The Data Stack Show - 250: The Chat Interface Debate: Is Text Really the Future?

Episode Date: June 25, 2025

Highlights from this week’s conversation include:Dashboard vs. Chatbot Discussion (1:40)The Future of Chat Interfaces (3:03)Vibe Revenue Concept (6:36)AI's Early Days Compared to the Internet (10:14)OpenAI and Hardware Collaboration (13:09)Challenges of Hardware Development (16:20)Productivity in Programming Languages (18:18)Critique of 'Language is Dead' Posts (21:34)Legacy Systems in Use (22:34)Deterministic vs. Non-Deterministic Workflows (23:37)Final Thoughts and Takeaways (23:45)The Data Stack Show is a weekly podcast powered by RudderStack, customer data infrastructure that enables you to deliver real-time customer event data everywhere it’s needed to power smarter decisions and better customer experiences. Each week, we’ll talk to data engineers, analysts, and data scientists about their experience around building and maintaining data infrastructure, delivering data and data products, and driving better outcomes across their businesses with data.RudderStack helps businesses make the most out of their customer data while ensuring data privacy and security. To learn more about RudderStack visit rudderstack.com.

Hi, I'm Eric Dotz. And I'm John Wessel. Welcome to the Data Stack Show. The Data Stack Show is a podcast where we talk about the technical, business, and human challenges involved in data work. Join our casual conversations with innovators and data professionals to learn about new data technologies and how data teams are run at top companies. How to Create a Data Team with RutterSack Before we dig into today's episode,

we want to give a huge thanks to our presenting sponsor, RutterSack. They give us the equipment and time to do this show week in, week out, and provide you the valuable content. RutterSack provides customer data infrastructure and is used by the world's most innovative companies to collect, transform, and deliver their event data

wherever it's needed, all in real time. You can learn more at ruddersack.com. All right, welcome back to the Data Stack show. Got Matt here with me, Cynical Data Guy, and we're gonna do a little segment on LinkedIn posts. So Matt, welcome back to the show. Thanks for having me. We're going to miss our fearless leader today.

So I don't know. We'll see how this goes. We got we we got it. We got it. All right. We're going to start out with this one. You're going to love this one, Matt, because I've yet to see a LinkedIn post that has to do with data that does not have the word dashboard in it for at least a month. That's true.

Yeah, it's, I don't know. I don't know what's going on with that. Okay, ready? Yep. If you, and this is a meme, and so I can't give you the full context here. Maybe Matt can help out. It's a Futurama meme.

Futurama meme, okay. So here's the meme. If you have to set up each answer in advance, is it a chatbot or a dashboard? Sounds like we're getting to like, it's not a dashboard, because it's just text. Oh, so it's worse than a dashboard. But yeah, and I think, and by chatbot, like, I think think obviously like AI chatbot is what we're going here and I do think we're gonna get into this thing where To get consistent outputs. You're going to have chat interface that end up with very deterministic results Yeah, it's gonna be picking through the Germanistic patient of that bro. Yeah

Right, and it's gonna be like wait. Did we need AI to do this? And then it's also going to be, should this be a chatbot? Yeah. I mean, the biggest thing that makes me think of is like, what's one of the uses of the dashboard? I can look at multiple things and compare them at the same time. Now I have to ask multiple questions and you sit together. Why am I doing more work at this point? Yeah. So this was a really interesting moment. For those of you, we've had a couple of shows recently released from data council that we attended. And one of the really interesting moments from one of the talks is there was, it wasn't a standing ovation,

but it was one of the only like, just kind of spontaneous cheering, it would be a good way to put it, when one of the presenters said that the chat interface is not the interface for the future, it's not the interface that every AI thing needs to use. Yeah, I think that's true right there. I mean, it serves a good purpose in a lot of places if you're asking questions and you need to text back,

but yeah, let's harangue everything into a chat interface is a bit much. Yeah. Though it might also be proving to then stats. It's like, yeah, his thing that he posted on multiple times in his blog, which is that everything becomes BI, everything becomes the dashboard, right? All companies eventually become a BI company. So there you go. Right. becomes the dashboard. All companies eventually become a BI company. So, there you go.

