Loading summary
Transcript40 lines
- [00:00]
A
Foreign.
- [00:05]
B
I'm Dan Runcy. Welcome to Trapital, where today we're going to talk about AI and music, where frankly, we've heard the standard questions by now. Is AI an opportunity or a threat? Will AI end music? Will AI replace artists? Will it flood streaming with low quality songs? Those questions all matter, but this conversation starts from a different place. What if we assume that the most extreme version of the AI argument is true? What if AI can eventually automate any cognitive task? If that happens, what does it not replace? What then becomes more valuable? And that's where the topic gets really interesting. Today we're joined by Jessica Powell. She is the CEO and co founder of AudioShake, whose company uses AI to separate audio into stems and unlock new possibilities for music, film, tv, sports, live sound, you name it. Jessica has one of the more nuanced views on AI that I've heard, especially from a founder in the space. She's optimistic about what the tools can unlock, but she's not naive about the disruption, the labor questions, or the flood of content that comes from it. And as an investor and audioshake myself, I've had a front row seat to see what a lot of this looks like. In this episode, her and I talk about the physical side of music and why it may become even more valuable in an AI world, especially with live performance instruments, human connection, learning a craft, and the value that we place on things that are inefficient on purpose. We also get into why AI generated may be less of a useful term over time, how digital audio workstations and AI native platforms may converge, and why streaming fraud may be a bigger issue than the creative tools themselves. We discuss that in a whole lot more. But the big question underneath all of this is simple. What if AI makes creation easier, faster and cheaper, but humans look for meaning elsewhere? I hope you enjoy this one as much as we did. Here's my conversation with Jessica Powell. Got some big news for you. The third annual Trapital Summit is back year three. We're running it back in LA on Tuesday, September 15th and early bird tickets are now on sale until July 15th. This summit is for the people that are shaping the future of music, entertainment, media, technology and more. It's for the founders, operators, investors and executives that are driving this business forward. What makes our summit different is simple. We go for depth, not breadth. No overloaded panels, no fillers on stage. This is for the one on one fireside chats with the execs and leaders that you want to hear from. And and it's filled with the room of decision makers to make your networking connections and relationships the most valuable they can be. We couldn't do it this year without the support of our wonderful partners, including Universal Music Group, Warner Music Group and several others. If you saw the highlights from past years and thought, I gotta be there next year, this is literally for you. Links to get your tickets to the Trapital Summit are available now. You can go to trapitol.com summit or tap the link in the show notes. All right, we are here with Audio Shake's Jessica Powell returning guest to Trapital. Welcome back.
- [03:16]
A
Thanks for having me.
- [03:17]
B
Of course. Great to have you as always. You were kind enough to do a guest essay for us in trapital. This is your second guest essay you've done, and this one is intriguing. It's about music coding and why music is going somewhere with the future of AI that may actually give people some peace of mind that, yes, not everything is just going to get disrupted. There still is a human connection to the business. So let's first start with what inspired you to write this.
- [03:45]
A
I was listening to a tech podcast where the host was interviewing the founder of Claude Code Sky Bohr's tourney. And the host was asking him, well, do you miss like, traditional engineering? And he was like, nope. You know, for me, it was always about getting from A to B. And I have superpowers and now I can get from A to B much, much faster. And I was listening to that and part of that describes, I think, how work with and experience AI too. But I also know tons of engineers that really, really miss the journey, right? Like, they miss the process of learning and discovery and figuring something out and the way that that doesn't entirely disappear, but a lot of it does disappear now that you have assistance like Claud Code. And it got me thinking, well, where is meaning created both in the act of creation and in the act of consumption, when cognitive labor essentially becomes zero cost, when it's all automated Again, whether we're talking about coding and engineering or something like music kind of to theoretically to extremes, right? At least in terms of how we talk about them societally, music and the arts on one side and stem on the other side, regardless, like, where do we find meaning in a world in which the driver of all of that activity was largely cognitive?
- [04:53]
B
One of the interesting things about cloud code and how much that has helped so many tech companies become more streamlined is, yes, the end product is still something that the end consumer wants, wants that will be more efficient and probably needs less human engineering. Headcount over time. But in music, there's still a bit of a gap there, where sure, yes, you can press push button creativity, get the output from a song, but if Taylor Swift or Drake was just to say, hey, here is an output of 10 songs that sound like me, boom, put it up on streaming services, I'm not sure that that would connect with consumers. Even if that is an artist at that level, they still need to have some story behind it to make it work. And that was another reminder of, okay, yes, music is still different than code.
