Why the AI slop problem is getting worse | David Okuniev (Typeform co-founder)

The AI slop problem is getting worse. AI can crank out code, decks, and landing pages in minutes, but somehow everything starts to look exactly the same. That sameness is not just an aesthetic problem; it is a trust problem. I sit down with David Okuniev (Founder of Typeform and Supercut) to unpack what “taste” really means when anyone can one-shot a product, and why people still crave signals of real craft.

We also dive into the hidden cost of easy creation: from small-team dynamics to a "65 PR backlog," we discuss how AI shifts the bottleneck from writing code to building a harness for shipping safely.

Beyond the philosophy, we get practical about the future of async work. Supercut’s bet is that the winning Loom alternative is not just about recording your screen, it's about searchable transcripts, AI Q&A, and agentic workflows that turn a video into a document your team can act on immediately.

If you care about product design, reliability, and building honest software in the AI bubble, this conversation will sharpen your thinking.

Transcript

AI Slop And Honest Building

SPEAKER_00 0:00

I'm not writing any code anymore. Neither is anyone in the team writing any code really. We've been receiving, instead of receiving decks, we give them a case study. It's always a lovable deck. And it always looks the same, the same way. It's like they they went to lovable, they said, hey, I want to make like an animated deck for supercut.ai. And it just creates this like, well, what we call AI slop, right? It's not so much like the using AI, it's just one-shotting with AI. Because I use AI. The point is that I'm I look at it, I iterate, and I get it to how I think it needs to be. And that is the human element. But when you get something that is devoid of that, that creates a complete disconnection. It's so easy to build and launch right now that you can do all this smoke and mirror stuff. You promise all this stuff that you're at that your product can do, but then you try the product and it's never up to standard. I just don't feel that comfortable just like lying about it and like bigging myself, bigging up myself and my company up, which many people are just completely comfortable with doing. I'd rather just build that value and people try the product and not be disappointed, right?

Typeform Roots And Beating Loom

SPEAKER_02 1:15

Welcome.

SPEAKER_00 1:16

Thanks, Alex.

SPEAKER_02 1:17

Long time no C. Thank you for doing this.

SPEAKER_00 1:21

Last time we were recorded.

SPEAKER_02 1:24

Exactly. Lots of them, actually. But for good reasons, I presume. Um last time we spoke, it was at um one of our podcast episodes. We recorded in the pandemic times. Yeah and uh we discussed you know typeform at large uh because uh you've been a long time friend of the company. Uh I interviewed you, I think, for the first time at a StarPrain conference that we organized 2017, 2018, something like that. Yeah and been following you and and typeform. For me, Typeform was the uh one of the first companies, if not the first company, that made software sort of delightful and with taste. You guys made uh an incredible accomplishment, which was you you made it exciting to integrate web forms, which is a fucking big thing to be honest. Everybody gives credit credit to linear and superhuman and all of that, but I think it all started with type form. So I'm flattered, but yeah, nice. On the subject on the subject of taste, we'll be covering this a little bit later. You were mentioning before hitting record that uh you are going to follow the uh linear playbook to sort of uh overcome the incumbent in your space now with SuperCat, which is Loom. Loom is a company that is like five, six years old or something like that. It's older, it's older, but it became popular during the pandemic.

SPEAKER_00 2:50

Yeah, during the pandemic, it really blew up. Yeah.

SPEAKER_02 2:53

How do you plan to do this with Supercat?

SPEAKER_00 2:57

Well, I think it's a long slug of building product, improving it, working on the brand, getting key customers. It's just it's not one of these products which everyone just like suddenly moves over to because they're already using something, right? That does a similar job, right? So we are SuperCart does the job of recording your screen, but we're trying to make it do a lot more than that. So we're all about making move uh making work move forward beyond just you know recording your screen, like what happens after? Do you give the information to your agent? How does that get translated translated? What is available? How does it work in between teams, etc.? So we're really focused on trying to like bring in the next chapter of what video messaging, what we call, is going to look like in the agentic in the agentic era.

Why Supercut Goes Native

SPEAKER_02 3:52

What did you find broken in the workflow of using Loom or other similar tools?