Chat AI companies are becoming a BI company. Yeah. I mean, yeah, it is funny because all of modern SaaS, right, if you break it down, really is some kind of like web form storing some kind of information and presenting it back. Almost all of it is that. Well, no, it gets real bad when you start having to be able to have the,

when it starts asking you, would you want to export this to Excel? That's when you've truly gone to, you're just the BI company. Yeah. Well, and the other thing that I think is funny, obviously, I mean, the chat is a good interface for LLMs in their current form. But if we actually look at like the most engaging

and you actually have a, you have a poster you're gonna share in a minute that's somewhat related to this is if you look at the most engaging apps and the highest, whatever the, it used to be dwell time. Like now it's, you know, engagement time or whatever. It's video and picture. So to say that like oh everything is gonna be chat it's like well not so far we are a visual species right I mean maybe there'll be something involved with this all right you've got one for us okay so this is

this was posted by Tom Goodwin on LinkedIn. So thanks, Tom. This is every person alive. I really need to check my phone less and spend less time online. Every tech company alive. Wouldn't it be great if you could strap the internet to your face and have notifications as tiny electric shocks?

Ugh. Every person alive. I wish I didn't have to get my kid a smartphone. These things are amphetamines to the eyes and essentially rob our kids of childhood. Every tech company alive, why don't we make Tinder but for kids? And like use AI so you're not actually talking

to a sociopath, it's just AI pretending to be your friend. And that would lower our cost base and increase TAM. And then that was, he followed that up with a quick comment here where he said, I have said this many times before, but there are absolutely no more poorly equipped people to lead us into a wonderful future. And the people generally leading technology companies, they exist to optimize lives away against poorly conceived KPIs. John, what are your thoughts?

See, when you pick the posts like this and you get to read them, that means I have to respond to them first. So I feel like that's fair. But I'll play. Now I have to look like, okay, business. He's in consulting, it looks like. But yeah, I mean, I would say compare it. The end of it is like, well, compared to what?

They exist to optimize lives against poorly conceived what? Are they optimizing shareholder value? Yes. Is that what every for-profit company is doing? Yes. Could they do a better job and be more socially responsible, et cetera? Yeah, probably. Well, yes, definitely. I think he's probably going off of the engagement or the usage or those metrics. And I think there's merit to that. It's also like, you know, that's TV and radio and all that. We have it just on, you know, even growth hormone.

Yeah, here's a funny term I hadn't heard until recently. You familiar with vibe revenue? Vibe revenue. Yes. Okay. I've heard of Vibe. I don't know if I've heard of Vibe revenue. So the concept, I guess here, is essentially like a lot of these like rocket ship

fast growing AI companies that have like, oh, I'm curious revenue or hey, what is this revenue? That would be a revenue. And a little, gonna repeat very long revenue. Right, so, and there's probably other categories of Vibe, but the interesting thing is I do think the KPIs are gonna change. Like I think he has a point with the poorly conceived KPIs

is that there are gonna need to be more like better ways of tying KPIs to business metrics essentially. Where if it was before it was just like user engagement in general, like we need a better understanding, what are they doing? How does that transform their business or their lives or whatever like business you're in? Until they figure out how to stick advertising

into all of this, then it'll just go back to the same old thing. Yeah, well fair. Well, yeah, I mean, it's all gonna optimize around whatever the monetization of it is. No, I mean, to go back to the first part of his post, I do think we've seen this where like, I don't know if you've talked to any really like let's

call it in the bubble tech people, their answer to everything is like a social network. It's like, no, nobody wants that. It's like we have a loneliness crisis. What if we created another website with AI that could interact with you? And it's like, no, stop. Go that way.

