing around with 11:06 this feature so if you go on something 11:08 like perplexity what you can do is is 11:11 you can say find me websites that offer 11:16 free templates for figma that are Auto 11:22 layout so typically where it will root 11:24 you to is the figma gallery where you 11:27 can actually retrieve some free 11:29 templates that are Auto layout format so 11:32 if you go down here you can go to click 11:35 on this link for example and you'll see 11:38 the figma community design templates and 11:41 one little nugget here is you'll notice 11:43 that the ones that are our autolayout 11:45 are typically labeled Auto layout so 11:49 what I'll write here is I'll say Auto 11:53 layout design okay and then that keyword 11:57 should show up so you'll see see here 11:59 Auto layout Auto layout um this one 12:02 looks pretty interesting not sure if 12:04 this one's Auto layout so let's take 12:08 something like this a crypto wallet UI 12:10 design you can now open this in figma 12:14 and now one thing we want to do is you 12:15 want to click on this cover and see if 12:18 it's its Auto layout doesn't seem like 12:21 it is so I'm going to exit this one 12:24 because seems a bit misleading so let's 12:27 go back and let's look for something 12:30 like 12:32 this okay so let's do open INF figma so 12:35 that one wasn't Auto layout format and 12:37 you'll just maybe have to check a couple 12:39 of them cuz sometimes they say Auto 12:40 layout but they aren't so let's click on 12:43 a couple so I'll open this in a new tab 12:46 I'll open this in a new tab and maybe 12:49 this in a new tab and let's do this and 12:52 let's see if any of them hopefully all 12:54 of them are truly Auto layout so if we 12:56 open this 12:57 inigma I'll do do this for this one as 13:01 well and this 13:03 one and then the last 13:07 one we should be able to quickly see 13:10 here so if I go on this layout Builder 13:14 and I hover 13:15 over hover over any one of these so 13:18 these look like they are Auto layout 13:20 components which is not bad not sure if 13:24 any of these are worth actually 13:25 importing into lovable so let's see what 13:27 else we got here so if we go to this one 13:30 the content 13:31 calendar I don't see auto layout on the 13:34 side so that doesn't seem no that's no 13:37 bueno this calendar is eligible so let's 13:40 go back to the calendar one we just saw 13:42 cuz out of the ones I just tested this 13:44 one seems to be the best let's click on 13:46 this one let's do auto layout what we're 13:48 going to do is we're going to go on the 13:50 Plug-In or action section on the bottom 13:52 screen if you've already used the 13:55 builder. integration it should show up 13:57 right here we're going to click it and 13:59 now we're going to do easy mode expert 14:02 to code it's going to do its thing then 14:04 let us know it might ask us which 14:06 version it wants us to import we'll 14:08 click on it and then we'll do the same 14:09 process where we import it into lovable 14:11 as expected it prompts us to choose one 14:13 of these versions let's pick this one 14:16 here let's do open unlovable it's going 14:18 to do its thing give us that open 14:20 project button and we'll do the same 14:21 thing all right so we got the launch 14:23 project button so we'll click on it 14:26 it'll import into lovable now and then 14:28 we should be a ble to have that same 14:30 experience where it looks very close to 14:31 the original calendar in this case icon 14:34 that we saw and on Landing we got an 14:36 initial error here that some code wasn't 14:38 ding so I just clicked fix and then it 14:41 was able to resolve it and we got again 14:44 some semblance of what what we 14:46 originally had so this was let's go back 14:49 here the original an actual calendar and 14:53 we got a a version of that if we open 14:55 this up in full screen okay so this 14:57 actually looks a lot better now we we 14:59 put in full screen this is much closer 15:00 to what we were going for if not exactly 15:04 what we're going for so let me just do a 15:06 quick compare it even has the icons in 15:09 the right places by date so you can see 15:10 here 3 11 17 all have blog and freebie 15:14 then same thing here so even the fonts 15:16 are identically the same so not that 15:18 this would be a useful springboard to 15:20 build an application out of