SPEAKER_00 3:58

I I think most things out there just feel quite cobbled together and not really love for as far as the details. Um not to pick on Loom because they're not the only ones. And I want to set aside companies like uh Screen Studio that, for example, do a really good job, but we don't see ourselves as screen recording. We see ourselves as as video messaging, and it's not like video production. It's more about you know using video as a medium to get your your message across. And uh what we see with the leaders in this space is that there's not really much care for like the details. The software is usually quite clunky. Um, for example, uh we hear a lot, we hear a lot of this from Loom users that you know they lose recordings, it stops, fails in the middle, like it just it's a it's quite buggy. And that's probably because it's based on web technologies, and there's so much you can do to really keep a video stream alive on the computer. It's quite difficult because it's a balance between what's going on in your machine with memory, et cetera, and sort of sort of managing that, especially given the fact that we don't just record your screen and camera. Well, we we don't just record your screen, we we record your screen and a separate stream of your camera. So we can composite that on the fly into a video. So, for example, you can change the layout later. Doing that is also resource intensive. So for us, it made complete sense to like build this stuff natively so we could manage resources as best as possible to have as much reliability as possible. So that's one side of Supercut that we heavily invested. There's also, you know, design, UX, and really trying to make a more beautiful looking player on the out, uh, let's say, for people when they share videos, uh internally as well as externally. And you can also brand these videos so you can add your logo, you can add backgrounds, everything kind of adapts. But it's really the main thing is really about recording with stability and being able to uh share the link as soon as possible. So as soon as you finish recording, you can share it and someone watches it. So that's the core technology. And then on top of that, of course, you've got design uh and and brand. Um, and now we're really starting to push into the more genetic workflows.

Agentic Video Messaging Workflows

SPEAKER_00 6:24

What happens after you record a video? So, for example, right now we have an AI assistant that's on the video page. You can ask any question of the video, it'll it'll look at the video frames, it'll look at the transcript and answer you questions or generate documents for you. Uh, we also have an MCP coming out, uh, which basically has access to all your activity on Supercut. So even a comment that you said on someone else's video, you can ask your agent anything about your activity on Supercut and it will take actions for you. So, one example could be uh morning stand-up, we do async. Um let's let's have our agents listen out to what we're reporting each day and post a Slack with screenshots of relevant stuff so that everyone is up to date, um, so that no one has to watch the video, right? Video is a great medium for for for expressing yourself, but sometimes, sometimes it's not so great for ingesting.

SPEAKER_02 7:25

It's not great for productivity, it's not optimized for it.

SPEAKER_00 7:29

Exactly. It's like you've got a 10-minute video that you need to catch up on. You don't just want to watch the whole thing. So you'll watch it on double time. But beyond that, maybe it's just best to you know get the the quick insights, get the summary, jump into key areas, ask questions, and just you know, get what you need from the video. And that's what I'm saying, like using agentic workflows on that can really speed you up. And you know, we've got some some customers that are now really switching onto this because everyone is scrambling to, you know, move faster and put agents to work. And I think SuperCUT is a really good place to play in becoming your eyes and ears of your screen, right? Because it's high fidelity information that just writing down doesn't capture everything that you want to say. That's why people use voice mode or paste screenshots into clawed code into clawed code just to give it context. Video gives that real rich context layer.

SPEAKER_02 8:27

Exactly. One of the things I realized, for instance, we we tried going async during the pandemic, right? Because everybody was going remote and therefore we lost our edge in terms of competing against other companies for talent. Then we decided to try async. Also, we were always sort of async, but we decided to go almost full async. Yeah. We tried video tools and we realized that video messaging, even though it's really good for the sender, yeah. Like you're able to kind of like brain dump, convey more information, and and maybe lose the constraints of I'm not really good at writing stuff or the blank canvas syndrome. Yeah, it's not good for the receiver.

SPEAKER_00 9:08

Yeah, yeah.

SPEAKER_02 9:09

It's kind of like when you're look looking for something on Google or perplexity nowadays, and you just want a fucking cake recipe. And no, now you've got this 10-minute video that explains how to do this recipe. No, man, I just want to quick glance, show me step three and four, right? How do you solve that?

SPEAKER_00 9:24

I think it depends on the use case. Yeah. I mean, for example, if I was looking for a recipe video, I do actually want to see the video and how they they do stuff. Yeah. There's many videos where I just want to extract the the information and what is the key information that I need to know about this video. Yeah.

SPEAKER_02 9:41

So like I said, you're working on that because you mentioned you got the transcripts and the you can have you can chat to the video.

SPEAKER_00 9:47

You can chat to the video, but also you can extract transcript frames uh uh and list of all the videos that you uh that you've recorded, categorized, etc., for your agent to ingest. So if you report a bug report, you can ask your agent to be listening out for that and post it for you somewhere. So that part is is really

Video Is Hard To Consume

SPEAKER_00 10:10

important. Just be able to get through the vid through the content of the video as fast as possible, especially for internal use cases, right?

SPEAKER_02 10:17

So do you do you do you see that the use or the consumption of the video is less than the production of the video? Like do you do you feel like people send more videos, but then they're not consumed by humans, but for by AIs?

SPEAKER_00 10:30

Well, I think that's that's the part that's challenging. Okay, right. I think traditionally you record a video and then someone sits down and listens to it on 2x speed.

SPEAKER_01 10:40

Yeah.

SPEAKER_00 10:40

And I still listen to a lot of videos, but there are some things, like, especially for my team, it's like just one like get to the point. I just need the essentials. And they understand why that person uh recorded it in video, because what what happens when you when you write, you you just don't have time or or you just get lazy to like include all the context around. But when you talk and show stuff, you can you're just kind of riffing. So you can just go off on one, and you in in that there's like really important information that can be extracted. Yeah.

