Cosmos raises $15M Series A to build a creative-first visual search engine and inspiration-driven commerce platform

sounds like we have Andy in the reream waiting room. So, let's kick off our Lambda Lightning round and let him in to the TBPN Ultra Dome.

What's going on?

How are you doing?

What's up, guys? How are you?

Good to see you.

Welcome to the show. Uh, first time on the show.

Wearing green, too. A great sign of respect.

We love greens racing jacket. You guys had a green hyperard.

Yeah. We were doing a quick tour of

uh anyway, first time on the show. Please kick us off with an introduction on yourself and the company.

Yeah, for sure. Um, I'm I'm Andy McHune. Uh, we're based here in New York. Cool.

Uh, 20 person team. Um, building building the new home for inspiration on the internet. Um, visual search engine. Uh, just announced Star Series A today. Super exciting. Uh, $15 million round. Um, and sliding around.

Amazing.

I'm fired up. Congratulations. Yeah, there's so so much to talk about right now and it's such an interesting time to be building this company because

there's so there's just like an explosion of content. You get a pretty picture now. You don't need a you don't need a camera.

Anybody can generate one from their own home. Uh why don't maybe we start and you can give us kind of the history of the company, what the initial

maybe even go back further pre uh about the previous company you worked on.

Yeah, for sure. Um yeah, in 2017 I I uh founded a company called Unfold. Um so we are like a mobile content creation tool kind of Canva adjacent. Um ended up selling that company to Squarespace. Did a few years there. Uh leading creator product strategy and I've been working on Cosmos for the last four years or so now. Um and yeah, I'm I'm a really creative person. always working on a lot of different creative projects and I very much grew up on Tumblr back in the day and ever since then I've been an obsessive collector and curator of things on the internet um always just saving things to a million folders on my desktop and folders on my phone things just in all of my different bookmarking folders on different platforms and um so Cosmos just started out as me wanting to build a place where I could bring all of these references into one home um where everything would be tagged by AI so I could quickly search and resurface anything that I had saved. I wanted to be able to work with collaborators in these different projects. Um, so yeah, I mean just started out as as building a side project for myself as everything I've ever built has kind of started out as um started getting into the hands of my most creative friends and people were like this is awesome. Um,

and so were those friends the ones that were uploading the original content? Were you scraping the internet to bring on the first and like seed the community? Like where was the actual content coming from? Yeah, we we weren't scraping anything. Um, it's always come from the users. Um, we have great importing tools and people can upload links from a lot of different platforms and around the pl around the internet.

Um, so yeah, all of the content is is seated by the community. Um, which has grown in a really really beautiful way. We have tens of millions of images on platform now. Um, an amazing global search, you know, all of the recommendation layer that you would expect from a product.

And what were the early adopters looking for? Or is it more like because I I you know when I think about image search I think about people that are doing home renovations but also designers that are looking for inspiration for clothing or web design. There's some people that just uh sit there and scroll just to enjoy their favorite images like you enjoy your favorite books. Um what were the early drivers of growth?

Yeah, I think you know we we amassed a weight list of a few hundred thousand folks before we launched the product. And I think if you look at the cohorts of users that were coming in really early, it was, you know, designers across every discipline. We're talking interior designers, architects, graphic designers, brand designers, photographers, folks like this. Um, which really helped seed such an amazing, you know, original corpus of content, but really where we're seeing a lot of the growth come from now is sort of this, you know, Gen Z wave. Um, you know, I call them kind of the next generation of Tumblr kid. Um, you know, they they they love to mood board for fun. you know, maybe they're not saving typographical references on a day-to-day basis because, you know, they need it for their client work, but maybe they're just dreaming up um a version of themselves that they aspire to be um and they want to surround themselves with beautiful things.

Yeah. And then how is how have you been grappling with the AI generation, image generation boom for you started the company four years ago, pretty much pre-I content essentially. Uh now it's indistinguishable. uh what is what what's the journey been like

and there's so much pressure on each platform to take a stance right now when you look at most of the platforms the stance is like AI content is welcome here we may get tagged

yeah we may try to tag it

uh but it's being harder and harder

for a lot of platforms you just have to search okay I only want to see images before 2022 [laughter] and and it's like almost [clears throat] like an ar like an archive that can be really good uh but yeah just how have you been processing Yeah, for sure. I think that, you know, we're definitely seeing a lot of backlash um within the ecosystem against a lot of other platforms where AI generated content is kind of proliferating all of the feeds and um you know people are engaging with stuff that they don't even know is AI generated content and maybe they search the thing and then they click into it and they see that it's labeled by AI and that they find that really frustrating. Yeah.