You're going in the wrong direction. So I think there's something there, but yeah. I also just, you know, it just hits that nice little cynical side of me. That's what I love. That's just... We're gonna take a quick break from the episode to talk about our sponsor, RudderSec. Now I could say a bunch of nice things of me that he loves. to make sure that data is clean, downstream tools. Now rumor has it that you have

implemented the longest running production instance of rudder stack at six years and going. Yes I can confirm that and one of the reasons we picked rudder stack was that it does not store the data and we can live-stream data to our downstream tools. One of the things about the implementation So if you need to stream clean customer data to your entire stack, including your data infrastructure tools, head over to rudderstack.com to learn more. All right. So this one, this one's from Joe Reese. He's been on the show before. So shout out to Joe.

Joe! Love, I really like this take, especially just because it brings back memories. So you ready for this? All right. With AI, we are still in the AOL, CompuServe, EarthLink, et cetera days. Wall Gardens, or BBS2, Wall Gardens like today's AI. At that point, the web browser was very new. So think like Netscape, right?

It had changed everything. Then there was mobile, you know, et cetera. Perspective matters. It's very early in this game. No idea what happens, but it's likely going to be cool, scary and very unexpected. The fact that I know what all those things are, does that mean we're getting old at this point?

Yeah, you've got mail. I actually never had an AOL account. I did, but you know why I had it? Why? AOL and Sim messenger. Oh yeah. I think I did have that, but I didn't actually use the mail. I Yeah, well, Instant Messenger. Ah, yeah. Yeah. I think I did have that, but I didn't actually use the mail.

I had a hotmail account. Yeah, I don't remember using the mail very long, but I had it because of Instant Messenger. Yeah. That for those of you, it's Slack, but different. Slack, but it'd be more basic. Yeah. For a being-year-old.

But I actually like this perspective a lot in that it's easy to go down the well trodden paths of like, of course it's gonna like be multimodal and there's gonna be like better video things and better photo things and you know, all those things, which is the path the internet traveled. Yeah. But I don't think it is that deterministic.

I don't think it's like, well, of course AI is gonna like do video, then it's gonna do photo, then it's gonna do this. I think it's different enough, there'll be novel things that come out of it, just like with the internet, that will legitimately surprise people. Well, I also think, you know, if we're going through this stage of it, we're probably also going through the, you know,

internet bubble stage too. There was a lot of like, you just put it on the internet, put it on the internet, and you had companies that had, we're losing tens of millions of dollars as public companies and have no basis for how they were ever gonna make money, and yet their stock went up in the late 90s before it all came crashing down.

So there's, we're starting to see, I think, a little bit of a pullback on some of the worst excesses of the AI expansion, let's call it. And I do think he makes a good point there, and I agree with, which is nobody really knows what it's going to look like. Of all the things that mainstream people were predicting, mobile was not something that people saw in 1998, particularly because mobile phones were not a big thing. Right.

And then they were just making calls. They didn't hook up to anything. There was no screen on them, really. So, you know, you had all those things that like, it's just, it's a good reminder too of just, we don't actually know what's gonna happen. And while everyone loves to try to make straight line predictions, the future is almost never a straight line prediction from any moment you're in right now. Yeah. Did you see the announcement with OpenAI and Donnie Ive's group for the hardware company? I think I saw something about it. I didn't read any of the detail on it. There were very few details. Oh, so I basically did read the details. Oh, good.

The details. No, I mean, it's interesting, right? Like for pretty much all of you know, it's like, all right, let's take the guy who was instrumental to all of these like truly transformational devices through his time at Apple. And let's collaborate with OpenAI and the best of AI and then creating hardware.

Like, cool, yeah, good concept, but it's quite hard to imagine. And apparently they have like a prototype and like people use it. There's no details about the prototype that I'm aware of. It's a prototype, I don't know about that. There's some sort of device. But here's my question to you.

Where do you, what do you think you do? So as far as like, is this meta glasses style? Is this some kind of handheld device style? Is it, where do you think you go with that? So they didn't say this. Not that there's no, maybe some, maybe there's rumors I don't know about, but I think we have essentially zero information. Where would that, I don't know.

So if you're trying to do something, cause like, so obviously with Ives there, very instrumental in a lot of Apple's designs and how they work. I kind of think it's going to be something voice related. Yeah. But like exactly what, I mean, glasses, feels, I mean, obvious or whatever. I don't know.

