From Vibe Coding to Production: Closing the AI Impact Gap with Databricks Apps

Summary

This video explains why vibe coding has not yet delivered enterprise impact—it is easy to create polished demo apps but hard to infuse them with enterprise data context, meet production-grade quality standards, and govern them at scale—and how Databricks Apps addresses these three gaps.
Four announced capabilities are introduced: App Spaces for governance boundaries around app deployment, AppKit for a consistent enterprise-ready framework, Genie App Builder for creating production-ready apps from a single natural-language prompt, and Unity Catalog telemetry for centralized app observability.
Serverless Apps with scale-to-zero economics make personal custom apps cost-effective for individuals and small teams, and a live demo shows a Genie app built and deployed end to end from one prompt.

From Vibe Coding to Production: Closing the AI Impact Gap with Databricks Apps

You can't open social media today without another post proclaiming a new age enabled by vibe coding. But in the enterprise, vibe coding has yet to deliver the impact we're led to believe it should. That's because it's easy to vibe code a polished demo app, but it's hard to infuse that app with your enterprise data context, ensure it meets production-grade quality standards, and centrally manage and govern these apps at scale. this video lays out how Databricks is tackling these challenges to help you close the gap from vibe coding hype to real enterprise impact.
Talk By: Cong Xu, Product Manager For Databricks Apps, Databricks ;

Chapters

FAQs

What is the vibe coding gap in enterprise AI?

The vibe coding gap refers to the disconnect between how easy it has become to create a polished AI-powered demo app using modern coding tools and how difficult it remains to deploy that app in a way that meets enterprise standards. In the enterprise context, a working demo is not enough—apps must be infused with real enterprise data, meet production-grade quality requirements, and be centrally managed and governed at scale. This video explores how Databricks Apps is designed to close that gap.

What are App Spaces in Databricks Apps?

App Spaces are governance boundaries within Databricks Apps that define where apps can be deployed, who can access them, and what data and resources they can use. They provide a structured way for IT teams to manage both central custom apps built by engineering teams and personal custom apps created by individual users. This video explains how App Spaces allow organizations to maintain governance without restricting developer flexibility.

What is Genie App Builder and how does it work?

Genie App Builder is a capability within Databricks Apps that allows users to create a production-ready application from a single natural-language prompt, without writing code. This video includes a live demo showing an app being built and deployed from one prompt, demonstrating how Genie App Builder lowers the barrier to app creation for business users. The resulting app is integrated with enterprise data and governed through the Databricks Data and AI platform.

How does Unity Catalog telemetry support app observability?

Unity Catalog telemetry collects and centralizes usage data, performance metrics, and access logs for all Databricks Apps, giving administrators visibility into how apps are being used across the organization. This enables teams to understand app performance, identify bottlenecks, and audit data access patterns without building custom observability tooling. Combined with App Spaces governance, Unity Catalog telemetry provides the control plane that IT teams need to manage a growing portfolio of enterprise apps.