Um

and you know, we're definitely we're definitely growing a a lot of users now that we kind of consider like refugees from this. um you know, we really want to be a platform that champions um human creativity and human created content. Now, that's not to say that we're drawing a hard stance against AI generated content, but I think that it's really important that people are able to decide what they want to see. Um and so we've done, you know, what some of these other platforms have have done as well is every image that gets uploaded on platform um we're running it through, you know, multiple different algorithms to detect if that is AI generated. And

so every user has settings within their profile where they can um choose to show or hide AI generated content.

Um

yeah,

interesting. And then

how how confident are you that you're catching all of it?

Yeah, I mean it's it's an uphill battle, right? I mean it keeps getting better. You have to keep retraining the models.

Um and it's only going to get more difficult. Yeah.

Um, so you know, I I think that it's only a matter of time before it's, you know, pretty difficult to do.

There's also some AI image models I've noticed where I will upload an image, ask it to do something, and it kind of just spits out the exact same image, or maybe it just like adds a piece of text, and it feels like it didn't actually run a new diffusion model. It just sort of use some tool to like layer over text and it's like, well, it's kind of just acting as an agentic Photoshop at that point. not really creating an AI image like all the details and fingers are the same because it's literally the same pixels that it's just using and so to that count yeah it's going to be all hairy. Um I'm interested to know how you think about for you pages algorithmic feeds versus search and being a like a more of a search engine. Is that line blurring? What are the important trade-offs? What do you like and dislike about algorithm design?

Yeah, for sure. So we have a for you page on on Cosmos that you know learns your taste and learns your interest as you say things. I think it's important that you have something like that. So people have a new vector that they can discover content through. Um you know what's really interesting is the search problem is actually very similar to the for you problem especially when you get into the world of serving people personalized search right so you train your for you. you know, we have an idea of what your tastes and and your aesthetic leanings are and then when you search something, you're actually just adding that keyword over sort of this vectorized database of um you know visual aesthetic that you've already identified with. So the problems are actually very related um and intertwined. I think one thing that we've done that's really interesting is we've started um developing in-house aesthetic prediction models. Um, so we've actually been using all of our own data, user data that, you know, user user uploaded content. And we've been training models that understand kind of like what good versus bad looks like. Um, and we're using we're using these aesthetic predictions to loosely rerank our search results in our for you pages to try to keep, you know, the slop off of the platform. Um, you know, we believe

slop wars.

Yeah, you're deep in the slop wars. Frontline soldier here in the foxhole. [laughter] I salute your your your service. Uh

I think like we believe that your outputs are a reflection of your inputs and so it's like very philosophically important that we're we're feeding people high quality inputs because you know so many creative projects start with the mood boarding phase. Um and so if the inputs are going to become the output, it's really important that you know we're we're giving people a place where they can find the highest quality things.

What's been the growth loop for you? What's been the key to growth? I know that you've been uh you know featured by Apple as one of the top 25 apps for 2025. You're 500,000 followers on Instagram, but uh is there a traditional like referral flow? What what are you using from like the web 2.0 boom era versus so viral social playbook? What's been the big growth engine? Have you had to just run ads? Like what what's worked?

Yeah, we haven't been running ads. I mean, it's really been or organic up until this point. Yeah. Um, you know, we had an early like invite friend flow which was pretty successful for us. I think that that's not as successful as it was a few years ago when we ran it.

Um, but I think, you know, the nature of Cosmos is really collaborative. Um, you know, you you start what we call a cluster is like a collection. You start a you start a cluster, you start inviting, you know, your collaborators to those. Um, the experience gets so much more valuable. Um, and we're seeing really high density as well within some of the best creative teams in the world. I mean we have you know hundreds of users on platform on nike.com emails. We have you know teams at Chanel using Cosmos, teams at Apple using Cosmos. So that density within creative teams has been really really powerful for us.

And then monetization I feel like you're going to be anti- ad just because as soon as you have an ad, it's hard to influence like okay you can't say like well you can only serve pretty ads. $50,000 you can post the sloppiest AI slop with six fingers, but you're we're going to charge you a pretty penny. No, no, I want to know the real uh monetization strategy.

Yeah, for sure. Um, so we're not monetizing yet. I mean, we have a premium subscription right now, but it's pretty demoted within the product.

Sure.