this but you 15:22 get the general idea now and this is a 15:24 way you can navigate and be confident of 15:26 what designs might be import worthy 15:28 versus is not import worthy okay so now 15:30 I want to introduce the major hack I 15:32 mentioned at the very beginning of the 15:34 video so it's using a site called 21st. 15:39 deev that allows you to actually copy 15:41 paste prompts for components for entire 15:44 webpage designs so you can directly 15:47 recreate it in lovable. deev without 15:50 having to actually recreate the style or 15:52 screenshot the actual designs and say 15:54 try to replicate this using this trick 15:57 will let you recreate all kinds of user 15:59 interfaces that might be above you in 16:02 terms of being able to describe or 16:04 actually create the code for so 16:06 something like this when we actually 16:08 dive into it looks super sleek and super 16:11 futuristic even the font looks very 16:13 clean and these are components that if 16:16 you go to the right hand side here at 16:18 the top you will typically just see copy 16:20 prompt on 21st. deev but if you click on 16:23 here you can actually get this prompt to 16:25 recreate this UI optimized for lovable 16:29 for bolt for site Brew or versel so if 16:33 we go 16:34 to something that looks pretty let's say 16:37 this 16:38 one hello waves hero waves are cool so 16:41 let's take this and we'll copy paste 16:45 this let's do the optimize version for 16:48 lovable and let's copy The 16:50 Prompt let's go back to lovable and I'll 16:54 say 16:55 create an 16:57 application for 16:59 generative 17:01 AI apps using this UI design and then 17:07 what I like to do is put some quotes 17:09 paste the prompt and then there's one 17:11 component here that I noticed that 17:13 typically gives some headaches which is 17:15 this remember do not change the 17:16 components code unless it's required to 17:18 integrate or the user asks you to so I 17:21 like to remove this part of the copy 17:24 pasted prompt just because I noticed 17:26 that sometimes in working with this when 17:29 I ask for a change down the line for 17:31 some reason it doesn't listen to me and 17:33 it might be the fact that it's trying to 17:35 not make any changes and it doesn't 17:36 remember the user asks you to so I'm 17:39 just going to paste this as is here and 17:42 you're going to see it's going to be 17:43 able to recreate that design almost 17:46 seamlessly with no additional Intel for 17:48 me all right and you can see here we 17:50 have this application that looks pretty 17:52 much identical to this application just 17:55 with our tailored copy for whatever I 17:58 mentioned so the next step here is this 18:00 didn't even have buttons this had 18:02 buttons what if we wanted to make these 18:04 more shiny buttons so we could go back 18:07 to 21st. deev so you don't just have to 18:09 use stuff from landing pages you can use 18:12 individual elements like chats buttons 18:15 calendars cards Etc and if we go down 18:19 here if we go long enough we'll start to 18:21 see other components that are not just 18:22 landing pages so we're going to see here 18:25 some fancy little swivel designs you'll 18:28 see some some buttons so if we take this 18:31 here and I copy The Prompt for these 18:34 buttons I can go and say um for the two 18:38 buttons on the 18:40 screen make them a Sleek 18:44 blue format similar to this style and 18:49 then 18:51 again and then again I will paste this 18:55 and I'll remove this condition here and 18:59 then I'll let that go and we should get 19:01 two buttons that might not jive with the 19:03 background stylistically but it should 19:06 emulate this style here where if you 19:08 hover over they're nice and reactive and 19:11 very futuristic and there you go you now 19:13 have two of these buttons here get 19:15 started and learn more where we now 19:17 imported the button component from 21st. 19:20 deev again without any understanding of 19:22 the underlying code or even how to 19:24 actually describe this and you can see 19:25 when you actually open this in a brand 19:27 new tab it'll look look a lot sleeker 19:30 and we can try this with another design 19:32 just to really drive the point home so 19:34 let's close this let's create a brand 19:36 new chat let's go back to the dashboard 19:39 all right so I'm going to reload the 19:41 