Full transcript

[00:10] All right, welcome everybody. People are still flocking in. I'm Song. Good to see you all. I'm a product manager for data bricks apps and we have a very ambitious title here today. From vibe coding into production, closing the AI impact gap
[00:25] with data bricks apps. Very ambitious title. I do think we have a very cool solution for you all that will help definitely get partway there. So excited to show that to you. Quick show of fans. Who's seen the keynote from Justin early
[00:38] today? Who saw that? Okay, so you can basically all leave the same thing more or less with a bit more detail maybe. Um, okay. So before I get started, there's a lot of space in the middle.
[00:51] Wherever I'm closest to is the most space. So that's on the edges a lot right in front of me. Um before I get started, so who in here has sort of really never heard of apps until maybe
[01:03] this week? Like it's very new to data bricks apps. Okay. Couple people. Okay. And who's sort of dabbled in building a couple apps but not like Okay. And the rest you guys have more than 10, 20, 30 apps. Is
[01:16] that sort of the right way to think about it? Okay. Cool. I will do a quick like what is database apps? If you're completely new to apps, maybe come find me after if something doesn't make sense. This is definitely um not
[01:29] the more not the extensive what is data bricks app session because I did that yesterday. So you should have been there yesterday. Um okay, cool. So let's get started. First of all,
[01:43] we're going to show a bunch of terms I'm sure you've all heard before and I'm curious, you know, what your reaction is to them, right? Cescopolis, fast fashion of SAS, the end of third party software, VIP coding. I'm curious how many of you
[01:57] have there's a lot of space here. How many of you have this thing happening at your organizations where there's a lot of vioded apps popping up everywhere? How many people have that experience? Okay,
[02:11] good amount. And for how many of you would you say is it on margin a benefit versus a burden or a distraction? Okay. Nobody. Okay. Makes sense. I mean
[02:24] self- selected audience I guess. But yeah the idea is we've all heard these terms right? We are not sure if we felt the benefit. And it sounds like this room here would agree and we at data bricks definitely agree. So before we
[02:37] dive into the solution, what do these terms actually have as the underlying promise? What is the sort of pot of gold at the end of the rainbow? And we really think it's, you know, fundamentally a
[02:49] third era of software. Without trying to be too Steve Jobs, we're going to just go through this quite quickly. You had the first era which was onrem. We said it's 90s to mid 2000s, but we've all
[03:02] been familiar with on-prem software. If you have not, I guess congrats. You are young. But most of us have dealt with on-prem software. It was good for productivity. It sort of made the PC real in the workplace, but you know, it
[03:15] was quite clunky. Then came cloud and SAS. If you haven't experienced cloud and SAS, you're probably too young to be in this room, but all of us have experienced SAS. So SAS, you know, it continuously evolves. It's a big leap
[03:28] forward. But how many people in here are just very happy with their SAS tools of choice? have no complaints. Yes, it's a very negative a very negative questions, you know, but it is
[03:41] true. I think we all fall in that camp. We all know the limitations of SAS. And so the promise that we are being told is really we're in a third era, right? this era of custom apps where you don't have
[03:53] to deal with the monolithic interface of apps but rather you can build something that is custom made to your use case and if you don't need it tomorrow you can dispose of it or you can persist it over
[04:06] time. So that's really the promise that all these terms on the prior page are trying to sell us and we at data bricks have seen two patterns of this and we think for both we've only scratched the surface. The
[04:19] first pattern, well, there's a lot of seats here if you want to sit, but if you want to stand, there's two patterns. The first pattern is the custom the centrally built custom apps, right? So,
[04:31] these are apps that maybe all a lot of you have built centrally. You know, you have say an app for the sales team to do something they couldn't do historically. It's something that you build centrally and then you sort of propagate through the organization. And we think that
[04:45] that's already happening. A lot of our customers are doing that today. and we put a arbitrary number but you know a lot of this is still unrealized potential but we're seeing some traction we're seeing some real value being generated there but of course this is
[04:57] limited to what central platform teams can build or central software teams can build the much bigger opportunity and also the much bigger pain is the personal custom apps
[05:09] are apps that the people closest to the problem or closest to the decisions can build themselves they don't need to large apps. They can be apps they built just for their own use or for their
[05:21] small team's use, but it's tailor made for their problem and they're the best ones to solve it. Right? So, the personal custom apps is where the massive productivity gains are going to come from in our opinion, but it's also
[05:33] where there's a lot of unrealized potential. Today, we barely see anyone successfully doing this. And I I don't know if some of you have figured this out. If so, you should please come to me. I want to learn. But we're trying to
[05:45] make this happen. And so what is preventing these personal custom apps from becoming a reality, right? We don't think it's creation. We all know creation has been solved and this chart was shown earlier. But we all know