- [05:38]
A
Yeah, in its own funny way, I think music in some ways is more durable than more strictly cognitive tasks like coding. Which is not to say that all the engineers are going away. There's still a lot of tasks that are going to require humans. On the engineering side, when I think about music, the aspect of music that is really, really hard to automate is, I'd say things that are tied very much to the physical body, like these corporeal elements, if you will. It's about embodiment and ident. So that could be things like physical performance, live performance. We're very, very far off from automating that. And what quite are we gaining from automating that? Do we want that automated? What exactly is being replaced there other than the human? So I'd say like playing an instrument, having to learn an instrument, seeing people perform, connections between artists and fans, these are all deeply, deeply human activities that even if AI was able to automate all of the cognitive act of music, there is still this physical part of performing for people and connecting with people that I think is incredibly powerful. And when we look at what gets valued, just like historically over time, things that are difficult, things that are scarce, right. When things are abundant, those things tend to be cheaper. When things are harder to obtain, they tend to be more expensive or have more value. And I think that physical act, like physical acts and physical connection will be more and more valued over time. When we talk about the arts and specifically music, so much of music lies really in the body.
- [07:08]
B
And as you were building this out and thinking it through or stress testing it yourself, are there certain aspects of coding or software engineering that also are similar? Where okay, what is the part that still requires human intervention that is still difficult? That for those few people that the cloud code creator had spoken to that still missed it, are there certain parts of that creation process that they do miss or do become more valuable?
- [07:33]
A
I think what we're talking about today and in the near future, across the board, whether we're Talking about engineering tasks or artistic tasks, taste, wisdom, experience are still very, very important. I think that when I talk to folks in the valley on the engineering side or the research side that are getting the most value out of these tools, they actually tend to be the most experienced people because they already have a huge number of frameworks and patterns in their head around how to harness these things because these tools still make a ton of mistakes. Also, there's a lot of things tied to deployment, I would say security, privacy. There's a lot of aspects of taking something from research into production that today I don't think anyone's entirely comfortable with having that be fully automated. But I think if we zoom out and say, what are machines really good at? Machines are really good at detecting patterns, finding anomalies, being consistent and systematic about how problems are approached. A lot of those tasks actually do intersect with taste. You know, we wouldn't necessarily think about them the same way. But machines should be really good theoretically at say, security related tasks. In some ways a shakier moat, right, than when we think of these tasks that are very much about human connection, watching a human wrestle with a problem. Rightly or wrongly, we value difficult things. We value scarce things. There's a reason, silly though it might be, that luxury handbags exist, right? Anyone can go into a Target or the equivalent in their country and buy a bag because that exists. Then people will search for the bag that you can't get at Target. Similarly, plenty of people that can afford the doordash will entirely develop a whole hobby of learning to cook or gardening or whatever it might be, because they want to wrestle with difficulty and they want to touch things. We were given hands and we tend to like to use them.
- [09:21]
B
And I think the nuance with music in this way too is that yes, a lot of it is the art. It's hard to replicate. But to your point, some of this same pattern matching exists there. You look at something like Max Martin, of course, say that Max Martin has taste, but there is melodic math. There is a pattern to any of those songs from the 90s or even today. So if you were to go into a tool to say, create me a song, that song sounds like the song of the summer. You can probably get close to what it would sound like. Whether that song catches on is different. And that's probably that gap that we're talking about.
- [09:55]
A
Yeah, I think there's two aspects to it. Like, you know, one is to say, let's just say that in the future, I don't think we'd think it's there 100% right now that an AI with the push of a button and the perfect prompt can produce a song that is as perfect as something that Max Martin wrote. There'd still be the issue then that I would say that if that machine can produce the perfect Max Martin song or something that would be as resonant as the Max Martin song, everyone has access to that machine. So then that is where the human taste element comes in. Because you're going to have to figure out some way to get between your human brain and whatever the machine. If you're working with the machine, how do you do something that's sufficiently differentiated? Because all of a sudden you're going to be competing with a thousand Max Martin track. Right. So the human there still matters. And then we get to your point, which is. All right, again, let's assume that the machine can do everything in the craft aspect of it, or the machine plus the human together can do everything. Okay, great. We actually already have that problem. We had that problem before AI, which is there's a ton of amazing music or amazing videos that are on TikTok or YouTube or it's an indie artist distributing their music. Again, whatever the platform, and that isn't getting heard. There's already a technology discovery problem of how do you break through. That doesn't go away. So again, write the perfect song. You still have to find a way for it to connect with humans. So those are some of the really interesting challenges tied to music, even when AI can solve, so to speak, the craft part.
- [11:23]
B
Yes, you can create something that looks engaging. Maybe it turns into a viral clip that's on X or some other platform, but then the demand for it just goes away. So it's tough. Again, you could create the thing, but the connection just isn't there.
