The Future of AI App Building Explained (n8n + Supabase + Lovable)

0:00 Hello everyone. In this video, I'm going 0:02 to take you through some of these new AI 0:04 builders that are blowing up on social 0:05 media right now. I'm talking about tools 0:07 like Lovable, Bolt, and Cursor. I'm 0:10 going to talk you through exactly what 0:11 they are, how I think you can use them 0:13 the best, and what I think is FUD and 0:15 what is actually gold and something that 0:17 you should really take note of to make 0:18 the most of this booming industry, which 0:20 I think is going to 10 times at least in 0:23 the next year alone. If you're not 0:24 familiar with who I am, my name is Adam. 0:26 I'm a top-rated freelancer on Upwork, 0:28 which is where I started my AI and low 0:30 code career. Doing over 50 projects 0:31 there and becoming expert vetted, I 0:33 transitioned into a global low code 0:35 development agency called APG Software. 0:37 And since then, we've developed over 250 0:39 projects in the low code and AI space 0:41 for clients all over the globe. So, I've 0:43 been in this industry for a while. I'm 0:45 pretty familiar with some of the ways 0:47 that people are building these products 0:48 in rapid time frames. And I wanted to 0:50 give you my opinion on exactly what 0:52 these tools are and where I think 0:54 they're going to be headed. So, the way 0:55 I've laid this out is we've got our new 0:57 AI builders on the right of this 0:59 whiteboard here. And I'm going to talk 1:00 through what they are, the benefits, 1:03 some of the trade-offs that exist there 1:04 right now. I'm going to compare that to 1:06 some of the traditional ways of building 1:08 things and also some of the low code 1:10 ways which work and still are really 1:12 popular, but why I think this is 1:14 beneficial in certain areas and this 1:16 approach is beneficial for other areas. 1:18 And then I'll talk you through what I 1:20 think we're going to end up with for 1:22 building really scalable rapid 1:23 applications in the next year or so. So 1:25 I'll go through walk you through the 1:27 pros and cons of each of these and then 1:28 talk through what I think the best stack 1:30 is going to be in the future. And if 1:32 you're someone that's looking to get 1:33 into this space and learn some tools, 1:35 I'd really recommend exploring this kind 1:36 of combination. Now to start off with, 1:39 let's talk about these tools, what 1:41 they're useful for, and can you actually 1:42 build million-doll applications that can 1:44 scale on these tools. Now, what these 1:46 tools are really, really good for is 1:48 rapid prototyping. Now, if you've played 1:50 around with any of these tools and 1:51 you've put a wellthoughtout prompt into 1:53 a tool like Lovable, you'll see a pretty 1:55 incredible response come out. You'll get 1:58 a number of pages generated. They'll be 1:59 mobile responsive. You can click around 2:01 in there and those prototypes used to 2:03 take months just to build. So, what used 2:05 to take months can be done in around 2:08 about a minute these days. And so, this 2:10 is really, really powerful. Now, there's 2:12 a lot of benefits to this and I think 2:13 this is really going to shake up the MVP 2:16 development industry in general and 2:17 there's a number of reasons why 2:18 something like this is preferable to 2:21 what used to be what low code tools are 2:23 used for which was rapid prototyping and 2:25 MVP development. The benefits of this 2:27 over using tool like bubble to build an 2:30 MVP are you get access to your source 2:32 code. Now, we work with a lot of startup 2:33 founders, a lot of which have investors 2:36 or they're in some sort of startup 2:38 accelerator, and having access to the 2:40 source code is really important for them 2:42 because if they were to scale to a 2:43 million-doll application, they don't 2:45 want to be locked into a vendor and not 2:46 have access to their own source code. 2:48 So, whilst I love Bubble, I think it's a 2:50 really powerful tool. One of the things 2:51 that scares a lot of founders from using 2:54 a tool like this is the fact that they 2:56 don't have access to their own source 2:57 code. And if something was to happen to 2:59 a company like Bubble, it may damage 3:01 their application. Now, I don't think 3:03 this is a big issue for anyone that's 3:05 trying to scale an app, and there's a 3:06 million ways you can get around 3:07 something like that. A lot of companies 3:09 trust big CRM platforms and all sorts of 3:12 big platforms to store their data. I 3:14 think that's really something that 3:15 you're going to have to sacrifice in 3:16 your company in some way or another. But 3:18 