[05:59] vibe coding coding agents have solved largely solved the creation of software and you can see this chart again. It's actually quite cool. I think the 50 billion is bit of a crazy projection but the range of projections is 14 to 50
[06:13] billion for GitHub commits this year. So it's a 14 to 50x increase. Insane either way. So it's not creation, right? So then what is it? Well, what I'm sure you've all observed is that it's very easy to
[06:27] vibe a demo app. But then it's very hard to make it real, right? At scale in the enterprise that is heavily governed and then driving real business impact. Is that a statement most of you would agree
[06:40] with or has someone figured it out? because if you figured out, you should come up here. Okay, awesome. So, we're still all on the same page. So, then why is that, right? What is the journey from demo app to business impact? We think there are
[06:54] four critical things that are missing to take you from a prototype you VIP coded that looks really good to something that can drive real impact in your business. So, what are those four things? You might have picked up on this theme
[07:07] already throughout the two keynotes. We love a lot of words that start with C. this year. I don't know why. So, we have four words that start with C. So, the first one is context. I think we've talked about this at many keynotes. A
[07:19] lot of the tools today don't know your business. They don't know your data. So, you might be bad cooking something and then it's time to move it into your data and they don't really know how to do that. So, we need coding agents and we need apps that know your data. Number
[07:32] two, control. That's probably the biggest one and I'm sure the one that creates the most headache. How do you govern all these apps, right? How do you make sure that it do what you want it to do? The third one is consistency. We hear
[07:45] this over and over again where people are vibe coding apps, but they all look different. They all use a different stack. They all rebuild the same components that already exist somewhere centrally. So you want consistency
[07:58] across all these vioded apps. And then the last one is consolidation, which just means when you have thousands of these running, you want to be able to govern to administer those easily at scale.
[08:11] So context, control, consistency, and consolidation. You'll hear this a couple more times throughout the talk. And by the way, I should have mentioned I'm going to talk at you for a bit, but then there will be time for Q&A. I know how painful it is to just listen to me talk
[08:24] out of from personal experience. So to summarize, number one, creation we think is somewhat largely solved, right? A lot of people can create, but then context is missing from these tools and
[08:36] these apps today. You do not have control over them. And that, by the way, that includes both governing what they do, but also cost. That's a big factor. You can see how I'm already leading you to the things we announced at the
[08:48] keynote if you paid attention. Consistency. Everything should look the same, use the same components, and then consolidation. you can easily administer all these apps at scale. So like I said, we think there's a bit
[09:00] of a gap here because first one is largely soft, the last four are still somewhat unresolved. And so that is partly why all of you are in this room and partly why we have all these issues.
[09:12] So I already said this, but just to make it very clear, we think there's a gap in the market here that we are trying to fill. There's it's good that I'm repeating a lot. There's still some seats up here if you
[09:24] want to sit but you can also stand. So what is this gap right? We believe today and this is on the left hand side there is a tool that exists
[09:37] which is a coding agent that allows you to democratize creation but what is missing is the platform to support the tool i.e. In other words, you have personal software, but you need it as an
[09:49] enterprise service or PS AE. Not quite as, you know, easy to say as SAS, but if you want an acronym. So, personal software as an enterprise service. Again, the tool we all know,
[10:02] but the platform is what takes care of the other missing gaps that we have just seen to get from demo app to business impact. And you might not be surprised to hear that we at database apps are trying to
[10:16] become that personal software as an enterprise service platform for you all. Okay. So at this point let me do a quick refresher on what database apps is.
[10:28] Couple pages just so you're aware. Um database apps is not just an app otherwise that icon will be way too small. There's more more to it. It's an app that is embedded inside the broader
[10:40] data bricks platform, right? It's a datab bricks app. This is a complicated architecture and and between us I think it might be different from what was shown a couple hours ago. So maybe don't take too many pictures. The versioning of this has to
[10:52] be has to be fixed. It essentially says the same thing. Essentially says the same thing. But if you want to simplify it, database app makes it very easy to integrate with everything else in the platform. So out of the box you
[11:04] basically get these things at the bottom you get the app back end of the future. What I mean by that is you heard talks about lakeflow lake base lakehouse a lot of lakes it's like Minnesota but
[11:16] basically what we're saying is the apps today most of them just have a database. We give you a database but we also give you an analytical stack. We give you a real-time stack. We give you a way to ingest all the other data you need. And we think with agents longunning tasks
