Dashboard Slop rant with Sonny Rivera

Speaker A: Sam. Mhm. Special edition live from Austin. Episode of, uh, catalog and cocktails.

Speaker B: Rant session.

Speaker A: Rant session. That's true. Yeah. These Fridays are Friday.

Speaker B: Rant session. We used to do just the two of us.

Speaker A: Yeah. Uh, we need to do them more often. But we have a special guest today.

Speaker B: The second time we have a guest in our friend on our rent sessions.

Speaker C: Awesome.

Speaker B: Honey. How are you?

Speaker C: I'm doing well. So really happy to be here in Austin. So super excited. And the weather's great.

Speaker A: It's

Speaker C: a great spot

Speaker B: in our. My local brewery here in Austin. Lazarus.

Speaker C: Yeah. I was like, what the hell?

Speaker A: There

Speaker B: you go.

Speaker A: You got the UT game tomorrow, right?

Speaker C: Yeah, I got the UT game tomorrow and uh,

Speaker A: then F1 for those who are doing that. So it's a busy

Speaker B: weekend too.

Speaker A: Yeah, I think so. I think so.

Speaker C: It's a lot going on this weekend.

Speaker B: I can't imagine. Cheers. All right. What are we drinking or drinking? Beer, but

Speaker C: some pills. So I'm ahead of you guys?

Speaker A: Well, this is my second, so I'm m kidding.

Speaker C: All right. I would tell you that I'm not great at ranting. Right. I've been told like, you're a really nice guy, so I might m not be great at ranting. So let's see how this goes.

Speaker A: Well, you had, you had most of a beer, so. Yeah. Um, what's on your mind? What are you excited about right now

Speaker B: or not excited or depressed?

Speaker A: What are you upset about?

Speaker C: Upset about. Good.

Speaker B: By the way, quick hello. We got a lot of people. Ryan.

Speaker A: Hey, what's up, Brian?

Speaker C: Ryan. All right.

Speaker A: Welcome to the Friday rant session.

Speaker C: Awesome.

Speaker B: So

Speaker C: would be great to have Ryan here. Ryan.

Speaker A: I know. Ryan, what are you doing? Come to Austin, man. Hang out.

Speaker C: There's a flight

Speaker A: come on a Friday. We got beer for you.

Speaker B: Hopefully your. The. The. Your Detroit date and the D. So it was, uh, I have to say, pretty envious. Uh, that lead looked like it was going really well. So congrats.

Speaker C: Yeah. Uh, what is on my mind? You know what's on my mind? We. A lot of people have been kind of ranting about AI Slop. From a blog and a writing perspective, I think what's on my mind is Dashboard Slot.

Speaker A: Okay,

Speaker B: now, now, Claude. Uh,

Speaker A: now, now. Dashboard Slop is a first class citizen.

Speaker C: It is.

Speaker A: It's never been easier to make Dashboard Slot.

Speaker B: So Dashbo not dead. They're just. Even if they were dead, they just literally came back alive with the most. They're

Speaker A: so democratized. A, A random, uh, kid can do it.

Speaker C: Right? Exactly. You could just talk into your phone and create a dashboard and who knows where the data or the insights come from.

Speaker A: Yeah. So good thing, bad thing.

Speaker C: Uh, I, you know I, I like the idea of being able to make dashboards easier or to, to be able to use insights easier and then dive deeper and dive deeper. Um, but I kind of like the uh, my colleague Jane Smith proposed this model to me which was the driver's license model. You can't do it until you've you know, done driver training or dashboard training and you have some proficiency in a little bit of AI. What's happening?

Speaker A: Yeah, that uh, definitely uh, in addition to you mentioned, mentioned AI there like this whole data fluency conversation, it worries me a little bit because now it's really easy to just like make a bar chart and, and you know, or, or ask it to do some analysis and it's going to do something that looks super smart. But do you know, you know is random sales manager asking Claude random questions about their pipeline going to result in good things? I hope, I hope so. Um, but yeah,

Speaker B: I, we need to let this happen.

Speaker A: It's

Speaker B: like a good problem. I don't know. The genie's out of the bott. It is going to happen. So just let it happen. And then let's see, uh, people are going to crash into the wall. Right? It's gonna, there's gonna be some stuff. Yeah. We just need to. And we're being pes. I'm being pessimistic about but like maybe good things do. We'll see what the good things will happen out of it. But I mean this is now truly the self service analytics is now true. And uh, uh, we can be all top down about it like well you got to do it this way. But it wasn't

Speaker A: turning off.