in saying that, for building a an 3:21 application, this is a really big plus. 3:23 In addition to that, you could make some 3:25 pretty rapid iterations. So you can put 3:26 it out for feedback and these AI 3:28 builders can make changes rapidly. And I 3:30 actually think that pretty soon founders 3:32 will not need software agencies to do 3:34 this MVP testing stage. So typically 3:37 what happens is a founder gets funding, 3:39 they approach a software development 3:40 agency, they go this is our idea and the 3:42 software development agency gets that 3:44 idea and builds the MVP because founders 3:47 often are not that techsavvy and they 3:49 don't know how to convert their idea 3:50 into a tech product. This is going to 3:52 change the game because they can use 3:54 natural language to prompt to some sort 3:56 of AI tool and get an MVP without 3:59 needing the input of a software agency 4:00 at all. Now, there are some red flags 4:02 and I'll talk about that in a second, 4:04 but MVP development and validating ideas 4:07 is about to be something that's 4:08 accessible to anybody and can be done in 4:11 just a couple of minutes, which is an 4:13 absolute gamecher to how things have 4:15 been operating for the past 20 years. 4:17 And what low code solutions like bubble 4:20 were trying to do was enable people who 4:22 had no tech skills to build apps. In 4:24 saying that though tech is very 4:26 complicated and if you don't understand 4:27 the fundamentals of how tech works, you 4:30 are still going to have trouble using 4:31 building scalable applications using a 4:33 tool like bubble because it gives you 4:35 the customization to build scalable apps 4:37 but you need to understand how to set 4:39 tech up properly because what you can 4:41 easily do if you're not familiar with 4:42 tech is build what they call spaghetti 4:44 code or spaghetti apps. So apps that are 4:46 built in a way that is not scalable and 4:48 workflows and and your database 4:50 structure and everything is not built in 4:52 a scalable way and early stage founders 4:54 who haven't been in the tech industry 4:55 for a while tend to make a lot of those 4:57 critical mistakes. And what I can see 4:58 happening here is that being magnified 5:01 because these founders are generating 5:03 applications. They don't actually 5:04 understand what's being done in the 5:06 background. They don't understand the 5:07 code that they're generating. And so 5:09 whilst this is going to be awesome for 5:10 building MVPs, building scalable 5:13 applications on top of that is where the 5:14 challenge is going to be. And so what 5:16 I've done up here is I've said 0 to 0.8 5:19 very very quickly, which is awesome. And 5:21 for getting just a couple users in your 5:23 local area or validating a market with 5:25 some beta users, this is probably going 5:27 to be enough. But from getting from 10 5:30 users to 100, a,000, 10,000, whatever 5:33 the case is, that's going to be a real 5:35 challenge on tools like Lovable Bolt and 5:37 Cursor at the moment. Now, I'm not 5:38 saying that that's not going to change 5:39 possibly in the next couple months, but 5:41 currently for a founder that has no tech 5:43 skills, they can get to here and then 5:46 they have to hire traditional tech team 5:47 and it's going to be slower from that 5:49 point forward. And some of the things to 5:51 be aware of when building with these 5:52 tools, one is AI errors and rabbit 5:55 holes. So, people that are building with 5:57 this, you can build really, really 5:59 quickly, but you don't understand the 6:01 code that's happening behind the scenes 6:03 and the way certain processes are being 6:05 undertaken. And what can happen there is 6:07 you can have some pretty serious issues 6:09 with privacy. So data privacy, who can 6:12 view what data, what data is being 6:14 exposed in the client, what data is not 6:15 being exposed in the client, where data 6:17 is being sent. You might be making a 6:19 number of API requests and you don't 6:21 actually understand what data is being 6:23 sent and received, what's being exposed 6:24 and what's not. So I can see a lot of 6:26 issues popping up at least at the moment 6:28 in the privacy space. And next up is 6:30 rabbit holes with the AI agent. Now you 6:32 might build 0 to 0.8, But then when 6:34 you're trying to set up possibly a third 6:36 party connection to let's say Stripe for 6:38 example or you're trying to make small 6:40 changes sometimes the AI can get really 6:42 confused and it can have trouble 6:44 debugging your issue and fixing your 6:46 code and this leaves founders pretty 6:48 helpless because they're sitting there 6:50 with a whole stack of code that