specific URL cuz sometimes I get lost in 19:43 the navigation here so I'll just load 19:45 21st dodev so we see all the components 19:48 and then let's try to emulate another 19:50 one of these designs here so if we go 19:54 into something like this this globe this 19:58 shiny moving Globe we can now copy this 20:02 prompt and now we're professional so 20:04 we'll just say create this UI we'll then 20:08 paste this command remove the remember 20:11 part take that out paste that there and 20:14 we should be able to get it and then you 20:15 can actually take it from there meaning 20:17 you can start to say you know what let's 20:18 create a button that has a weight list 20:20 let's change the copy so it looks a 20:21 certain way so so we'll do a couple 20:23 edits of this just to show you how that 20:24 works and then move on to actually 20:26 building our application so we get the 20:28 spinning globe design and we can make 20:31 some changes and say one let's say 20:34 changes here let's say 20:37 one let's F focus on making it full 20:43 screen right now it only 20:46 takes one small part of the screen and 20:49 let's actually screenshot this so it can 20:52 actually 20:54 understand the relative size and then 20:56 we'll say let's 20:59 make 21:01 globe let's make the 21:03 word globe and I'll put it in quotes so 21:05 it doesn't actually change the globe 21:07 itself the spinning orb 21:10 into join our weit 21:13 list and then add a button that 21:20 says enter your 21:23 email and after the user enters their 21:28 email 21:29 they can click 21:32 join and it will submit the response so 21:37 you can see here the proportions are 21:38 still a bit off but we have this enter 21:41 your email we have the join we have the 21:43 spinning orb and we could probably by 21:45 just clicking on this div say make this 21:49 span the entire 21:51 screen just so it actually collapses and 21:54 we don't have any white space or spots 21:56 so most likely this might take one time 21:59 or a couple iterations to do but you can 22:01 imagine that just describing or bringing 22:03 in this image or icon of a spinning orb 22:06 alone would probably take you 10 or 15 22:09 prompts if you knew what you were doing 22:11 so it didn't necessarily actually follow 22:13 what I just said but this actually looks 22:14 better than what I actually suggested so 22:17 this looks pretty sleek and obviously 22:19 the next step you could take is create a 22:22 superbase instance where you can create 22:23 a database to start storing emails for a 22:27 weight list then you can start nurturing 22:28 in a list and you can take it from there 22:30 and you can do this in literally under 5 22:32 10 minutes and build something that 22:34 looks like you actually hired an actual 22:36 developer for so let's take this hack 22:38 and actually apply it to build our own 22:40 app and like I said before we're going 22:42 to do something a bit more sophisticated 22:45 than a to-do list app which is to 22:47 actually create a prompt engineering app 22:49 where you can ask for a task it will 22:52 take that task make it into a prompt and 22:54 then ideally you can click a button and 22:56 simulate how that newly formed prompt 22:59 will actually perform on a particular 23:01 large language model so let's go back to 23:03 level B one last time here and let's go 23:06 to 2dev and look for some 23:09 inspiration so if we go down I want to 23:12 go for something more minimalistic in 23:15 nature so if we go 23:17 through let's wait for some more 23:19 components 23:21 here that's kind of minimalistic I like 23:25 this Bento grid looks very clean let's 23:28 look into this 23:30 one so you can see here when we hover 23:32 over all of these 23:35 components you can click in looks clean 23:38 so let's take this okay and now I'm 23:41 going to give it some form of more 23:43 robust prompt so it knows kind of what 23:45 I'm going for and then we'll actually 23:48 apply this template so I'm going to 23:50 write I want to create an 23:53 application where I can write a request 23:59 I have or a task I want to complete and 24:04 the application will take my task and 24:08 then send it to a large language model 24:14 open AI to create a prompt out of it and 24:19 then use said 24:22 prompt to simulate how the prompt would 24:27 output 24:29 if 24:31 ran I'm going to show you