SPEAKER_01 11:13

So it's just rich as well.

SPEAKER_00 11:15

Yeah, a lot of fluff. So you know, someone doesn't need to watch a 10-minute video, right? Yeah, they just need maybe like three or four minutes of of that video, or just the key insights for it, or the key things that they need for them that concern them. So that's why GenTech makes complete sense here in this uh it makes complete sense.

SPEAKER_02 11:34

For instance, I I for one, I don't listen to audio voice uh voice notes on WhatsApp anymore, just forward it to another conversation I have the transcripts.

SPEAKER_00 11:42

I wonder why WhatsApp doesn't transcribe those audio messages. Because that it's that's really annoying when you get an audio message that you don't get the transcript, right?

SPEAKER_02 11:50

You're a product person, you probably can answer this better. My take is when a company with infinite resources doesn't do something that is blatantly obvious, it's because it's delivered, it goes against their metrics.

SPEAKER_00 12:01

Yeah, yeah, yeah, yeah, yeah. Or it's just a deliberate product decision. I don't think it's like, hey, we just forgot we'd have time. It's like so blatantly obvious.

SPEAKER_02 12:09

No, but then then maybe then it goes against, like, you know, oh, um, it would maybe make the I don't know, the usage KPIs of audio go down, therefore somebody else without context would see that and see like, oh, audio doesn't work. Yeah, yeah, yeah. Why are we investing in this, right? But anyways.

Small Teams And AI Coding Reality

SPEAKER_02 12:27

Tell me about the team, because one of the things that uh by, you know, in the recent times or in the last year, year and a half, we're seeing massive changes to the composition of the teams, especially affected by you know the impact of AI, right? And we're seeing that with fewer people you can build faster. I'm not saying better, just faster. How big is the team now, and how big would it have been without AI?

SPEAKER_00 12:54

So we're gonna be eight people soon. Um I think at this point to get as far as we we have, uh yeah, I would definitely say we would have needed a bigger team if we didn't have AI. That said, uh at Typeform, I was after I stepped down as CEO six years into the journey, I started Typeform Labs with the idea that we could work really fast with less people. I do still, I do still believe that that is the case. Obviously, AI is an enhancer on top of that. But having really good focused people that you really trust and everyone's really like batting in the same direction, you could you can already move fast. I was coming from a company with like hundred engineers, right? And we saw in type four labs that we're just with a very with a few people, we could like do like loads and loads of stuff, like launch entire product lines. So yeah, like I say, I'm a believer in in small teams, and Float Labs was set up with the per with the idea that we can build a world leading product with a small team. Like with that idea, I I I didn't maybe it's just because like I was scarred from like Typeform and like seeing all the problems that come that come when you have so many people in a company. Uh maybe that is the case. I was scarred from that, but I did really want to make sure that my next gig was just really small and controlled. So when I left Typeform, I left with uh Neil Kinnish, who's my co-founder now. He was also in Typeform Labs. A few other people joined from Typeform Labs over time. Um, but we've remained small and moving fast. Now, what AI has done has really enabled us to move even faster. Um sometimes it's it's it's it's it's a bit cloudy how much faster you're moving because, for example, I'm I I've been coding for five years, mostly front-end. Uh and now I'm coding all over the stack, but also I'm not writing any code anymore. Uh, neither is anyone in the team writing any code, really. Uh, we're just looking over code, testing stuff, iterating. And uh the amount of PRs that we have, which we haven't even shipped because we just haven't had time to look over the work, and other things have become prioritized is crazy. Um, so I'm wondering, like, yeah, you can you can build much faster, but are you actually shipping much faster as well? I I'm sure we are shipping to some degree faster for sure, but it's hard to know how much when we have a backlog of 65 PRs, right? In in GitHub to look over. Some which have gone stale, some which we've like left over. There's a lot of junk in there. But this is the thing. You end up producing a lot of junk because we even have an internal tool that can produce PRs in the cloud, and uh we just set it off on an idea, so we've got all these things, but then you're gonna never have time to actually get to it and actually take this on board because you can't just merge these PRs into production. Uh most of it's gonna be slop and you need to go through it and get it good, etc. So yeah, I I I think we're moving faster, uh, but I don't know how much I can't quantify

The Hidden Cost Of AI PRs

SPEAKER_00 16:32

it. Um, we are moving faster. I I want to say we are, but I don't know if it's two times, three times, ten times.

SPEAKER_02 16:39

I feel you 100%. I've my understanding is the cost of creation of code has gone almost down to zero. At the same time, the cost of maintenance hasn't been been reduced. On the contrary, I think it's increasing because you're creating more stuff that you're leaving behind.

SPEAKER_01 16:56

Yeah.