I mean, it would be a tough one. I really hope it's not like a pin. Yeah. That's of no interest to me. The like, let's put an Alexa button on your chest. Right. No thanks. Well, I mean, I don't know what was that company that like did the pin that I think already is like, have they already like called it? Yeah. It wasn't like Humane or something.

Yeah. They're already done. They're already done. Yeah. Cause it was weird and dumb and did you refer to it a lot? Yeah, I mean, I think, here's my thing with voice. That's, now that you say that, I do think I read some predictions and I actually even talked to somebody earlier today.

It was like, it's gotta be voice. Personally, yeah, okay. I think I see that. Professionally though, like, what if all of the interfaces in the office, in your office, like changed to be more voice heavy? Like, I don't know, that just feels weird.

It would have to be through headphones at that point. You'd have to get up to like, you know, to some type of like AirPod or something. I guess. I mean, yeah, it's, but all this stuff, I mean, it's even, okay, so you're gonna do something with that. It's like, for it to really work well, there's gonna have to be something heavy somewhere on you.

And that's, I think, one of the biggest problems is like- Like some kind of battery, some kind of power source, yeah. I mean, even if you look at like Apple's, you know, like VR and stuff, it still had that rick of a battery that you're like putting in your back pocket or something, right? But I also think one of the things that people forget is because I think these are like Oh cuz Apple did this and it's like hey guys, you gotta remember Apple

Started as a hardware company, right? It has been a hard work right ever since Consistently. Yeah, yeah, none of these other people who try to copy them have ever been hardware companies, they've been software companies. And there's a big difference between those. So I always view these types of things with a bit of a skeptical eye, just because hardware is hard. There's only a few people who do it really well, and it takes a long time to get there. And it's still super capital intensive.

Super capital intensive, which I think like, you know, the Zuckerbergs of the world have found that out the hard way. Right. And it's just, you know, it requires a lot of expertise that you can't, like, it's not going to exist internally with you. You have to go out and get it worked. It's not something you can train up in six months. Right.

This is a, you know, you need people with decades of experience. Right. This stuff. Right. Okay. I think you had another one. All right, so here's my other one. This is another LinkedIn post. I'll leave the name off for the moment. And it says, R is completely dead as a language.

And I say that as someone who was once a strong proponent of it. The only use case I can think of for R at this point is to verify that my Python or sometimes Node.js code is correct. That is literally it. If you are building an R, you should stop and switch to a mainstream stack. Making this change will allow you to go at least 10 times and probably 100 times faster. Bolt egg. I have thoughts on this, but... You need me to anchor you, need me to start.

I can start if you want. I got thoughts. One, I, art just has a really special place in my heart, so I would never declare it dead. Two, it's open source. So anything that's open source has been around forever and has super high adoption Lindy effect. Like it's not gonna like die because you said it well,

just because, you know, that's just not how life works. But I think my biggest qualm with the post actually is, and I think this is referring to productivity. Yeah. As far as like 10 times or a hundred times faster. I don't think it's talking about code execution. I think it's talking about productivity. Right. I do't think it's talking about code execution. I think it's talking about productivity. Right.

I do not agree with that. I think R, at least for the workloads I've used, like analytics is very efficient. Like I've never had to like, wow, this is like so verbose. Like I got to switch to Python. That's never been the thought I've had. I haven't written a ton of R, but like I've never had that thought writing R.

All right, what's your take on it? So I've got two levels of takes. So one is about R specifically, and one is about the genre of posts. Sure. So the first one about R, so that was the primary language I used for like the first, I mean, I used it as an analyst,

I used it as a data scientist, used for a long time. So like you, it kind of has a special place in my heart with there. Where I did eventually have to, when I was running Teams, which over to Python was when you wanted to productionize it and you wanted to hook it up to like existing tools. Sure. Like pre-commit and a lot of the GitHub automation type stuff, GitHub actions. It was so hard to do with R because people just don't build.

Yeah, right. Whereas Python was easy. Yeah. And Python was easy. Yeah. And like I was in a Google shop, everything they were like, if you use Python, it'll just work. Oh, you want to use R? You're gonna kind of do that all yourself, sorry.