an 24:36 inspiration design that I'd like to use 24:41 as the foundation for our 24:43 design and I want to 24:47 have three 24:50 sections one 24:53 section that allows me to enter the task 24:57 I want 24:58 and then turn that into a prompt let's 25:02 make this a bit cleaner here so let's 25:04 say sections and then I'll say one okay 25:09 one section that lets me do 25:11 this one section that lets me copy paste 25:17 a prompt and then simulate the result of 25:23 the 25:24 prompt and then one that lets 25:29 me 25:30 simulate a 25:32 conversation with said prompt so the 25:37 last one will be more complex we'll see 25:39 how far we get let's now forgot the most 25:43 important part here's the 25:46 design to use as your base so now I'll 25:50 put this in 25:52 quotes Okay and then we'll remove that 25:54 remember component and then we'll send 25:57 this over and we'll see what we get 25:59 started with the next natural step is to 26:01 be able to enable the llm functionality 26:04 we're going to create a superbase 26:06 database because we're going to have to 26:07 create Edge functions to support all the 26:10 functionals we need all right so we got 26:12 initial little error here had a bit of a 26:14 hiccup we clicked fix we got to this 26:17 stage where it looks like it from this 26:20 perspective but what it did do that I 26:22 didn't expect it to do is actually put 26:24 the functionality directly on the card 26:27 so now I want to say 26:29 we want the cards to look 26:34 minimalist and spam the entire 26:38 screen and then the functionality of 26:42 entering the prompts or simulating the 26:46 prompts or 26:48 conversations happens when you click on 26:52 the underlying 26:55 cards button to 26:58 go to the next page so I'm just going to 27:02 go back here so if we go to here it says 27:05 learn 27:06 more I'm just going to use that as 27:09 inspiration and paste it here and say so 27:13 if you 27:14 see in the image when you hover over a 27:19 card it should 27:22 say execute 27:25 task and then we'll paste that it should 27:27 look a bit better it should have a 27:30 secondary page now that leads from the 27:32 cards that make the original page more 27:34 minimalist to a page where we can 27:36 actually execute these workflows so now 27:38 they're all resized and if you hover 27:40 over you can see execute task and if I 27:42 click on one of them hopefully it should 27:44 lead me to a 27:45 page okay so if we go 27:49 back let's go to here I'll just say 27:53 immediately one changes one the title is 27:59 in the 28:02 background now I can't see the actual 28:07 text on the first 28:10 page when I click execute task I don't 28:15 see a second page now one thing I 28:18 noticed is I don't know if I fully 28:20 waited for the app to run before I 28:22 clicked on those buttons so let me go 28:25 back to index here refresh 28:28 refesh and let me just click on execute 28:31 task okay no none of the pages are 28:33 working I don't see a second page for 28:35 any of the 28:37 cards they should lead to being able to 28:42 enter the respective input for the 28:46 respective 28:48 functionality we need okay so now when 28:50 you click on any of these cards you do 28:52 see a second page now whether or not 28:54 looks ugly is a secondary issue let's 28:57 just check that all of the open up to 28:59 something okay they do so far let's 29:02 check the last one okay cool so next 29:05 issue is this glaring title here so 29:08 let's actually screenshot the issue 29:11 let's go back to here okay and I want to 29:13 screenshot it versus selecting a 29:14 specific element just so you can 29:16 actually see the overlap so there's an 29:21 overlap 29:23 overlap with the 29:25 title over the cards 29:30 make the cards slightly shorter to make 29:33 room for the 29:36 title on the secondary Pages please make 29:42 the UI of 29:45 the functional prompt 29:48 inputs not only more userfriendly 29:53 but prettier and better fit to the 29:59 page as some of the elements seem to be 30:05 going off of the page so that should 30:09 hopefully resolve this title issue once 30:11 and for all and obviously right now I'm 30:13 just focusing on the shell of the app 30:15 versus the underlying functionality 30:17 because once this looks like what I want 30:19 it to do then we'll actually focus on 30:22 integrating the features we need so