SPEAKER_02 16:57

You're creating stuff that you're just overseeing but not actually really inspecting it or scrutinizing as well as you should. And AI is really good at certain kind of tasks, but not at removing the technical depth, for instance. And so you kind of accumulate stuff that doesn't really make sense there. And because the context is limited, sometimes it just like cuts from here, this 18% or this 15% that was left out of the context window matters. Yeah, and somehow you forget about it. And so the accumulation of PRs, some of them go stale, there's more conflicts. Yeah, you know, oh, you deploy a PR that is two weeks old. Yeah, maybe it just carries over some other stuff that shouldn't have been there. I'm seeing lots of stuff. Uh, also after 15 years of not coding, uh, I've gone back to coding this year, and and so I'm seeing all of these things, and I'm worried because I am technical. I'm a software developer by trade. And even though I haven't been developing for 15 15 years, a lot of people who are producing more critical code haven't been a software developer or don't have the this kind of critical eye, right? And so the the slump is just generating at an alarming rate.

SPEAKER_00 18:14

That that's why you need to have a harness for for everything you put out. And this is something that you have to develop, and that takes time. And by the way, that takes away velocity from production because a lot of well, not my job as such, but for example, one person in the company, Martin, he he he has undertaken most of this work in terms of like how can we ship better with more confidence, etc. And that's that takes time to build up that capability, and that takes away from from shipping. That's why I'm saying I don't know our real velocity because we're also building stuff so that we can go faster, etc. So, but that's obviously super important. You talk about um code quality and and um you know consistent um so that everyone is coding in a consistent manner. It's like we're not inventing new APIs to solve problems. There's stuff in the code that can be reused. That's really important because it will make AI faster and you'll have your ship with less bugs. And of course, there may be one day where we're just not looking at code anymore at all, but I think it will make it harder for AI to reason with the code if it's not well organized, let's say. So yeah, yeah. I mean, all this stuff is work in progress. This this year's just been absolutely just bonkers, like where we've gone from. It's like I was I I think Claude Code kind of changed the whole game where we we were using AI in 2025, but we were very much in cursor, looking at every change that it does, pressing accept, blah, blah. Uh, and and and then it moved to like, well, we're just in a terminal and it just does its thing, and then we'll review it later, kind of thing. So we got to that level of trust with AI, and since then, no one's coding, basically.

Taste, Craft, And Lovable Deck Tells

SPEAKER_02 20:12

We uh which brings me to my next point, which was gonna be one of the central parts of the conversation, which is the the subject of taste, right? You posted publicly some opinions on this. I I wrote an article about like when whether taste will die or not, right? Because with the eruption of AI and uh we've got so much slob going on, everybody's a creator, everybody's a coder. But I'd like to think maybe I'm a romantic, and I'd like to think that the people who are really good at their craft, they will still retain it precisely because that's their mode, right? Um I'd like to think that we regret it.

SPEAKER_00 20:51

I think the ones that are really good, there will still be a demand.

SPEAKER_02 20:55

I mean, it levels the field of play for pretty much everyone, but not for the top person uh percentile, right? So um for me, one of the one of the key questions of taste, for instance, is not only like does it look good aesthetically, is it does it really make sense? The priorities, uh how you prioritize the the roadmap, for instance, right? How do you how do you let go of certain parts of the project or how do you shut down site projects and stuff like that? So this concept of taste at large, something that becomes much more important. And you and in type form now with float and super cat, uh, you have been always super opinionated about the software that you design and produce, right? Um, which brings me to think that also you you create this this other mode that you've got is the opinionated software creates strives, people who are willing to go the extra mile for you. And when I see the comments on Twitter and all the social media, like, oh uh forget about Loom. I just cancel my subscription. SuperCAD is the shit right now. Yeah, it's really good. Super fans. Yeah, yeah. How do you find them? How do you keep them in?

SPEAKER_00 22:03

I think they're touching on something where they can see that the a team has crafted a product and has worked really hard on it. And I think you often come across like a website which may be well designed, but you you know there's a few telltales that like it was kind of one-shotted.

SPEAKER_01 22:25

Yeah.

SPEAKER_00 22:26

And kind of you it kind of loses its soul a bit. I think there's an interesting thing with like lovable going on right now, which is they've got this new thing which allows you to well, you you can you've always been able to one-shot design, yeah. But now they've given it some pretty like decent design chops. Um and yeah, this starts blurring the lines, but then can you inherently feel when a product has been like handed over to AI as far as like the decisions, the taste? This is what I don't know, this is a question by the way. Turing too much. Because I think that makes a difference, right? You can feel it when you use something where it's like humans have been involved versus this is more like just very much one-shotted. I I give you an example, and this is pre-Lovable's new design update, but when we've been hiring people, we've been receiving, instead of receiving decks, we give them a case study. It's always a lovable deck, and it always looks the same, the the same way. Yeah, it's like they they went to Lovable, they said, Hey, I want to make like an animated deck for supercut.ai, and it just creates this like, well, what we call AI slop, right?

SPEAKER_02 23:41

Yes.