Um, I think, you know, there there will be a lot that that we have yet to reveal. Um, but I think that where we're going is, you know, going deeper on sort of focusing on those those creative teams that are using Cosmos and how can we better serve those folks. Um, and then we're also starting to to build into commerce as well. So, we launched a commerce pilot um last year and you know, you can imagine, you know, we're running hundreds of thousands of searches a day right now on the platform and if someone searches for brown leather couch, they're probably looking to buy a brown leather couch. And so, how can we

Yeah.

Yeah. How how can we create a compelling commerce experience that doesn't feel like we're trying to force products down your throat, but it's really commerce through through inspiration.

Yeah. Yeah. It's like aesthetic first, taste first, as opposed to like feature, set, price, like all the filters that you see when you're on Google shopping. It's a like those are probably going to be way lower in the funnel than the aesthetic that you're

Yeah. shoppable mood board. You can think of people doing inspiration for, you know, interior design, their wardrobe, all this stuff. It's like that's that's a high intent moment when somebody's trying to

um

Yeah. Can you can you explain a little bit more about the aesthetic trends that are popular? I I I've seen these different uh like like phrases used to describe how certain people describe their whole aesthetic like what what is uh uh how would you describe the communities that have formed around these like modern subniche aesthetics?

Yeah, it's a it's a really interesting question. I mean, we we launched a new feature today. Um, so we gave the the profile a really big refresh. Historically, everything that you saved had to go within a cluster. Um, the the release that we put out today gives you a core feed of images that sit at the forefront of the profile, which is really akin to like your Tumblr blog back in the day. It's really like, you know, a snapshot of your taste and and what you're into. Um, we're starting to go really deep on, you know, the datas and the trends um that we're seeing on the platform between, you know, what people are searching, what clusters are they creating, all of those sorts of things. And we're going to start putting out trend reports, which we're really excited for. And we've started sort of seeding these into our newsletter, which has been really, really fun.

That's great.

Um, but yeah, it's been really cool to see these different sort of niche aesthetics blossom. Um, I think historically a lot of the usage has been just around again the core use cases that we were talking about like graphic design, interior design, whatever. Um, but if you think about something like Tumblr, right, there was like fandoms that were popping off and, you know, like emo aesthetics and culture, right? So, we're starting to see some of these pop up, but it's a little early to tell. Um, but we're really excited to start publishing these more publicly.

Okay, last question. For your personal aesthetic, the panone color of the year for 2026 is cloud dancer. It's a soft, lofty white that symbolizes calm, peace, and fresh starts. Underrated or overrated? You know, I'm I'm a neutral palette guy, but I know it was really uh you know, contentious within uh within the creative community. I think we could have gone a little bit more creative with it, but uh I'm never going to hate on a neutral palette.

Okay,

this is pretty out of this is pretty out of pocket for

Okay, so it sounds like Cloud Dancer might be working its way into your

It's a good name.

Spooky in the X chat says, "How do you avoid creating an automatic house aesthetic if you're algorithmically filtering slop?" Uh any thoughts there? Like is this part is part of this like it is so generated that somebody might have like you're not trying to like have an overarching aesthetic other than

like somebody could come on and create a bunch of clusters for an aesthetic that that is not aligned with your personal taste or cosmos.

The art station look became the midjourney look and then now that just looks like the AI art look. But, you know, what if you're actually just into like, you know, matte painting? Like, what you're going to get filtered maybe. I don't know.

Yeah. I think like the point of the aesthetic prediction model is is less to

um pedal what we think is good and is more to set a bottom bound on what we think is bad. And so, that bar is set pretty low.

Um, but it's it's a great question. You know, we're continuing to retrain the aesthetic the aesthetic models that we have. Um, and we're kind of going through this right now of where we're trying to widen the aperture of that um to service more folks and and and and to reach a broader audience. Um, so I don't think there's a clean answer to it other than just like, you know, continuing to iterate on your models and pass them into production and and measure and and learn from there.

Very cool. Well, thank you so much for taking the time to come hop on the show. Great to meet you and good luck. Congratulations.

Great to see you.

We'll talk to you soon.

Goodbye.

The New York Stock Exchange. Want to change the world? raise capital at the New York Stock Exchange. And with that, we will Oh, you can hear I hope you can hear it.

I don't know if it's coming through.

The 21C is fired up. That V8 is roaring. Uh, and we have Michael from FatFit Fun and Deep 33 in the [music] Reream waiting room. We'll bring him in the TBN Ultra Double. How are you doing?

Welcome to the show.

Welcome to the show.

Hey guys, doing great. Good to be here.

Thanks so much for hopping on. Uh, give us the news. I want to ring the bell. First of all, this jacket is fantastic.

You have news for us.