this 30:23 looks way better now we have the title 30:26 separated when we click on execute cute 30:28 task this doesn't look awful if I check 30:30 the other ones this doesn't look 30:32 awful perfect none of them actually look 30:35 awful so that's a big M stolen forward 30:38 so now step two before we even talk 30:41 about functionality let's create a 30:43 superbase database and if you've never 30:45 used this before I would strongly 30:48 recommend at this point watching my 30:49 first video on how to master a lovable 30:52 to see how this works but the crash 30:55 course is you need to be able to create 30:57 a database to store you know the 30:59 different prompts we enter in this case 31:01 to be able to enable functionality like 31:03 calling open AI so we're going to create 31:05 a new project here I'm going to call it 31:07 let's call it um prompt buddy and then 31:11 we're going to make a password for our 31:14 super base instance we're then going to 31:16 click create new project this will take 31:18 anywhere from 3 to 5 minutes and once 31:21 it's done I'll be able to go on 31:23 superbase here and then over prompt 31:25 advisors and be able to select the the 31:28 right super base instance so let's go 31:30 back and click on create new project and 31:33 we'll flick back as soon as it's ready 31:35 all right so now our prompt buddy 31:37 superbase is set up and now if we go 31:39 into lovable into superbase we should be 31:43 able to see it here and we do so we'll 31:45 click on connect click connect again and 31:48 then this will allow us to hook up to 31:49 super base and now we can start 31:51 integrating open AI to make this 31:54 functional instead of just a shell all 31:55 right so super base is connected so now 31:58 I'm going to say we want to enable 32:02 functionality for all of these 32:06 components so I'm going to do it very 32:08 structured so to maximize the likelihood 32:11 of to actually get this working from the 32:13 first go even though most likely it'll 32:15 take a few iterations let's first 32:17 actually tell it to integrate open a in 32:20 general and to ask us for our open AI 32:22 key so we get that out of the way we 32:25 want to use gbd 40 mini for all the 32:31 functionality 32:34 here let's actually let's 32:39 integrate this into 32:43 superbase so set it up and ask me for my 32:48 open AI key so now this should signal to 32:51 it that it should start doing that work 32:53 on the back end it should prompt us for 32:55 an API key and then we should be 32:57 hopefully be able to now delineate all 33:00 the tasks for the different cards so 33:02 we're going to do add API key going to 33:04 do 33:05 submit and then it didn't set up the 33:08 actual changes yet to superbase with the 33:11 draft of our database but hopefully I'm 33:14 trying to get away from doing that so 33:15 that I can first tell it exactly what 33:17 the goal is before we worry about 33:19 setting up the actual databases 33:21 themselves all right so it looks like 33:22 we're ready to use these little 33:24 functions here so let's actually tell it 33:26 exactly what we're going for so let's 33:28 tailor each one of these function 33:33 sections task to 33:36 prompt this section should allow the 33:40 user to enter any 33:44 description of a task and then upon 33:50 submitting the behind the scenes prompt 33:55 should be along the lines of this and 33:59 then I'll go with my typical prompt for 34:01 meta prompting which is you are a prompt 34:04 engineer you write detailed prompts 34:09 based on the basic user information 34:14 you're 34:16 provided output The Prompt taking the 34:21 task that is provided to you in layman's 34:25 terms and transform it 34:28 into a robust yet cinct prompt in 34:34 markdown and then once we return this 34:39 prompt from open AI we parse the mark 34:44 down properly and display it in the view 34:49 so that's section number one prompt 34:53 simulator this should be easy it should 34:58 just allow the user to copy paste any 35:03 prompt and upon 35:07 submission they 35:09 can yield oo fancy word the output of 35:14 said 35:15 prompt okay so that one should be 35:17 straightforward and then conversation 35:18 simulator similar to prompt 35:22 simulator except you can have a back and 35:28 fourth 35:29 conversation where the last few messages 35:34 are injected in the memory of