SPEAKER_00 23:42

Um, it's immediately identifiable, and you immediately like lose like the tables, the arrows. Yeah, yeah. There's just all these tells. So you immediately lose that kind of connection with that person, right? As opposed to if that person really like even worked with the AI to get something. It's not so much like the using AI, it's just one-shotting with AI. Because I use AI. The point is that I'm I look at it, I iterate, and I get it to how I think it needs to be. And that is the human element. But when you get something that that is devoid of that, that creates a complete disconnection. So when you take that to the product level, um, when you get a feeling that there's a this is just completely automated behind, you just feel less personally. I feel less invested in that product versus like, hey, there's a cool team of people that are thinking about this space and doing this, and they, you know, clearly everyone is using AI, but they're really iterating with it and like putting all the magic and details and doing different things because AI always like refers to like the same patterns, right? It's just it just copies, right? What's come before. So there's no space for like, hey, the little surprise, oh, that feels different to how it's usually done, etc. And that comes about when a human with their own personal quirks and taste work on something and something gets crafted, right? That's the craft. So is the craft going away? That's the question. Um, or is AI gonna be able to, let's say, quote unquote, craft and get to the point where we're just like, wow, it's like this AI is so amazing and creative that I don't care anymore that there's a human behind it. It's just beautiful. I don't know. I don't have answers to those questions. I really hope. Like I'm one of these people that like I'm I'm embracing AI because I have to, but I I go back and I, you know, I I miss the days when things were hard to do and you had to like put the work in, right? It was kind of nice and visceral, and now it's just everything's just too easy and boom, and like anyone can do anything, and the bar, like everyone can jump in and think they can design something, etc. And so it's a tough sorry, it's a tough, quickly evolving space, which I I I just don't know where it's going as far as design. It's a it's a hard one to put my finger on.

SPEAKER_02 26:09

I I think it's sort of like uh I I don't know if you're familiar with the concept of chronocentrism, which we tend to believe that things were better when we were young, right? Because they were of course, right? No, but that's the thing. Like you used to like uh things that required a certain level of effort and you did it and you felt accomplished, then they don't make any sense anymore, right? Why would you want to learn how to drive manual nowadays when actually it doesn't really have any good this is what made life rich.

SPEAKER_00 26:44

Now I'm gonna sound like another guy. In a way it does. We used to call each other on the phone. Now it's like, oh, it's weird. Why are you calling me? It's weird. It's pushing it.

SPEAKER_02 26:52

Like not my generation, but feels like uh no, but I understand the rejection that that you you you that you feel when you encounter these telltale elements in in you know websites, for instance. I can see them. Like I produce lots of websites for internal projects, sites, you know, things that I use internally for me, like my own accounting uh app. It's something I produce with the app. It's only for me. Yeah, but I see like all of these, like for instance, the arrows that go like kind of like a corner, like this, that I use in many, many websites. The eyebrows on every section, you know, the eyebrows with the little thing on top of the super thin lines that are the borders of the tables, yeah, yeah. Things like that.

SPEAKER_00 27:33

It's you know, it's not like the M-Dax. Everything looks, yeah, exactly.

SPEAKER_02 27:36

But 20 years ago, we had I faced the same problem, the same situation with WordPress templates. Yeah, every fucking website was a fucking WordPress template. And for me, it was like, why? You know, I I I sort of resorted back to when I was designing a creating website when I was 15. Everything I had to do it manually and all of that. And so, in a way, for me, it felt like a it was part of my ego talking and saying, like, yeah, everybody else shouldn't be able to create websites, which is not right. I know the right thing is like everybody else.

SPEAKER_00 28:16

Correct good designers' work, right? It's a different thing. Now, now it's it's like anyone will be able to create a uh a beautifully animated, like super show-off y site, like one-shotted. Yeah, and it will look great. The question is, will will the perception still be there that this is AI or not?

SPEAKER_02 28:43

I think so.

SPEAKER_00 28:44

But up to which point? You take it five years into the future, take video, everything, like yeah, where's it all going, right?

Marketing Without Selling Your Soul

SPEAKER_00 28:51

I don't know. When the point and it will be better than any human designer can produce, it'll be so so good and so surprising and so creative that but I think there are several other vectors that you should factor in.

SPEAKER_02 29:05

Like, for instance, the the timing, right? Timing a product launch or a communication, yeah, is something that maybe uh will make you send out, right? You can craft the most beautiful design using AI, perhaps. Yeah, yeah, but if you launch it, nobody cares. Yeah, like there should be something else, like a rising uh tight lifts all boats. Right.

SPEAKER_00 29:27

I think we need to care that there's people behind it that care about the mission and uh uh passion about it.

SPEAKER_02 29:32

Yeah, you can place your fucks somewhere else, right? Um instead of investing so much in, oh, I have to create a website, I have to invest in design. Maybe like the real core value of the product isn't there. So let AI handle that. Focus on execution distribution, uh PR, and I don't know, surprising people. I think that that's my take.