So like when I was doing that, when I ran a whole department, we switched to Python because I was like, unless you absolutely have to do it in R, if it's gonna go out, this is the fastest. So if you're talking about productivity in the sense of like, can I get it out there and make changes quicker? Yeah. Yeah.

That would be that way. I can see going to like a Python works just because everything's made to work. Yeah. Yeah. Like this is one of the first things it is. But I mean, I also, I ran a team that we owned production code and it was, you know, and it was pricing loans and it was all in R and it was fine. And we had a CI CD for it, but I mean, we, plus and it was, you know, and it was pricing loans and it was all in R. Right. And it was fine. And we had a CI CD for it, but I mean, we thought it was custom built for

it, but you could still do it. It wasn't slow or anything like that. Well, as long as you didn't try to do like dplyr in production, don't do that. No pipes, none of that. I had to give the data scientist another, just make the time came through about that one. So, so like, yeah, it's not. I don't think it's dead. It's also like, I don't know. So I first started, you know, I had to learn SAS, like SAS. Yeah. And as I was learning that and got certified, everyone was like, SAS is dead. SAS is not very much not dead. Yeah. No, like Bank of America has like over a million of code of sass. It's not

going away. But these things don't go away. So that kind of leads to my second one, which is kind of the genre of the post, which is, I am so tired of this. Why are we still arguing over this stuff? Why are we still trying to like, what's the point of saying, well, it's dead. Really? Don't use it. Who cares? Because. Because none of this stuff dies. No. Hobalt is still out there. Fortran is still out there. If it got heavily used, it's still out there. And it's like, there's a lot of places, and there's a lot of regulatory situations where like,

you have to use R, because it's the approved language. Sure. It was the same thing with SAS at the time too. So I just, I get very tired of these. They feel very clickbaity. They feel very like, I I'm gonna rage bait for clicks and stuff like that. It's like, because I remember the old Python verses bar,

which one was the right language? There is no right language. Well, and here's what I think the interesting thing about, because there's a whole circle of SQL is dead posts going around again. We had a break. We had a couple of months break and they're back. Yeah, I didn't notice that.

I miss the old days. But here's my take on all the dead is dead post is if you're able to solve a problem completely deterministically with any programming language, why from everything we know currently about LLMs, why would you move it to a non-deterministic solution? Yeah. Like there's so many things in the world

that need to be improved, that need to be automated, that need to be streamlined. If you have a streamlined working deterministic solution, I think those are the very last things to get replaced. If they do get- If ever, if ever. And the only reason that people will start replacing them is essentially nobody knows how they work.

So they have to write it in this like non-deterministic way. Cause I don't know how the deterministic way works. That will happen. That will probably happen. It's not ideal, but I'm also pushed back. I wrote an article this morning about how there's all of these places that are still running things off of dos three point up and windows Windows NT and stuff like this that it's like, you know, you're like, oh, those are like, you know, those aren't major things. Like, oh no,

that's like in, I think it's in Germany, the display panels on all of their trains are still in DOS. You know, these things don't go away, particularly because if it works, why am I going to replace it? Why am I going to take the time to go deal with that? And to your point with the deterministic thing, there's lots of stuff that these LLMs can work on. I mean, I'm working for a company that does that. It's a lot of things with summarization, going and finding answers in unstructured data and stuff like that.

But you don't, it's like, well, but it's going to replace SQL. Why? There's not a and stuff like that. But you don't, it's like a little bit where it's gonna replace SQL, why? It's not a good reason for that. There are deterministic workflows. They probably should be deterministic. The thing I need to do is deterministic. Let's use a tool that's deterministic. Right, for sure.

All right, I think we're at the buzzer. Any final words for all those snows? No, I think we're good. We think we did it. Remember people, languages don't really die. They just kind of fade away to be fuzz that will try to kill you. It will haunt you.

Yeah, exactly. All right, thanks everybody. Stay cynical. The Data Stack Show is brought to you by Rudder Stack. Learn more at rudderdersack.com.