the 35:38 conversation to create a thread so that 35:42 it's continual and each response from 35:47 the AI has the context of the entire 35:53 conversation in scope all right so try 35:57 to be as detailed as possible here to 35:59 increase the likelihood that we don't 36:01 have to spend another hour on this so 36:03 let's send this over and see what 36:04 happens all right so it ran I tested it 36:06 initially wasn't working clicked fix and 36:09 now we have a version where we write 36:10 something like write a SEO blog for my 36:15 mortgage brokerage business if any of 36:20 that English makes sense and it should 36:22 actually generate using gb40 mini output 36:25 in a way that we can easily copy paste 36:27 but obviously it has a couple flaws and 36:28 you'll see that shortly you can see it 36:31 is in markdown but it doesn't actually 36:32 wrap the text so the first thing I'd 36:34 want to do is wrap this text so let's 36:37 put this 36:38 here and I'll say it works but we should 36:43 wrap the 36:46 text all right so let's in the meantime 36:49 whilst making that update see if we can 36:51 sneak and see if the second 36:53 functionality works now since the first 36:55 one didn't work the first time wondering 36:57 if the second one will work the first 36:59 time and it had an error as we expect so 37:01 let's try to fix this while we go back 37:04 to the other card and see how this 37:07 actually changed right now so let's say 37:09 write SEO blog something super generic 37:12 just to see if it actually worked and 37:15 that's one thing I like to do is instead 37:16 of just twiddle my thumbs as I try to 37:18 test other functionality while I'm 37:20 waiting for other things to fix so in 37:22 this case this looks a lot better the 37:24 text is wrapped we see the Bolding we 37:26 can easily C copy paste this one should 37:28 be good so let's hope the second one 37:31 comes along properly because it should 37:33 be straightforward and obviously one 37:35 thing we're missing here is some form of 37:38 storage so that we can see the last 37:40 prompt we did but one step at a time so 37:43 let's just do this again write an SEO 37:46 blog it's a good way for us to also 37:48 battle test that it works consistently 37:51 so we should get some form of output 37:53 here we'll take the copy we'll paste in 37:55 the next card and see what happens we'll 37:58 take this we'll go back we'll go to 38:00 prompt simulator we'll paste this and 38:04 we'll see if that works now yeah one 38:08 thing that I noticed is it kicked me out 38:09 of the actual tab so I'll click this 38:12 again and one thing I'm looking at is 38:14 I'm wondering if any of these little 38:15 special characters like that are called 38:17 markdown are going to cause any issues 38:21 here so it said prompt simulated 38:23 successfully but I don't actually see 38:25 anything uh okay so I stand corrected so 38:30 this is the input and this is the output 38:33 which is actually a pretty long output 38:34 it already did the wrapping so that's 38:36 good so this one seems to be working now 38:39 the one I'm the most skeptical that 38:41 would work on the first time is this one 38:43 so let's type our prompt and let me just 38:47 take out this little special character 38:48 stuff so we don't have to deal with that 38:51 let's take that out uh that's that's our 38:55 system prompt so let's send that see 38:59 what happens if anything happens so as 39:01 expected we have an error here I'm just 39:03 going to see what kind of error we have 39:04 so okay so it seems like the edge 39:07 function is not 39:08 working so we can make some changes 39:11 there in the meantime so this is a time 39:14 where I'm going to actually start typing 39:16 my next prompt since I already know what 39:18 it's going to be assuming we can 39:20 actually get this to work is it's going 39:22 to be can you 39:26 ensure that all prompts that are tested 39:31 are 39:32 generated 39:34 tested or chats that 39:37 occur are stored in 39:40 superbase okay so I'm just going to copy 39:42 paste this elsewhere and we'll use it 39:44 for later assuming we can get past our 39:46 little hiccups here all right so 39:47 apparently things are fixed so let's go 39:50 to here and let's create a basic prompt 39:53 and let me just say as a user Make Me an 39:57 SEO blog about generative AI That's only 40:03 