they 6:52 have no idea how it works and being able 6:54 to debug these things is really really 6:56 difficult. So, as the LLMs get better at 6:58 writing code, this issue is going to get 7:00 smaller and smaller and smaller, but it 7:02 does exist currently. And at the end of 7:04 the day, you're still using code. And 7:06 this is what I'm going to talk about in 7:07 terms of the future I think this is 7:09 going. But at the end of the day, you're 7:11 still using code. And software 7:12 developers are expensive. And building 7:14 with code is expensive. The reason for 7:17 that is because it's still slower to 7:19 build with code than to build with 7:20 visual development interfaces like 7:22 Bubble has, especially for its back end. 7:24 It's a lot easier to visually look at 7:26 bubbles backend workflows and understand 7:28 exactly what's happening than to look at 7:31 some code and understand exactly how 7:32 that's performing. So this still has the 7:35 advantage in that space. So to talk 7:37 about where I think app development is 7:39 going to go between these two approaches 7:42 to building apps in the future is I 7:45 think we have three different kinds of 7:46 apps. We have SAS applications. We have 7:49 what I call two-sided marketplaces. And 7:52 we have internal tools and CRM. Almost 7:54 all of the apps that comes through our 7:55 agency door is one of those three 7:57 categories. And in my opinion, I think 7:59 SAS applications and two-sided 8:02 marketplaces will start to get built 8:04 more on tools like Lovable, Bolt, and 8:06 Cursor than on tools like Bubble. And 8:08 the reason for that is because these 8:09 apps are really hit and miss. A lot of 8:11 these apps fail in the first 5 years 8:13 because they're really difficult to get 8:15 off the ground. But if they do get off 8:17 the ground, they can often scale to 8:18 hundreds of thousands or millions of 8:20 users. And so having access to a 8:22 scalable tech stack like Superbase and 8:25 using Lovable for your front end is 8:27 going to be more preferred by founders 8:29 than being locked into some sort of tool 8:32 like Bubble, not getting access to their 8:34 source code and having to deal with a 8:37 backend that is not as efficient as 8:39 something like Superbase and doesn't 8:40 scale as well price-wise. So, I think 8:43 we're going to see MVPs for SAS 8:45 applications and two-sided marketplaces 8:48 going to be really common in this space. 8:50 Where I think tools like Bubble still 8:52 have the advantage is for internal tools 8:54 and CRM. And the reason for that is 8:57 internal tools and CRM typically have a 8:59 couple hundred users, but they're not 9:00 trying to scale to millions of users 9:02 because it's kept internal to a company 9:04 and they're just supposed to process a 9:06 company's internal operations. And 9:08 what's really important in a company's 9:09 internal operations is exactly how the 9:11 data is moved around and stored. And 9:13 what bubble and no code tools still have 9:15 a large advantage over over these tools 9:18 like lovable bolt and cursor is one you 9:21 get truly no code which means you can 9:23 understand what's happening especially 9:25 in your backend workflows a lot easier 9:27 than with a tool like lovable and 9:29 superbase. And the benefit of this is 9:31 companies have really specific 9:32 processes. is that when a certain user 9:34 clicks this button, they want to send 9:35 data to three different separate third 9:37 party tools and then post data into 9:40 their database and then upsert that into 9:41 a vector database and all sorts of 9:43 technical things. And so you really have 9:45 to understand a business's processes and 9:47 be really specific in those backend 9:48 workflows. And that takes a deep 9:50 understanding of processes and it's 9:52 going to be something that's hard to 9:53 communicate to an AI tool like lovable 9:55 bolt and cursor at least at this stage. 9:57 So I think there's still the advantage 9:58 there. The other advantage is that these 10:00 tools handle all of the hosting server 10:03 management. It's one-click deployment. 10:05 And so for an internal company that just 10:07 wants a simple internal tool that's 10:09 going to sit there, it's going to do 10:10 their processes and they want to have as 10:12 little maintenance as possible. I think 10:13 you're still going to have some benefits 10:15 in this area for that industry. And the 10:17 other thing is is a single source of 10:18 truth. So you've got your database, your 10:20 front end, your back end, and even your 10:21 mobile application all sitting in one 10:23 place. It's simple to manage. It's easy. 10:26 It's visual to see. And that's what a 10:28 lot of internal companies are looking 10:30 for. They want simple, easy to manage, 10:32 but still do all of the processes and 10:34 handle all the complexity that is in 10:35 their business. And so for these 10:37 companies, vendor lockin is not as big 10:39 of an issue because they're not 10:40 predicted to have the scale to millions 10:43 of users that typically SAS and 10:46 two-sided marketplaces have to worry 10:47 about if they can get off the ground. 