SPEAKER_00 29:53

Yeah, I think that's the game now, right? Correct. Not for everyone. It's like you can you can be building as much as you want, but if you can't market it, like how are you gonna get out there? But there's a big challenge with this, which is to what to what level are you willing to stoop down in order to get noticed? Because and then taste comes into this. Like maybe someone else will be be able to copy exactly what you're doing, yeah, make it look relatively good, but they'll just market the shit out of it and be super aggressive and say really aggressive things, which probably are shown to like convert more people, but it's the vowed of this kind of a special, let's say this is a special product with a sp with it with special people behind it, right? So it's like where do you I don't even know where you start there because maybe you're just at a loss. Immediately you're at a disadvantage if you want to be more tasteful, let's say. I'm not talking about design, I'm talking about your approach to how you go out there. There's uh because there's so much noise, like to get notice, you just have to you've got to start getting comfortable doing things which you're not comfortable with with doing, uh which you haven't been comfortable doing before, right? And that's that's a challenge because there's so many people in the market, right?

SPEAKER_02 31:23

Exactly. There's another element to it is longevity. Lots of more companies are being created, but at the same time, the lifespan of companies is shorter than ever, right? So I think that being there for the longest time, it's also kind of a maybe not a mode, but if it's a differentiation value, right? So um a company that has been there for 10 years, it earns immediately more trust for me uh than a company that launched last week. And um just because everybody else can create a company, but sustaining a company after you know certain VC cycles or uh global geopolitical things hasn't changed.

SPEAKER_00 32:08

Okay.

SPEAKER_02 32:10

What's your take on that?

SPEAKER_00 32:10

Like I mean, you got more competitors, but I I think it's still a slug, right? Yeah, we're all also because we're seeing some of these AI companies just like take off so quickly. Yeah, and but those those those are rare, and everyone's trying to replicate that. I think for most companies, it's like a it's it's a slug, right? I mean Supercut's doing really well, and we'll hit a million in revenue in our first year, and that used to be like like or more, hopefully. And that used to be like, wow, the unicorn path, right? But now that's like that's yeah, that's a standard. Yeah, yeah. It's like like nothing compared to like, hey, we need to build a an AI company, and then we'll just just take off. But those that really take off, I don't know how you know, out of each 10 that you see, or every, well, each 10 companies you see launched every day on Twitter because there's so many, yeah, how many of those are gonna either die a quick death or just like keep on going and maybe hit it lucky later? And maybe there's one a week that actually does might do something, who knows? But it's just it's so easy to build and launch right now that you can do all this smoke and mirror stuff. You promise all this stuff that your AI can do, or sorry, that your product can do. It's all obviously AI driven, but then you try the product and it's never up to standard. How many times has this happened to you recently where you get hyped into a product and then you try it and it's like, yeah, I didn't get that result with it. Many times, this is everyone's experience, right? 100% because everyone's playing this game of hype, yeah. And to the point of like, are you willing to join in that hype game? That's that's everybody else because I don't want to I like I don't want to sound magnanimous or anything like that, but I just don't feel that comfortable just like lying about it and like bigging myself, bigging up myself and my company up, which many people are just completely comfortable with doing. Uh I I'd rather just build that value and people try the product and not be disappointed, right? And that's building it up, you know. Then you can make bigger promises later when your product is up to standard. But early on, everyone just wants that quick kind of take off. So they're they make this flashy video where it creates the most amazing thing. Like you see these website and app builders, which can create these most amazing things, and then you try it and it's like, well, it doesn't look like that. Like my video looks like shit, right? But of course, all these things are carefully manicured that if you really spend a lot of time with these things and you know what you're doing, you might get a result, like the result we're putting out.

SPEAKER_02 34:49

It's never so magical. Like most AI products don't go past their the the test of the demo, right? For for a small demo, yeah, it works. For production ready data projects, yeah, they don't work. And now I'm I'm speaking as an angel investor. I have not invested in any, I've only invested in one AI company uh in the last three years because I know that all of them are about customer acquisition, but they're churning people left, right, and center. Um they are not good at converting, they're not good at retention, yeah, and they're just burning cash and hoping for a Hail Mary solution that maybe you know cycle, right?

SPEAKER_00 35:36

That's a hype cycle. So you hype them into trying a product, correct? They sign up, most of them will just drop off because they don't get the value. Correct. And then um some of them will get through, they'll get some value, and then they'll be off onto the next product doing the same thing.

SPEAKER_02 35:51

So it's and at the same time, I just hope that the AI as an accelerator, like it is an accelerator, it just accelerates also the demise of a lot of unneeded companies. I'm not, you know, wishing the the the the misfortunes uh to anybody, but at the same time, we want these people to focus on solving real problems, really real companies, not companies full of fluff or hype or snake or charmers, right? Um, so for me, uh as an Intel investor, I've seen a lot of my portfolio companies lingering really long deaths. I really don't want that. Especially in Spain, it's particularly painful. They tend to be in zombie state for many, many years, especially if they go after down rounds, rich rounds, extension rounds, and all of that. They just they they refuse to die, right? And so I don't know, maybe with AI is like, ah, just move on to the next thing, you know?