one paragraph okay and let's just send 40:06 that so now what's interesting is it now 40:09 looks like an actual chat card okay 40:11 that's a pretty quick response and my 40:14 ultimate test to know is there context 40:16 being held in the conversation is can I 40:19 say something like make it shorter and 40:21 it will actually remember what it should 40:23 make shorter so if I say make this into 40:27 only one sentence okay and then I send 40:31 that it should show that here 40:36 okay so that that was pretty quick uh 40:40 let me try this again uh make this into 40:44 only 10 40:46 words 40:48 okay okay so it's hyper quick so it's 40:50 actually doing a really good job it's 40:52 keeping the context and then I'm going 40:54 to say Do You Remember what my original 40:58 request 41:00 was okay and I'll send this 41:03 here yes your original request was an 41:05 SEO blog about generative AI but 41:07 condensed into only one paragraph So 41:09 what this tells me is it's keeping 41:12 probably the last couple messages into 41:14 context which for me is pretty good 41:16 given that we're just using this as a 41:18 starter but yeah so all these now quote 41:20 unquote function and the last thing we 41:22 can do is copy paste that prompt so 41:25 we'll go here we'll go and say can you 41:27 Ure all the prompts that are generated 41:28 tested or chats that occur I'll say the 41:31 threads are stored in super base uh 41:35 let's let's be specific for task to 41:38 prompt and prompt 41:41 simulator we should be able to see the 41:46 past prompts that have 41:49 been 41:50 created and past prompts that have been 41:56 simulated 41:58 we don't need to show past chats for 42:01 conversation 42:03 simulator so this is important to be 42:05 specific about what's worth keeping on 42:08 the actual user interface versus just 42:10 keeping in the back end okay so 42:11 apparently it's ready to go so if you go 42:13 to here and we say create 42:17 a business plan structure for my law 42:22 business and I click on generate prompt 42:26 hopefully we should not just get it 42:28 generated prompt but some form of 42:30 History if we refresh the 42:33 page so we got a little error here we'll 42:35 try to fix it um let's see what the 42:37 error is looks like some form of issue 42:42 with the URL so we'll just resolve that 42:44 and go from there so the next error I 42:46 got was this where it says please sign 42:49 in to generate prompts even though I had 42:51 no authentication to begin with so this 42:53 is something that pops up sometimes in 42:55 lovable where you just have to tell it 42:57 you know what I'll be the only user 42:59 assuming you are the only user so I'll 43:00 show you the exact prompt I actually set 43:02 so it says please sign in to generate 43:04 prompts we need to get rid of that so I 43:06 can use it I'll be the only user and 43:08 you'll notice here that it has something 43:10 called rowle security where it has like 43:12 some fail safes that you can't execute 43:15 some functionality Sometimes using 43:17 superbase unless it removes that 43:19 restriction so once it removed that 43:21 restriction you can see here I just said 43:23 make me a law blog it makes the blog and 43:26 at the very bottom bottom now we have 43:28 history that's stored and if we go back 43:31 into superbase click into table editor 43:34 and you can see here generated prompts 43:38 you can see the original text I had and 43:40 the generated prompt and what's cool is 43:42 you can even see here this little unlock 43:44 icon and this is one hint you can use if 43:46 you get that error is if you go to 43:48 superbase you'll see this lock icon 43:50 telling you that there's some form of Ro 43:52 level security that you probably have to 43:53 unlock so the whole point of this was to 43:55 show you how to build something that's 43:56 not a to-do list app that's actually 43:58 functional and helpful using 21st. 44:01 deev's designs as the foundation to take 44:03 care of the pretty part while you focus 44:05 on the functionality part if you found 44:07 the tips tricks and demo helpful let me 44:09 know down in the comments below 44:10 hopefully this was helpful for you and 44:12 gave you some nuggets to make you that 44:14 much better at doing text to web apps 44:16 using lovable. I'll see you in the next 44:18 video and like share and subscribe