10:49 They just want something that's simple, 10:50 easy to manage, and does what their 10:52 business needs. And so I think tools 10:54 like bubble and no code are still going 10:56 to be pretty strong in that space for 10:57 the foreseeable future. Now that being 11:00 said, what I think could be a really 11:02 powerful benefit in the future and if 11:04 you are looking to get into this space, 11:06 learn some low code tools, learn some ad 11:08 tools and start to sell solutions, a 11:10 stack like this, I think is going to be 11:12 really really powerful in the next few 11:14 months and years. Now what have I got 11:16 here? Where I think these tools like 11:17 Lovable and Cursor are still lacking at 11:20 this current point in time is having a 11:22 really nice and easy to see back-end 11:24 workflow builder that tools like Bubble, 11:27 Make.com, Zapia, and N8N have. So, what 11:30 I think is a really powerful stack is 11:32 using a tool like Lovable for your front 11:34 end to be able to build really modern 11:36 looking user interfaces for all your 11:38 different screen sizes rapidly. Now, 11:39 Lovable is a great solution for your 11:41 front end. Superbase is a great solution 11:43 for your database. It's very scalable. 11:45 It handles authentication. It handles 11:47 whatever you need to handle. It's fairly 11:50 userfriendly and it's easy to use. But 11:52 setting up edge functions and stuff like 11:54 that. You're dealing with code and there 11:55 is a bit of a learning curve there. And 11:57 we've got these tools that have come out 11:58 now like tools like NAN which are 12:00 essentially where you can build backend 12:02 operations to process data. You can 12:04 self-host these solutions and they're 12:06 really powerful for building AI agents 12:08 and integrating AI into your back end. 12:10 So if you were to use some sort of 12:12 backend like NAD, call it via API 12:15 through your tool like Lovable and post 12:18 the data from NAN into a database that's 12:20 in Superbase. I think that's a really 12:22 scalable tech stack. It's fairly no 12:25 code, so it's fairly easy to understand 12:27 and it's really optimized for AI and AI 12:30 development. So something like this I 12:32 think is a really solid tech stack and 12:34 it's something that our business will 12:36 look to shift to if we start to build 12:38 more SAS and two-sided marketplaces in 12:41 the next few months. We're fully aware 12:43 that this is the direction that tech is 12:45 heading. But whether or not now is the 12:47 right time to jump into this, I still 12:49 think there's a couple of months left 12:52 where I'm just waiting for the LLMs to 12:53 get a little bit more accurate, waiting 12:55 for some of these tools to fix a few of 12:57 the little nuances that are currently 12:59 still in them. So, we're still building 13:00 a lot of these solutions with Bubble, 13:02 but we're looking to niche down into 13:03 building internal tools with Bubble cuz 13:05 I think it's still a great place for 13:07 companies to build simple internal 13:09 tools. But SAS and two-sided 13:11 marketplaces and even to internal tools 13:13 in the future, I think a tech stack like 13:15 this is the future of where software 13:17 development is headed and how you can 13:19 build rapid MVPs in one or two weeks. 13:21 So, this is a really, really solid tech 13:23 stack. In saying that, I'd love to hear 13:25 your opinion on where you think the 13:27 future of tech is going for building 13:29 apps for companies from 1 to 100 13:32 employees and rapid growing startups and 13:34 what your opinion on a really solid tech 13:36 stack for building these MVPs on 13:39 scalable infrastructure in just a couple 13:41 of days or weeks. Feel free to let me 13:42 know in the comments what your opinions 13:44 are. If you're interested in staying up 13:45 to date in everything low code and AI, 13:47 feel free to join my completely free 13:49 community in the link in the 13:51 description. We have weekly discussions 13:52 where we talk about all things AI and 13:54 low code. And yeah, I hope you found 13:55 this useful. Let me know in the comments 13:57 your thoughts on the directions that 13:58 this tech is going.