[11:29] more complicated tasks that's the future of the application that's the future back end so you have that at the bottom then you have existing governance and security all unity that's the middle
[11:41] layer the app will inherit that governance and security so one thing I want to point out is one of the most popular features about apps is what we call on behalf of user off it basically means if two people have different
[11:53] permissions inside of data bricks the app will respect that and they will see different data when they open the app. So it inherits the existing security and governance. Same thing with network controls. If you set that up at the workspace level, the app will respect
[12:06] that. And the third thing is connect to your data and AI. That is the fleet of genies that we have for you and also agent bricks. So it's your data, it's your AI. We make it easy to connect to
[12:18] those tables, volumes, whatever it is. And so it's an app that sits on top of all of that platform. Does that make sense to the folks that are hearing about it for the first time? Yeah.
[12:30] Cool. And then apps has been very popular. I think Ali said it on Monday. It's growing very quickly. We have thousands of customers building applications on data bricks. So this is very much something that a lot of you in the room have tried. Maybe some of your
[12:42] logos are up here. If so, I hope it's all approved. I think so. Okay. So, let's go back to what we were saying. Personal software as an enterprise service database apps. That's where we want to be. How do we get
[12:55] there? Right? What are the things that we've announced? I'd actually be curious if someone remembers the things we've announced, but I'm going to re reiterate it for you all. So, we've announced basically a set of things that if you
[13:08] summarize them allows you to do govern vibe coding with data bricks apps. That's fundamentally what we're trying to do. You can vibe code. You can unleash the creativity from the people that are closest to your problems, but
[13:20] you can keep control and governance centrally. That's really the tension, the fundamental tension that we're trying to solve with this set of offerings. And so what does that mean? What is this governed vibe coding platform that we
[13:33] are trying to establish at data bricks? So at the bottom we have probably the announcement that got the most cheers. Makes sense. Serverless micro apps, right? What those are is basically
[13:48] apps that will scale down to zero when they're not being used and then scale back up to one within a reasonable latency. It's not going to be the most latency. You cannot have the most latency sensitive use cases on there,
[14:00] but within sort of 10 seconds, we're going to try to get that thing back up. And for a lot of our customers, I'm sure including all of you, you already have use cases in your head, right? that are apps that are sort of infrequently used maybe once every couple of weeks, couple
[14:12] times a week and the end user doesn't care if it takes five or 10 seconds to spin up the app. It's not mission critical. But historically, you had to use the long running instance which was too expensive. So we want to make sure
[14:24] those use cases are all economically viable on apps. So we introduce this at the base foundation of the govern vibe coding platform because as you can imagine if you start vibe coding 99% of
[14:36] the apps will never be used again. They're all just random prototypes right so we don't want those to incur any cost. Oh yeah that's why the little control thing here is on the right hand side. Can you see that? I spent a lot of time
[14:49] trying to design that little button. Anyway, control cost control primarily cost control. Oh, number two, app spaces. That is maybe the one that got the least cheers, but
[15:02] probably the most impactful. This is really going to be your governance boundary. So, where this came from is we already have workspaces. Why do we need another space, right? Workspaces, as some of you might know,
[15:14] are usually very difficult to configure and set up. There's a lot of configuration. They're quite heavy weight. They're also very powerful. But with apps, there's two unique characteristics about apps that made us decide we need a new governance
[15:26] boundary. The first one is they're the most untrusted artifacts inside of data bricks, right? Yes, you can do some stuff with notebooks, but apps you can do anything. And so as a result, you
[15:39] need much more fine grain controls over how to govern these apps. The second one is especially with vibe coding the number of apps we expect organizations to create is going to go through the roof. Think back to the personal custom
[15:52] apps. I mean if everybody can create ephemeral apps for their project that's that changes every week right you will have a skyrocketing number of apps and so we think a new governance boundary to
[16:04] make sure that all of those are governed centrally is very important. So that's the app space and so you have control and you have consolidation at the app space level. Third is the app kit and the app kit is
[16:17] actually already available. I'll spoil it right now. The rest of the stuff is not going to be in prevent until July. So I apologize for giving you the sneak peek, but AppKit is already available.
[16:29] You can use it outside of the stack or inside the stack. It's basically the way to ensure consistency. And so on the one hand, it's consistent in how it interacts with the rest of data bricks. So you don't have to worry about how to connect to a SQL warehouse, how to
[16:42] authenticate into a table, all those things we take care of. It's also consistent in that and admittedly we will still we're still trying to figure out the best form factor for this. It allows you to customize your templates, customize your