- [11:36]
A
Yeah. And I think there's another aspect too, in terms of just the state of the tools and then the state of the discourse. But I think a lot of times the discourse around art and AI tends to talk about everything in extremes. Right. It's either the artist, right, on one side of the spectrum, using entirely traditional tools and never touching a computer at all. And then on the other end of the spectrum is the person that is just pushing a button and pumping things out. And that exists, of course, on both sides of the spectrum. But there's so many people, and increasingly in the future that will be in the middle. Meaning you are an artist and you want to create and you will use all the different tools at your disposal because you're so excited about creating. And sometimes that'll be something that uses AI. Maybe it's generative, maybe it's not. Sometimes it will be absolutely nothing to do with AI, but you're going to be making art, and that, that's going to excite you. I think that's the reality of, like, where we will be moving to. And none of that, of course, gets rid of the debates around really legitimate problems and debates around things like training data or around streaming fraud. But if we're just talking about how people will create, I think a lot of it will be in the middle. But again, when we talk about how people talk about this, I think for a lot of people for whom creation is a difficult task, or they want to create something to get from A to B faster, or to give an idea or to be able to convey that idea to someone who is an expert, these tools can be very useful. So I love playing music, but I don't know how to draw. I don't know how to do anything visually. And being able to have something where I could say, oh, like imagine a horse on top of a bridge or something like that that I could show to my kid, that's a very different use case.
- [13:17]
B
And this also ties into the trend we're seeing from the streaming platforms today of labeling AI music and even the decision of whether or not to label the music. Because if we believe that the middle is where this goes, where some people will use it to some degree, they may not use it all the way fully, then the future of a lot of music will have some AI creation involved with the process. But now the. The media around it can be so polarizing, where you either have a Jack Antonoff that is saying, no, anyone that uses this or is a loser. I forget if that's how he said it verbatim, but it was something along those lines. Or you have Timbaland that's like, not only do I use this stuff, I invested in the company. And here's me in a music video next to these AI artists. And those of course are the loudest voices because they're on the extremes. So maybe there is more that's actually in the middle there. But that's where it feels like a lot of this is heading.
- [14:09]
A
Yeah. And just going back to Anthropom, I hadn't heard that. But it's interesting because let's say the training data debacle didn't exist. Let's say all these systems had behaved ethically from the start and artists were participatory but let's just say then you probably still would have people saying what he's saying, but the argument would maybe not be as blurred. Like, is he saying that because he believes that artists were ripped off in the creation of these tools? Or is he saying that because it feels like cheating? And I think you'd still have people that would say it almost in the bucket of it feels like cheating. I think I saw something the other day that Queen originally posted something or posted what am I talking about? In the 1970s, Queen posted to LinkedIn. But I mean like Queen said something when they released their album of like, no synths were used on this album, you know. But I imagine you'd still have people that would say, I don't want to use these tools because I value so much the difficulty. And they might not say explicitly those words, but that would. Where it's almost like a cheat to be able to have a co collaborator or someone filling things in. So it's interesting. And that's where I say sometimes a lot of these things are conflated or harder to pick apart because there's so much that is heated around generative AI between data and between the creation act that you're kind of wondering which is the main driver here. I think in a lot of cases it's actually the training data that drives a lot of the animosity. Understandably. To go to the AI labeling part, I don't envy any of the DSPs, the distributors, the labels, the art, anyone that has to think through this, because it's both a seemingly silly debate or moment in time while also being totally necessary. Now, if we're trying to tackle streaming fraud and we're talking about push button creation again, even before generative AI, that was a problem, and that is something that deserves a huge amount of attention. But when we think about how works are being created or will be created that are largely in the middle, right? Someone used stem separation to do this or that, or they use style transfer, or they maybe composed a piece but then used a, let's say an ethical generative AI tool to fill in the drum part of the track. What are we deciding is okay? And isn't. Is 20% AI okay? Is 80% AI okay? If every single step of that track creation was human in the loop and a human directing the machine towards what they wanted, where are we deciding that something is? Where it's okay for a synthesizer to be there, but not for a more modern version of a synthesizer, so to speak? I don't think any of this labeling will exist. There will be something and that lots will be tackled around streaming fraud. But I just don't think the way it exists right now is how it's going to exist in two years. At the same time, you probably have to do something to move the debate forward. But I think it's a very hard space because it's like the tools that are deciding what's AI detected, who's auditing the auditors again, who's deciding where you're putting the bar in terms of what's okay and what isn't. I'm very empathetic towards the people that are trying to tackle it, but I'm also like philosophically, except at the extremes. I don't love it.