SPEAKER_00 36:44

Yeah, uh that's a pattern I'm

Multi-Product Strategy And Focus Tradeoffs

SPEAKER_00 36:46

seeing. Companies doing multiple products now, not just small companies, but big companies, just say we've got to become a multi-product company. Because they have to. It's it's also so easy now. You can just put a small crack team from the company and and and work on stuff and see what sticks.

SPEAKER_02 37:02

One of my takes on this, particularly, is that one of the strongest modes you can have nowadays as a company is being multi-product. Much like, you know, perhaps for less tech savvy people, Apple is really good at this. Once you buy the iPhone, you're like, I'm gonna buy the Mac because and then the iPad, and then you get Wi-Fi on this.

SPEAKER_00 37:21

And there's like physical products that works really well, obviously.

SPEAKER_02 37:24

Can't stop. Yeah, but also in in software, certain software work together. For instance, we use uh a couple of software products that I use uh that are called Harvest, it's for invoicing and forecast, which is for uh uh project planning, right? One without the other really don't make sense, but one of them is not really good, but we can't ditch it that easily because it really depends on the other, right? So, and you're probably familiar with a company that's called Every, yeah, right. And uh their products are not magnificent, but once you get into their subscription that allows you to use all of their products, you're kind of like they're in the ecosystem, yeah, yeah, and you're kind of like trapped because you get like free value because if they launch a new product, it's already bundled in, right? So I think that going multi-product, and I've spoken this with with Amir from Todois, for instance, it was like, yeah, we're going back into multi-product precisely because if we trap somebody into using to do is then we're gonna have somebody else. So potentially a new customer for our second product, our third product. So is that your strategy with Float? Or were you what's your take on this?

SPEAKER_00 38:27

I think the challenge that we have compared to maybe Amir is that they have a company of maybe 30, 40 engineers that were working on the main product, and they figured out well, we don't need that many engineers on that. So what do we do? Either we fire them or we see if we can build new new value with that. So that makes total sense. I think that the the trap that we're in at the moment is that Float Labs was set out uh to be a multiple pro uh multi-product company, but given the complexity of SuperCar and the size of the team, we've mostly only really been able to focus on that. We've done a few like B2C well consumer products on the side that just experiments. Like I did a a photo app called YapCam over Christmas, and Neil is like launching Sublime, which is a captions creator, you just record and it creates pack captions and you just publish it very, very fast. Um but these kind of projects are super scoped and they're easy to do. It's like they're kind of side projects. But taking on like a bigger kind of project the size of Supercut, right now we just Supercut just sucks us in the whole time. So it was the strategy, but we haven't been able to free ourselves out of Supercut to really be able to do it and think like bigger ideas. Hopefully that's coming soon, but you know, we've got a good opportunity with Supercut to really break through, so we're focusing there.

SPEAKER_02 40:02

Because I I understand that you might want to go into multi-product at the same time, it just takes away focus, right? And uh people, not resources, but like money, time, people, mental bandwidth as well. Um, if you got a main product that is going like this, that's going like a rocket. Um how do you calibrate when is the right time to say, like, fuck it? We're really going to devote one or two people on something else at the expense of SuperCat, but we really need in our case, it's really the founders that have to drive it, right?

SPEAKER_00 40:36

Okay. It's not just well, I mean, there could be a situation where someone in the team comes up with an idea and we're like, yeah, go off and do it, right? Uh but so far, again, we've just all been sucked into Supercut. But I think there has to be a day where where actually this is why float is called float, because the idea is that that Neil and I would start these products, build a team around it, and then get to a point where it can where it can float by itself, and then we would uh work on the next thing. But uh Supercut needs a lot more more work, we feel, before it could just be floated.

Letting AI Prioritize Founder Work

SPEAKER_00 41:13

Um, and that's you know, keeping that founder energy in the product. Uh it's a product that needs selling, right? I need to talk to people, I need to go out there as a founder and really sell it um directly to companies, right? Getting a team of 500 people off Loom is not easy. You have to go there and be a founder, go there, talk with the people, push, etc. But eventually you can do that. And that's that's what we have to do now.