[16:55] components, customize the stack you want to use, and then standardize around that. Right? We know you all have your own requirements for what app should look like, feel like, what stack they should use and so we want to make sure
[17:07] you can configure that. And AppKit is very much done to be extensible. So number three, AppKit consistency. And then lastly is the Genie app builder, which was what the demo was
[17:19] about earlier today. And the Genie app is about creation about context. The Genie app builder uses ontology to get all the context about your business. And then obviously it is the interface through which anybody no matter if
[17:32] they're technical or nontechnical can create applications. So that is the way we want to take care of context control consolidation
[17:44] and cost and consistency. Cost is part of control. Cool. I'll leave that up for one more second. So we can dive into each of these a bit
[17:56] and then I'll what I'll do is I'll dive into each of these a bit and then I'll show a quick demo of what we have. Again the thing is going to launch end of July. So you are literally getting a currently inactive development view at
[18:10] what's going on but I wanted to show you what we have cooking and then we'll do questions and answers. Okay. So number one, Genie app builder, I already mentioned it uses the ontology.
[18:22] So it knows your data. It's fully embedded inside of data bricks. So one problem that some customers had when we offered partnerships is that they were worried about data leaving the premises
[18:34] of data bricks, right? And so if you don't let the data leave, the agent can't really do much, but if you do let the data leave, you can sleep at night. So it's a weird trade-off. And so we decided, okay, bring it in platform.
[18:46] data never leaves the data bricks premises production quality by default that is what the app kit does. The app kit basically takes care of things like you know timeouts or like refreshes or like streaming results in so a bunch of
[18:59] things you don't want to think about that are somewhat critical when you're building an application and so app just takes care of that plus integrating with services plus letting you customize. So that's a gen app builder. You saw a
[19:11] demo of that earlier. I'll show you another short demo of this after this talk. Number two, AppKit. It's a framework for human and agent app builders.
[19:24] I I still have to get used to calling out humans that are separate from agents, but we'll get there. Um, again, platform native, you don't have to worry about having skills or making sure your agent knows
[19:38] how to connect to these services, when to use authentication model. It's all taken care of for you. Production quality by default. I already said that. And then extensible, very important. We want to make sure you can build your own
[19:50] plugins. Plugins are basically the components that you can define inside of AppKit. You can build your own plugins. You can create your own custom templates. That's really a critical piece. So that's AppKit.
[20:02] Number three, app spaces. App spaces for you. Most of you have built apps before. So one simple way to think about app spaces is today you have to configure things at the app level
[20:14] right every single app you configure resources you configure potentially on behalf of scopes tomorrow you can do that for a group of apps you configure it one time and then a group of apps has that configuration for their apps and
[20:27] we've seen a lot of use cases already for um for app spaces so they range from team app spaces so you have the same workspace with all the data say in depth You have one dev workspace and then you
[20:40] have a finance team and they have access to different data than say a sales team and a marketing team. Also very popular is personal spaces. So basically imagine a space where every
[20:52] app cannot be shared with anybody else. Right? So you can only build apps for yourself. It lessens the governance burden because these are not going to be shared widely and you can iterate, you can ideulate, you can innovate, but
[21:04] that's not a very popular use case to sort of govern what apps can do. So app spaces govern sharing again admin level controls and then scalable operations. So one thing I should mention is I don't know if how many of
[21:17] you know apps already integrates with hotel doesn't integrate it emits hotel telemetry in unity catalog. So basically it will come out of the box with telemetry enabled across all the apps in
[21:29] the app space. So you can basically have that in your UC tables and that is one of the ways we want to make sure it's easy for you to monitor a fleet of apps, right? Hundreds and thousands of apps. So this is maybe the
[21:43] least flashy but probably the most impactful. I mean this is definitely the problem that a lot of us struggle with the most. And I want to urge you all to think about appspaces not in its current form because admittedly currently it has
[21:56] a bunch of things and I'll show you what it has. But really the main push here that we sort of went through long cycles with internally is to convince the company and everybody that this is a new kind of
[22:08] risk that needs to be managed with this dedicated governance boundary. Right? So a couple things for example that are obviously going to be managed at the app space level. What packages can your app use? That's something that you didn't
[22:21] have to worry about before necessarily with pipelines or with dashboards even. But with apps, you do have to worry about that. What packages can my app use? So package repository will be something you can configure at the app space. Even if that was never something you needed at the workspace level,