Speaker B: It wouldn't be self service. So.

Speaker C: Well I, I do think you know, look, I don't have anything against anyone using uh, AI to generate dashboards or visualizations or insights. But if you walk into my meeting and you make a claim with it, with you know, an uh, insight, uh, whatever it is, a trend, then be prepared to answer the questions about where this came from 100

Speaker B: and then

Speaker C: and be able to describe

Speaker B: and

Speaker C: how did you get here

Speaker B: and, and I think that's where we're gonna let them crash. Let them, they'll be able to go do that. And I'm like holy, you convinced us all and we made this. Or you look like a freaking idiot and guess what, are you gonna do that again? Then we learn I think

Speaker C: right but what if. Yes. Uh, and, but, you know, in my world, I don't want to see people crash. Right. This is why I'm not ranting. Right. I want you to be successful. So Juan, here's. I want you to be successful. You know what I really would like you to do? I'd like you to have a good semantic model that you are building or layer that you're building your dashboard on. Right.

Speaker B: So, but, okay, but, but if this all goes well, right. So first of all, I agree. Right. You want to have that. Great. That, that, that, well, solid governed semantic model you can go trust. And if you do have that, then technically you go build out. You go build your dashboard and then it didn't work out well, so then this. So is the slop really because anybody can build a dashboard, or is the slop because anybody built a dashboard without the right foundations?

Speaker C: Yeah, I think, you know, I think it's the latter, not, you know, not the former. Definitely. If you don't have the foundation. Look, don't ask me to, I don't know, you know, pull a tooth. Okay. Get me the pair of pliers. Like. No, you really need to be, um, well versed in what we're talking about. And this is one of the reasons I think, um, the industry needs more knowledge in computer science and math, not less. Because look, when you're a builder and you're a builder and you're a builder and I'm a builder, we actually need to know how to build. We need more education here, more training, not less. And I don't think AI is relieving us.

Speaker B: Well, interesting. So Ryan, as you're saying, cloud dashboards were super impressive if you don't pay any attention to BI for the last three years.

Speaker C: That's very true.

Speaker B: So I, So the bigger question for me then is it's not that I think BI is not dead. I think it's just like the way of doing bi, like that has truly been, uh. It's not even the self service analytics, it's the, the self service of building your own dashboard. I think that is something that has been,

Speaker A: well,

Speaker B: kind of.

Speaker A: Can we possibly. Something here. I'm curious if you agree with this or not, that uh, the BI tool ecosystem, you know, uh, more and more of the emphasis now is moving, I think. You know, I always thought of BI being a couple layers, potentially more than a couple, but more, there's more the user experience part of it of like, I gotta build my dashboard. Right. And then there's the, like the semantic Model aspect of. Seems like more emphasis is going on the semantic model part now because the user experience, like it might be called, it could be these other things, right? Yeah,

Speaker C: well, and it is. We've had semantic layers and semantic models since we've had. Right. Um, and so I think it is coming more toward context layers, which are encapsulating more than just the business formulas. Right. What does this, you know, term mean and how do I calculate it? It's. It's more than that now. And so, yes, I think you're absolutely right. So there's a ton of emphasis on context layers. You know, uh, and. And I, as a matter of fact, I was talking to a customer about that today, and there's like, hey, we have. We need to figure out how we could do this faster. Right. And. And hey, we need our semantic layer faster. We need knowledge management or knowledge engineering, but we need it faster. So do you have a playbook that I can just run down and do this? Okay, well, maybe not just a single playbook to run down. It's like, well, I got to be able to get, you know, many, many domains in parallel, because if I don't, um, you know, I can't. I don't have years.

Speaker B: I'm curious on this. So, like, the customer you're talking to, just from what you're hearing, how much of that there ex is that they're expecting agents, AI to do that work? And. Or is this, uh, people today that they were doing more technical data engineering work, that we're expecting them to do this type of work? What are you. What are you seeing?

Speaker C: Yeah, well, just in general, I think one, there's just a complete lack of, uh, not lack. There. There. We are a lack of resources that understand knowledge management, what we might call context engineering or knowledge engineering, which I would say are two separate things. But I think there's just not a lot of those folks out there today that truly understand it. You have people that have done data modeling, and then they're being asked to do maybe knowledge engineering, which while you know how to model data, may not know how to model that knowledge. Right, right. So that's a different thing. Like, I'm pretty good at modeling data, Juan, but I'm not an RDF expert.

