I rebuilt my app with AI. Here's what I learned.
Description
I tested out Lovable by giving it some prompts to rebuild my company's web app, XRay.Tools. In this video, I'm going to show you the results, and cover the pros and cons of AI-powered development.
Discover related resources on XRay Workflow:
https://go.xray.tech/lovable-rebuild-resources
---
00:00 - Intro
00:19 - XRay.Tools: the original app
01:19 - Lovable's version of our app
03:55 - My thoughts
---
To learn more about how you can use automation and other techniques to improve your company's workflows, visit our website:
https://xray.tech
Read our blog posts for tutorials and deep dives into automation: https://xray.tech/blog
Need help with a Zapier automation? XRay is a certified Zapier expert: https://zapier.com/experts/xray-tech/
Working on a Pipedream project? XRay is a verified Pipedream partner. https://pipedream.com/partners
Reach out to XRay for an Airtable project: https://ecosystem.airtable.com/consultants/xray-tech
---
Music credit:
Different Heartbeats by Heston Mimms, Imuno
imuno.sourceaudio.com
Discover related resources on XRay Workflow:
https://go.xray.tech/lovable-rebuild-resources
---
00:00 - Intro
00:19 - XRay.Tools: the original app
01:19 - Lovable's version of our app
03:55 - My thoughts
---
To learn more about how you can use automation and other techniques to improve your company's workflows, visit our website:
https://xray.tech
Read our blog posts for tutorials and deep dives into automation: https://xray.tech/blog
Need help with a Zapier automation? XRay is a certified Zapier expert: https://zapier.com/experts/xray-tech/
Working on a Pipedream project? XRay is a verified Pipedream partner. https://pipedream.com/partners
Reach out to XRay for an Airtable project: https://ecosystem.airtable.com/consultants/xray-tech
---
Music credit:
Different Heartbeats by Heston Mimms, Imuno
imuno.sourceaudio.com
Summary
I rebuilt my app with AI. Here's what I learned.
In this insightful tech exploration, the creator tests Lovable.dev, an AI app builder, by challenging it to recreate XRay.Tools, an automation search engine that originally required hundreds of engineering hours to develop. The video provides a practical, hands-on assessment of AI-powered development tools and their current capabilities.
The demonstration begins with an overview of the original XRay.Tools, which helps users find automation capabilities across different platforms by searching for specific tools like Slack or Zoom. The creator then shows how they used Lovable to build a functional prototype of this complex application in just a few hours without writing code, successfully connecting it to a Supabase database.
While the AI-built version successfully replicates core functionality—allowing users to search for tools and view their automation capabilities across 13 different providers—the creator highlights important limitations. The AI-generated app struggles with data accuracy and technical complexity, showing incorrect automation abilities for some tools.
The video offers valuable insights for developers and business owners considering AI app builders. The creator concludes that tools like Lovable excel at quickly creating visual prototypes (getting from "zero to 6 or 7 out of 10") but aren't yet ready for production-level applications. They're particularly strong with aesthetic elements like animations and layouts but fall short on technical implementation.
For those without coding experience, the creator recommends sticking with traditional no-code builders rather than relying solely on AI-generated code, as the "last mile" refinements needed for public deployment remain challenging without the ability to modify code.
This practical evaluation helps viewers understand where AI development tools currently fit in the workflow—excellent for rapid prototyping and visualization, but not yet ready to replace skilled developers for production applications. The video includes links to additional resources about automation workflows and AI development on the XRay.tech website.