[22:34] right? We'll also try to figure out, you know, if we can scan your code for vulnerabilities, etc. All these things that are unique to apps, we now have a new governance boundary that does that for groups of apps. So it's both the
[22:46] scale and the fact that we have a dedicated place. So this really is a big big win for us internally even though maybe it's not as not as what's the word flashy. All right. Lastly, serless micro apps.
[23:01] This is the one that's the most obvious in terms of benefit and also the one that probably is the most technically challenging for us. But basically this is two things. One is you have a bunch of apps that no one really uses.
[23:13] These will shut off. two and what we think the opportunity really is based on our customers. you have a long tale of use cases that you would love to bring into data bricks apps but it's just not
[23:25] economically viable right because the apps are unfrequently used especially those personal apps I was talking about now hopefully it's econ economically viable even though knowing all of you I'm sure it still will not be for some
[23:37] of you and we will figure that out but this is a step in that direction to make sure that you can bring that long tale of use cases onto the platform shared in full isolation Justin mentioned this as
[23:50] Okay, before I go into a short demo, are there any other questions? Or maybe we'll do questions afterwards. Yeah, I'm going to kick kick off some stuff. This is I actually can't believe I'm demoing this. So, we are
[24:03] maybe four or five weeks away from having this in preview. So, we are still very much in development. Um, but we have a version of these that's much more polished, what Justin showed. And now I'm giving you the in development
[24:16] version. But I want I want you all know sort of what we're cooking up. Okay, it's pretty cool. So, let's see. Let's go here first. Ignore this. We're getting random errors right now. But the first place I want to
[24:29] start is the app spaces. So, you can see here we have a little app spaces, but let's go into one of them and we can see what this looks like. So this is just a visual way for you to
[24:43] understand what you're going to be able to configure at the app space level. A reminder what is an app space? It's basically configuring a bunch of stuff for all the apps inside that app space. So you have you can have hundreds of
[24:55] apps. They will all abide by the same rules configurations because the app space is configured that way. So right now you can already see you can configure shared resources, shared data, you can configure on behalf of user
[25:07] authorization like what can the app act as you for user management and then budget and compute uh policies.
[25:19] So this is just the beginning but as I said this should give you an idea of these things will apply to every app in the space. So you don't have to worry about you know any apps not abiding by these rules. The other thing you'll see is there's multiple apps in here and
[25:31] none of them are on. I'm actually curious if our W pool is working right now. But the reason why none of them are active is because they're all scale to zero. Now, if the scale to one is working, we will see in a moment. I will show you. But scale to zero. So now I
[25:44] pre-loaded some things. But if you go here, let's start to build. This will take a bit of time, but let's see. So I have an app space here. You can select different app spaces.
[25:56] And I'm just gonna say, you know, build me a Wonder Bricks app. Wonder Bricks is our sort of sample data set for um like Airbnb like data like a travel business.
[26:09] So you can you can give it a name here. We can figure this out later, but I'm just going to say DS demo. And so let's see what it will do. And I've pre-loaded some of these. This will take a bit of time, but what it will do
[26:21] is we'll basically open the Genie app builder, which looks something like this. I'll go back to this tab in a second. This is always good to see a face like this on a live demo. That's what you want to see. This is this is how you know it's live. But so I'm going to
[26:34] reload this. And so you can see here what basically happened is you give it a prompt. I gave it a more extensive prompt in these two cases. But I want to show you something. So I gave it a prompt like this.
[26:48] And I did this in two separate app spaces. This is I'm building an app in an app space that has no resources. You see there's absolutely nothing attached to this app space and then the exact same prompt here I gave it in an app space
[27:02] that has a bunch of resources right so you can see in the first app space because the app has no access to any resources even though I as the end user in data bricks can do a lot of things it
[27:14] tells me at the end hey I have everything I need to understand the data but there's no resources attached to the app space so both must be attached before the app can work And so it will stop you in your tracks and basically
[27:26] tell you look even though you as a database user can do these things your admin has constrained what you can do in this space. So unless you have access to other app spaces you're not going to able to be able to build this app.
[27:38] the same prompt, exact same prompt just in a different app space with the right resources builds you this right. It's a much longer output. We're gonna constrain this to be much more digestion friendly, but ultimately it builds this
[27:52] one shots a simple app and I think the genie works. Let me just double check here. What is Oh, it's not loading. Great. Well, it's live. But um it builds you a