Speaker B: But I do think that there is something here that is connecting the whole, like the data product management. Right. The data modeling, talking to users and getting the requirements. Like, that's like, all of that is what I would consider knowledge engineering and all of this graph and already, like, that's Just the technology, like you should be able to go do all that work in a freaking spreadsheet. And like, but like that's the type of work that kind of happens, uh, in an ad hoc matter that I think now that's the change that we need to be investing more in.

Speaker C: Yeah. Well, it needs to be a discipline and here's what's happening. Yeah, it needs

Speaker B: to be a

Speaker C: discipline. Where it had stopped before was knowledge management or whatever else would be. Let's just go build me a spreadsheet of my business partners. And that's about where it ended. Right. And maybe I can have some metadata around my database. Right. And it ended there. It didn't take on what was reusable, what were any rules that are actually in place other than what you could glean from that definition. So.

Speaker B: But that still is the basic stuff. But it's not, I think for, for a long time in the catalog world, like we, we thought that that was like the end goal.

Speaker C: Yeah.

Speaker B: Even though we knew that was not the end goal, but a lot of customers thought that they saw that as the end goal. Um,

Speaker C: well, yeah, let me just add in here. I know you're about to say something, but I'm going to cut you off and say something anyway. So what we do, you're

Speaker B: rant right there,

Speaker C: you know the value that we bring from our data. You know, you and I talked about data modeling. Isn't data modeling, it's modeling your business.

Speaker B: Yes.

Speaker C: Right. You and I had some good conversations about that. Well, guess what? That hasn't changed. That is what we do. Right. Modeling that business, how we do it has changed and um, will continue to change. And that was the source, I think of your question was like, is this something AI is going to do? Is this something you're doing? Or are we repurposing people and their data engineer roles and putting them into that role? I think yes, probably to all of those things, but how we do those things I think is changing, but the value that we bring by doing them is not. And I think it's true of dashboards and um, you know, as well. And bi. Like a dashboard, yes, it's valuable. But now we've moved beyond dashboards. I mean, for me, I use my phone every day. I mean I like pre game plan my day because I have a surface of what um, I'm looking for from whether it's a dashboard or analytics right. On my phone. And I can use AI within my app, within ThoughtSpot and say, hey, what's going on? Here, here's this KPI. It's just another surface for me to interact with, whether it's embedded or dashboard on my phone or dashboard on the laptop. Right. And so the value that they brought, um, doesn't change, it's just how we're interacting. So another surface.

Speaker A: Yeah, uh, that makes sense. One thing that is a problem area that I've been thinking a lot about is, you know, so we're talking a lot about context layers and semantic layers and things like that and having repeatable building blocks that we're using for sort of common trust, shared meaning. Um, but a lot of that conversation comes up in more of a design time context of like we need to, we need to, we need to build this semantic layer. But there's a lot of uh, making the call in the moment that happens when an analyst is working with a BI tool or a data engineer is building a data pipeline or building an analytics engineering job or something like that. Right. And they're deciding in that moment, you know, what I need to, I need to filter by this thing and I need to make sure that it's active customer and like they're making some choices there. And what we often don't do is we don't close the loop. We don't, we don't. That engineer doesn't go, make sure I update the semantic layer when I do this. Right. They just, they start layering that logic on top. He's been thinking about that problem area a little bit, you know, has thought, spot been thinking about that. Because I'm thinking about like, how do we, how do we maintain this context layer and not do it in just the way that Ryan said here, which is like, hey, why don't we just have a lot of context? Like I think humans are going to be building the context for at least a little bit.

Speaker C: Yeah.

Speaker A: With AI assistance. Right.

Speaker C: So, you know, kind of like Ryan was saying, one of the things I've seen is, uh, engineering teams or data engineering teams said, well, I'll just build a bunch of YAML that expresses the relationships of my semantic model or my, and I'll just feed that to whatever LLM that I'm using. Um, or I'll put context in GitHub in a bunch of YAML files and you know, uh, Claude or whatever else can MCP to GitHub and pull that information down and

Speaker A: dump it in the context.