SPEAKER_02 41:40

And but but that certainly doesn't scale. And um cannot to wrap things up, one of the main changes I'm implementing right now as a CEO and founder is I used to treat AI as another employee, right? As an army of an inf an infinite army of interns, right? And now I'm I'm I'm I'm flipping the situation and I'm letting the AI manage me. Like I'm giving it the ideas, I'm telling it, you prioritize, you execute, but uh you tell me what do I have to focus on? Where does everything go, which project takes priority right now, and stuff like that, instead of me making those goals, because on a bad day, I would make the wrong decisions. What do you use to do that? Okay. But I just built a system that um I brain dump anything I want into it because like I'm the kind of person that I don't know if you relate to this, but as a CEO, we're usually go-getters, right? We're type A achievers, and we want to chase the next shiny thing. And sometimes I'm like, ah, that's an easy win. You know, I'm gonna implement llm.txt for the website. I'm just gonna go there. It's just it's five minutes, but those five minutes could have been invested in something else more meaningful or something I have forgotten about that I should have done today, but instead of chasing this other thing, right? So right now I'm triaging all of these things using Cloud. And the AI tells me, now you this idea is is okay-ish, you can do it in the summer, or you can do it next week. Oh no, this one is really like just do it now, yeah, right, and stuff like that. And it's pretty useful. I'm still distilling it, and it's not very, very um fine-tuned, but I think I changed the mental model, and now I don't feel like I have to do everything and that I I could do everything. And it's it helps also to unload everything that you've got, all of the ideas that you're probably you're an idea person, you know this. And so you probably generate more ideas than you can actually produce. And so by unloading them, what's that?

SPEAKER_00 43:41

Not the moment.

SPEAKER_02 43:41

Not in the moment. By unloading them into the system, at least I'm just freeing up the mental space for other stuff. And like, hey, if the AI decides that I should not do it, I will trust it because it's more intelligent than me. Sometimes I'll go into funder mode and I'll fucking do it, right? But uh I don't know if you if you feel like this. Something that you know you're seeing it. I'm I didn't invent this, I just got it from a podcast.

SPEAKER_00 44:04

So yeah, but so just to kind of feed that back to back to you. You're feeling that you need to use AI to be able to organize and prioritize your work, otherwise you're gonna be working too much, or I'm just gonna be working on the stuff that really don't have the most immediate value now.

SPEAKER_02 44:24

Yeah, right?

SPEAKER_00 44:25

I I can tell, but sometimes you know when you're feeling triaging of priorities, yeah.

SPEAKER_02 44:29

Yeah, sometimes you know you haven't slept well and you're like, I'm just gonna be doing like quick wins today. Yeah, no, you have forgotten about this stuff that the deadline is today, you just forgot about it. Yeah, it's that kind of a system.

SPEAKER_00 44:40

Yeah, I I don't work like that. Yeah, I can see how many people do. I just uh I'm I'm extremely impulsive, and and so I'm I'm I'm driven by responsibilities mainly, but also a lot of impulses. So if I want to get something done, I'm just gonna get it done and I'm gonna find the time. So and I I I love working, so it doesn't matter. Yeah, me too, but I disappoint a lot of people. It's not like yeah, so I'm not trying to like, hey, I need to save my weekend or something like that. I always find the time if it's a late night or yeah, you know, uh very early mornings, like I just I you know it's not a stress, right? Yeah, but maybe if you've just got if you're if you if you are managing multiple different areas of concern, not just like multiple things inside your own product, right? In my case, supercar. But for multi-products, perhaps. For multi-products, maybe you're gonna, yeah, have a better way to triage like where you focus your time. Absolutely.

SPEAKER_02 45:40

Maybe you can calculate for you like the 20% allocation to the site.

SPEAKER_00 45:44

Maybe you're accepting too many things. I don't know. I mean, everyone's gonna have their own world in that area, I think. And but yeah, AI can definitely offload some tedious tasks, right, that you don't want to do. I I've got a bunch of stuff which I like uh I don't know. Uh dependabot uh fixes, right? Where you have to update the repos that you're working on, like going in there, creating a PR and getting that all merged, that's those things can completely be automated now. So there's that. There's things like you know, I started playing with um ingesting all our sprint work, all our PRs, all our lunar tickets, every async stand-up that we do every day to give us a um to give us a halfway point during the week where we can get back to see how we're doing, and at the end of the week, um you get the full report and see what you felt. Actually, Kath from my team set it up and then I've been iterating on it. But basically, like all these things that you can kind of like automate with AI to get insights are super, super valuable, I think, and save you time, right? Because otherwise you have to go through, look at the data, see what went wrong, maybe have a conversation, you do a sprint review, etc. In this way, like it's all like given to you up up front. So so yeah, those those things. I don't know if I answered your question properly, but yeah, okay.

SPEAKER_02 47:20

You did David. One last minute, uh one last question.

How To Help Supercut

SPEAKER_02 47:23

Uh, what can we do for you? How can we repay you the favor? How can we help SuperCat or David?

SPEAKER_00 47:28

Yeah, yeah. Just uh use SuperCat, use it in your team. Cancel Loom. Well, well, can Loom's trying to charge you a lot now, so I think what they've done is they're just going after a lot of enterprise customers, well, or bigger teams that don't mind the price increase, and you know, smaller teams are kind of suffering in that, or mid to small teams. Um, and so if you want to work with a if you want to use a product where there's a small team behind it which you can have access to and you know, hopefully doing innovative things, then I think Supercut's a good choice.

SPEAKER_02 48:02

But uh I'm seeing a lot of parallelisms with what you the type form, and I know this is going to be a very successful product. So thank you.

SPEAKER_00 48:09

Thank you very much.