Transcript
0:00
can you really build an app with AI
0:02
prompts alone that's what a lot of AI
0:04
tools promise these days so I decided to
0:06
try one out I fed it promps to rebuild
0:08
my company's web app that our Engineers
0:10
spent hundreds of hours putting together
0:13
in this video I'm going to show you the
0:14
result and show you how AI app Builders
0:17
might fit into your workflow let's get
0:19
started this is x-ray tools it's an
0:21
automation search engine where you can
0:23
search whatever tool you're using now so
0:25
we could say
0:27
slack and maybe Zoom
0:31
those two tools are X-ray certified and
0:34
they're on a whole bunch of automation
0:36
providers so the real value of this tool
0:39
is when I click on view automation
0:40
abilities you can see all of the
0:42
triggers actions and searches available
0:46
on every automation provider that's
0:48
pretty cool if you're wondering where
0:50
you should build an automation for slack
0:54
in this case it gives you all the
0:56
information you need and you can go
0:58
directly to each automation providers
1:00
page that talks about slack so you can
1:03
really see how it all works but this
1:06
took an engineer hundreds of hours to
1:08
build and I wanted to see if with these
1:11
modern AI Builders now we could actually
1:14
rebuild it in just a few days so I
1:17
opened up lovable looking at the home
1:19
screen here you see lovable. deev says
1:22
idea to app in seconds it's your
1:25
superhuman full stack engineer now I was
1:27
very skeptical when I landed on this
1:30
landing page cuz it says it can do a lot
1:34
but in just a few hours I was able to
1:36
prototype a real automation search
1:39
engine that actually pulled from a
1:41
database now I'm not even writing any
1:44
code you can see here on the left that I
1:46
went back and forth with this AI in a
1:50
bunch of different messages to try to
1:52
get this connected to my superbase which
1:55
is a no code database application that's
1:58
similar to Zan or MySQL superbase is
2:02
really a super scalable database and
2:05
lovable. deev connects directly to
2:08
superbase so let's actually give it a
2:11
try if I search slack here boom I am
2:14
shown slack and slackbot you can see all
2:17
of the different triggers actions and
2:19
searches and you'll notice that I added
2:21
13 different automation providers
2:24
there's new automation providers all the
2:26
time so I wanted to not only incorporate
2:29
the ones that we already had like zapier
2:31
and make but also add some new ones like
2:34
pipe dream and Trey and unito or even
2:38
Parabola Lindy Bine relay and flow wise
2:41
these are all different like AI based
2:44
automation workflow Builders so there's
2:46
a lot of nuance here and we're trying to
2:48
make x-ray tools better so I can go
2:51
ahead and click inside of slack and I'm
2:53
shown all of the different Integrations
2:56
on these tools now it's not perfect so a
2:59
product is only as good as the data you
3:01
give it in this case in if there's a
3:05
bunch of abilities that don't really
3:08
look very good and they kind of don't
3:12
make sense like notify new Twitter
3:15
followers is not a slack ability email a
3:18
message to slack is an ability Daily
3:21
Post about Fitbit doesn't really make
3:23
sense so it's not perfect but it's a
3:25
really good prototyping tool and I can
3:28
even click on view integration and it
3:30
takes me directly to in this case
3:32
parabis page about slack I was even able
3:35
to have it say here here's some confetti
3:39
that's something that I personally just
3:40
think is so fun am I going to replace a
3:43
full stack engineer with this chat-based
3:45
developer probably not but is it going
3:48
to be the fastest way to go from a piece
3:51
of text to a functioning prototype it
3:54
just might be lovable is extremely good
3:57
at going from zero to like 6 or 7 it's
4:01
not going to be perfect but it's going
4:02
to be a really good prototype and if
4:05
you're trying to do things that are
4:06
technical it's going to fall short but
4:09
if you're trying to do things that are
4:10
aesthetic that you're just trying to
4:12
show visually and you want to talk about
4:15
the way that something should look it's
4:17
going to be able to build something with
4:19
code based off of a description that you
4:21
give it so things that are aesthetic
4:23
like confetti or the way a border should
4:26
look or the way that some different
4:28
elements on the screen should be
4:30
visually arranged lovable is going to be
4:32
really good for that and if you're a Dev
4:35
you can just copy and paste that that's
4:36
going to be way easier than trying to
4:39
translate a figma diagram by hand but at
4:42
the end of the day I would still
4:44
consider this to be a prototyping tool
4:46
this is not going to be production ready
4:49
for anything that's that's really
4:50
meaningful you're still better off using
4:52
softer or using a low code or no code
4:55
Builder to build an interface for
4:58
something as opposed to being completely
5:00
reliant on the AI prompt to actually
5:04
generate usable code if you don't know
5:07
how to code and you don't know how to
5:08
modify code I would stay away from these
5:11
tools like lovable or bolt. new they
5:14
would be really great to prototype but
5:16
you'll quickly find that your hands are
5:18
tied when it comes to getting it that
5:20
last mile to try to make it actually
5:22
usable something that you can actually
5:24
publish publicly that could be useful
5:28
for your audience it's just just not
5:30
going to get there yet we'll check back
5:31
in 6 months and see but right now stick
5:34
with the no code Builder and the
5:36
limitations that come with that because
5:37
you will be able to get to a good point
5:39
where you can finally click publish at
5:41
the end of the day if this video helped
5:44
you out give us a like And subscribe to
5:46
the channel for more Ai and automation
5:48
tips tutorials and news every week you
5:50
can also follow me and my company x-ray
5:53
by searching x-ray automation on nearly
5:55
every social media platform for now I've
5:57
got to go until next time find your
5:59
focus and stay in flow