Speaker C: Is that the job? You want to be maintaining YAML files in GitHub for the semantic model that might or might not get the answer right?

Speaker A: Uh,

Speaker C: either way. So with that answer is wrong, then you get to go back and fix the YAML in GitHub and do it again. That doesn't feel like something I would want to do. Right. If I'm a, if I'm doing data engineering today, that feels like a demotion. Okay, so, um, but, but I do think, you know, you need to invest. And as I've told customers recently, you know that people want to go faster. Like. Yeah, but fast and wrong is still wrong. Right. So let's just make sure you maybe go. All right, let me step back a little bit. Ten years ago, what did we say? A single data warehouse, single source of truth, everything was there, turned out to not really be what we've done. Data is dispersed, it's federated. We have data in a lot of different places and we're still able to interact with it. AI is making that even more federated. Right. And what happened, we then laid, um, data products and domains on top of that data. And so now we kind of have this lack of a better term, a data mesh type of model where data products, um, and we have domain experts in that, that are focused on that particular area. And I think you see the same thing happening with context and with knowledge. I mean, you can't say, I want all data. You probably want to say, I need this, the set of data that I need for this domain. And I can keep pulling that thread all the way through. And soon I'll have a huge thick rope that I can build off of you just keep pulling through. Juan, you're shaking your head. No,

Speaker B: no, no. Is because I, I use this very similarity. I call it the iron thread.

Speaker C: Okay,

Speaker B: so you said it's a big rope. For me, it's like you create a thread, but it's iron. And then you keep building on that stuff little by little. Right? So you start with that very simple use case. And that use case is a question that you start understanding and you modeling out. But then you. But, but what you're modeling is, is the, the domain, but also how it reacts to the rest of the business and the action needs to be taken. And then you just keep adding to that. And then that thread, which is iron, gets stronger and stronger. And then my analogy is then that thread becomes your fabric or your mesh or whatever.

Speaker A: The context, though, that you kind of brought this up in is a little bit more of a bi context. And the question that I have though is that, uh, every time I hear people talk about semantic layers, they tend to articulate it In a pretty centralized way. And so is there such a thing as kind of like a decentralized, more data mesh oriented approach to semantic layers?

Speaker C: There absolutely has to be me and there is so one. You know, I'm not here to shill thoughtsbot, you know, you guys know I work there. But one of the things I do believe in is open and interoperable. Right. So thoughtsbot's part of Aussie Apache

Speaker B: ServiceNow too.

Speaker C: So

Speaker B: many of our colleagues are,

Speaker C: and we just published recently like within the last month, connectors to go into and out of ThoughtSpot GML. That's a ThoughtSpot modeling language, which is YAML. But now you can go in and out from both of those. There's a, um. Gartner explained it like it's a mosaic, Right. So you have this view and it might be a beautiful picture. Yes. But it's a mosaic of a lot of different components that make up that semantic model. And I, or layer what, what I think of is having a single semantic layer is a wonderful dream. It is a great thing, It's a great aspiration and doesn't mean you shouldn't try to build it, but you should recognize going to have data in context and semantics in different places and they should interoperate. And that's what I would focus on.

Speaker B: This whole single, uh, version of the truth was

Speaker A: really exactly what I

Speaker B: was thinking.

Speaker A: Single version.

Speaker B: But this, I mean that was just people trying to achieve. This was just a waste of, waste of time and energy.

Speaker C: I spent 10 years doing it. Thank you. Now.

Speaker A: Problem there, right?

Speaker C: Yeah. Yeah. Well, I think one of the things you asked was like, hey, if I'm this person and I'm working on this, am I going to go back to data engineering? Uh, and we do think about that. So I'll give you a quick example. Uh, imagine you have, you know, some, some business KPI, um, defined in a semantic view in Snowflake or in a semantic view in Databricks. Right. Well, I'm sitting here working on my qbr, my dashboards and I'm talking to my phone, say, hey, build me my dashboard. I got QBR with Cindy House in this week and we're going to talk about X, Y and this. I need my dashboard here and when I see something that I need to add to it, my QDR is tomorrow. I'm not going to the data engineering team putting in a JIRA ticket to get that semantic view updated,

Speaker A: nimble and be able to do your local thing that you need to do, Right?