[28:04] simple app. And I'm actually going to wait for this to reload. So that is the power of the app space. No matter what your user has inside the app access to, it will build. This might actually not be a problem with us. This
[28:16] might be a problem with something else. Let me see. Then what I've done is I've built another app over here in a different space. Again, this face is just so beautiful.
[28:28] I've built another app here. Similar prompt. It looks slightly different, I think. And what did I do here? Oh yes, let me sign in. I haven't signed in.
[28:45] What I want to do here and then we'll get into questions is I want to show you something quite cool. Well, first of all, let me start a new session. And what you can do here is a very visual and like low to no code interface. You can say you can click on
[28:58] something like this and you will see it selects this and you can just say you know delete this element. So it will do this and it will delete the element. Let's get back to this in a
[29:11] moment. The one thing I want to try right now and I I will tell you it may or may not work is right here. You can see we have an app deployed as well
[29:23] that was built in the app builder. We can actually go into the session and the compute has stopped. And what should happen and maybe right now I won't is the reason it stops because it auto scales down to zero. And so what we are
[29:35] going to do is we're going to allow you to basically scale this app up within a few seconds. You have to reload it once. As you can see, it's scaled back up to one. And now the compute, if you go back here, is going to be not stopped
[29:48] anymore. It's going to be running. So this is scaled to zero with a weird screen in between that you will not see working for you. Right? So that app was not incur any cost and then some user wanted to use it again. It is
[30:01] spun back up. This is also wonder app. So now what we can do here is we can open the genie app builder. There might be another session already ongoing. And I want to do one last thing here.
[30:14] Let's go back to the other app. I'm switching screens a bit but hopefully you can see here the fourth thing is gone. Right? It just deleted the fourth element. And so now we have three. So
[30:26] you can use various sort of visual ways to edit the app. But the last thing I'm going to do I can do it here or the other place. And this will take a bit of time. I'll just do it here. Is I can say something like I do not like the look of this at all.
[30:40] Use this repo to change the look. And I can attach a zip file called neo brutalism dashboard template.
[30:52] And what this is, some of you might know this, it's just an open source git repo. This will take a bit of time, but hopefully by the end of the session I would have done it. Neo brutalism dashboard. It's just a git repo open source that formats apps that look more
[31:04] like like this. And the point here is the agent can already adjust what it does based on some instructions, right? You can imagine instead of doing this, you give
[31:16] it your templates. You give it your look and feel. You give it a set of constraints that it should consistently build around. And the agent can already do this. We just need to find the right form factor to make you configure it. So
[31:29] while this runs, the last thing I want to show you is app kit. I've showed you scale to zero app space genie app builder is what you're seeing right here. Appit you can see in this code here and this
[31:41] will be the same across the code repo is going to look the exact same. One thing I want to point out is the server is like seven lines of code. very simple because a lot of these plugins we've abstract away the complexity and then on
[31:53] the config side you can see all your queries all the stuff that requires user input is neatly laid out as different files. So each query that the app uses to generate some of these charts is right here. So these are just some
[32:05] examples of appkit doing its thing right this is not some messy repo and on the other hand you have this as well here this looks very similar appkit you have the server so all of these use app under the hood to make sure there's
[32:17] consistency and quality okay so let's do one last thing before I open up to questions let's see
[32:33] where this is at okay it's still changing look and deal. I'm assuming nothing has happened yet. Okay, let's leave it at this and I'll come back to this in a second while we answer questions. Seven and a half minutes. Let me just go through one last slide and summarize what we talked
[32:46] about and then we'll do questions. And I'm here. How long do we have the room? Probably not that long, right? I'll be outside afterwards if you want to ask me questions. In summary, last slide, a wall of text. Amazing. So, number one,
[32:59] there's a real gap between vibe coding and production. We think it can be closed with context, knowing your data, control, giving you the ability to
[33:11] govern, consistency, making sure all apps look and feel the same and use the same stack, and consolidation, making sure you can manage thousands of apps pretty easily. Creation is solved as a tool of coding
[33:24] agents has enabled us to do that. The platform to support this tool and the new behaviors is lacking. And we at databicks apps with Genie app builder with the app kit with scale to
[33:36] zero or serverless micro apps and with app spaces we want to be that platform for you. Okay. So like I said gated preview July worst part of this presentation and reach out to your
[33:49] account teams if you want to sign up in the interim. You can go to DevHub our developer platform to uh check out some ready to go prompts to build apps without this stack. But we'd love to hear from you all.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.