Speaker C: That's right. I'm going to do the thing I need to do, and then I may push that back down. I'm not going to go to Ryan and say, ryan, I need you to, uh, go check out the YAML and GitHub and do this, this, and this and put that back in. Because Ryan

Speaker A: doesn't want to help you.

Speaker C: Yeah, he doesn't like me much either.

Speaker A: Just kidding. Ryan, we love you.

Speaker C: But if I have the tool that will let me do that and then I can go, ah, um, is this really worth having? Because sometimes these questions will be, why are you doing that? We already have this. Or we're not actually measuring that. Um, and you know, if it's valuable, then you pull it into the maybe more gold layer. And I think you're probably going to end up with these, lack of a better term. Um, uh, a medallion model. Right. For.

Speaker A: Yeah,

Speaker C: I

Speaker A: found myself, you know, and I don't know if I'm a crazy person or not. Maybe this is going to become best practice. But using the medallion concept for context a lot more lately. Like, you've got your gold semantics.

Speaker C: That's right.

Speaker A: Right.

Speaker C: That's

Speaker B: right. So. So I something I've been obs over probably the last decade of kind of extracting the last. The best practices. So I've been calling this the fix. I call it the fix flexible and the custom. So the fix. So which in a way is your, your goal? Yeah.

Speaker A: Well, and fixed is kind of your centralized. Right.

Speaker B: So fixes your centralized is stuff that you know that, that it has to be this way. Right. And uh, just for regulatory purposes, because this is how the business defines it. Like that's it

Speaker A: when we say revenue, the company.

Speaker B: And this definition,

Speaker A: there's a definition for that.

Speaker B: The other example I always bring up a lot is like what is account information? So account information, contact, contact information. Well, you have a phone number, phone number can be a cell phone, a mobile phone number, home phone number, mobile phone number means that I can send a text message. Can I send a text message? Yes or no. Can I send a voicemail? Next? So you have that. Um, then you're, you're flexible. Is. You take what is already happening, you're fixing, you extend it. And custom is. Doesn't exist. You'll build that stuff. Um, and then there's use cases for each one for the flexible ones here you have an application you want to go build and you look into the fix. Oh, it's almost there. But I can reuse it. But there's things that's not there. So I'm going to go over you. I'm going to reuse what I have and extend what I don't. And custom is. Doesn't exist. I'm building this all new. So you kind of have that. Different models. But the important thing is that to build a culture of sharing of, um, I built this new thing and I'm going to bring it back and put it back into my catalog, my metadata management system, people can go find that. But then some pushbacks that I will hear from folks is that, wait, but then we'll have different definitions of customer or different ways of modeling this. And no, we should not allow that. I'm like, uh. Or we should allow it. Let it happen. Let. Let chaos happen. And, and if something, if there is friction, that's an indicator people care about it. And now they're incentivized to work together to kind of align. And if, and if you have different definitions for something and nobody's. Nobody's complaining about it, then nobody cares, it's fine, or nobody's using it. So that's, I think that model that I'm seeing, which I think aligns with like, this whole. It's like a medallion layer for context. It's like, definitely there's not a single version of a, uh, of a context.

Speaker A: Your idea of sharing it is kind of like, don't, don't, don't embrace, like, shadow semantics where, like, people aren't sure. Like, oh, you created that one dashboard. Like, what's, what's underneath it. Like, make it visible, make it clear. And if people disagree, then they should disagree on m. Like, hey, why did you, why did you do it that way? Maybe it should be different.

Speaker B: Well, I think, I think. Sorry, I think Ryan's a bit jealous he's not here. It's an amazing thing when your, uh, friends bust your balls on LinkedIn from a brewery in Austin.

Speaker A: Appreciate you, Ryan.

Speaker C: You know, I do think, though, you know, transparency is really important. And look, I've been, um. I've run data organizations, and my biggest pain was not how quickly I could deliver something to market or the actual dashboard. It really was trust. That was. I mean, if you, uh, if I was going to be hurt by something, it was when Juan says, sonny, I don't trust what your team is doing. I don't trust these answers. That is, as a data leader, to me, that's our kind of kryptonite. And so we're trying to, we're trying to protect that at all costs. Right. So how do I make sure that I have trust. Well, transparency, interoperability. And then one of the things you mentioned was, well, we've got to test this stuff. We do need to know and have everyone says evals now. And what we do mean is testing. Because there has just traditionally not been enough of that in our dashboards. You know, I'm an expert. I know what this is. I put the formula in. I go,

Speaker A: right.

Speaker C: And I don't believe in duty valves and testing.

Speaker A: Yeah. I think we have to be thinking, taking a much more testing mindset to semantics and to dashboards and things like that. And I think one cool use case around agents that I've been seeing lately is, you know, it's. It's hard to do testing. Like, you know, we love to build, but maintaining and testing is. That's annoying. Right. But you know what? Agents can help you with testing and finding where there's inconsistencies and things like that. And, uh,

Speaker B: that's. It's an easy thing to have an agent say, go build a bunch of test cases for all the corner cases and go run that stuff.

Speaker A: Or go look at all the dashboards that people are building. And like, where do you think that there are people using the same definition for the same thing, but the semantics are different? You know, things like that.

Speaker C: Yeah. And absolutely. Those are, to me, fabulous areas. And. And I'm sure we've all written some stuff, written some code. I would not say I'm a vibe coder. I'm actually a computer scientist. I. I actually. Wait,

Speaker B: you're not Vibe, Cody.

Speaker C: I'm. I'm actually writing software. But yes, I use agents to help me, but I'm not a vibe. By

Speaker A: the way, this is a funny thing I've noticed. I

Speaker C: would never put mine.

Speaker A: Is that people who are computer engineers or computer scientists hate to say that they're vibe.

Speaker C: Absolutely.

Speaker A: Because they're like. No, no, no. They're like. We don't. Wait, wait, hold on.

Speaker C: I'm a professional. I'm trained.

Speaker A: I'm a natural language.

Speaker B: Wait, but. So, all right, we're gonna change the topic here because I'm a computer scientist, and I actually am. I. I think it's pretty cool that I. I'm vibe coding now. I think our definitions by coding may be different, therefore, so. So I think we're gonna agree on this. Our semantics are different.

Speaker C: There you

Speaker B: go. So for me, Vibe, Cody, what I vibe code is I spent so much time up front figuring out what I want to go do, and I. I spent more time, like, with the My one entire session of, um, it's not even. It's not even in Codex or quad, right? Clock code. It's just like we just. I'm just brainstorming what I want to go build, but I'm trying to go build this thing and I want to make sure that my architecture is as simple as. Let's figure out. Break that problem. Break the problem down as far pieces. They spend so much time into that. And then when I look like, okay, this actually makes sense to plan where you go do. Then I say, now generate the prompts like I said, the Codex or clock code. And then I send that in and then that's the Vibe code. And then it just happens. And that shit just works now.

Speaker C: But. Yeah, but that's where we. What we've always meant to do, right, was to actually understand the product that we're building, the requirements that we want, the design that we want. Because there's always trade offs, right? And when there are these trade offs, are you letting quad or copilot or somebody else make the decision on those trade offs or are you doing it right? And so that's kind of where I am, I think. Just in general, what you're seeing is coding it. Writing the code is no longer the long haul, right? Everything that's getting pushed upstream, um, to product management, what you were talking about, Product management is now the bottleneck. I need like one product manager for every developer. Right. Um, and from a testing perspective, um, testing is now the bottleneck.

Speaker B: All right. We've been ranting for almost half. Uh, the honest hope you guys is. I think we still have a couple more meetings. Any last minute thought, this has been a fantastic grade. I'm glad you were able to hang out. Uh,

Speaker C: what

Speaker B: are your final thoughts? We started off with Dashboard Slop.

Speaker C: Yeah.

Speaker B: Your final thoughts?

Speaker C: My final thoughts are, uh, try to stay away from Dashboard Slop. Right? You know, take your time, make sure you understand your sources, and make sure you can answer any question. Because guess what? AI isn't responsible for what's written. You

Speaker B: are. There you go. Love it.

Speaker A: That's great.

Speaker B: Funny. Cheers.

Speaker C: Cheers.

Speaker B: Appreciate the

Speaker C: ranch,

Speaker B: folks. Ryan. Or, uh, there's another one before we leave. What about this? This one from Ryan Duffy?

Speaker A: Uh,

Speaker B: when it comes to that. Yes. We need to have more of those screens. Just shut it off and just see who screams.

Speaker A: Especially the AI slot. Dashboard.

Speaker B: Yes.

Speaker A: Shut them down.

Speaker B: All right, all

Speaker A: right. Cheers, everyone.

Speaker B: Cheers, everybody. It.