- [17:14]
B
Let's take a break for our chart metric stat of the week. Deezer, the music streaming platform, recently shared that 44% of all songs uploaded daily to its platform are AI generated music, which account for nearly 75,000 tracks. This led me down a Deezer rabbit hole to see how do the top tracks on Deezer overall compare to the top tracks on other streaming platforms? Well, on Deezer, Michael Jackson's biggest songs, Billie Jean and Beat it, are currently 61 and 68 on its global daily streaming chart as of June 1, 2026. And if you compare that to Spotify and Apple Music, you'll see Michael Jackson regularly in the top 10, if not higher. Deezer itself is more Europe centric, predominantly in France, so there is a bit of a demographic difference there. But it's always unique to see these platforms and how they compare elsewhere. But I do think that the percentage of AI music, especially of Deezer's showing this, may not be too different on some of the other streaming platforms as well. Let's get back to the episode. One of the other things I think about with these generative AI tools, especially for you with audio Shake, you're not generating music, but you are processing music. Has this ever came up with you or any of your companies where the output of what AudioShake has created? There was some partner that wanted to display that. Oh, the separation and processing of this audio was made with AI. Has that ever come up?
- [18:43]
A
What can happen is people just don't know, right? They assume it's generative AI. I've been asked before what foundation model we use. So what we do, which may be useful to explain is we work on the image of audio. So if you could imagine what the spectrogram looks like, you would see a whole bunch of gray bars and a Machine doesn't know, is this a voice? Is this piece of music? Is it a drum, Is it a bass? And we essentially color in. We identify what are the different pieces of components or stems in that music. So we'll color in the bass all blue, we'll color in the drums, green. And then what we're going to do is we're going to silence all the parts. We're going to black out all the pixels that don't correspond to what we're searching for. So if we're trying to create a base isolation, we're going to black out all the pixels that are not base pixels, and then that leaves behind what you're searching for. So it is a subtractive process, it's not a generative process. You are working on the image of audio and you are removing, subtracting sound. So you don't generate anything, you don't interpret anything. There's no risk, for example, that you would hallucinate or introduce any information that wasn't already there. That certainly can be an education moment sometimes when you're talking to people because they just assume that if it's AI, it must be generative. In fact, if you're thinking about everything we've been talking about, the body and like and scarcity, what rights holders and content owners and creators do with our technology is they essentially monetize human performance and creativity, right? They're taking existing work and then they are isolating it in order to be able to create additional derivatives, sync licensing, atmos mixes, dubbing. If we're talking about film, it's helping them create even more longevity or relevance or greater distribution for the assets that they have. Whereas if you were using Generative in what we're doing, you would run into the danger of possibly creating output that doesn't sound like the input and wouldn't have the same value. Value migrates towards what is scarce and I think what is uniquely human and
- [20:36]
B
the output being traceable to what the origin source is, does remind me of autotune, which I know is another technology that's come up. When people talk about AI or there was a lot of debate 20 years ago about how people were using or not using autotune, Nas had a whole album about, you know, hip hop being dead. So I do think that even though generative AI does have additional challenges, and maybe AI more broadly, I should say, I still do think that the future of it, this is not going to be something that is going to be as important to label, because we're going to Figure this out in the same way that 30 years ago, a company being an Internet company was something that may have seemed relevant, but less so today.
- [21:18]
A
I just think ultimately we're creative beings. If you are an artist, if you are a musician, if you're a writer, the existence of these tools does not mean that you will stop creating. You may use them, you may not use them. The way that other people that are not creators in your space may use them is very likely fundamentally different in nature. So I love playing music. I also love writing. Visual art is not my thing. And so the way I would use a generative visual tool would probably be somewhat infrequent and very utilitarian, right? Like, how do I communicate this idea super fast to someone who actually knows what they're doing? The idea that there could be something I could be riffing off with when all of a sudden it strikes my fancy at 11 o' clock at night where I could be playing something on the piano, and then there could be this tool that's suggesting other kinds of sound. And then maybe during the day, I then take that to a friend of mine who does know how to do that, and I'm like, what do you think about this? That, to me, is really, really exciting. And I do think that there are these larger debates and legitimate issues that have to be resolved. But if we think about this, again, as a very smart wall or drum or synthesizer or table or tool, like a tin can, whatever it is, we will find as artists, ways to create with it that get us excited.
- [22:33]
B
And one takeaway from this for any executive, founder, investor out there is if you're trying to find what is the best layer to build experience or business or opportunity on. Assuming that we've addressed the table stakes of making music easily accessible and easily creatable as well, one of them is investing in these things that are scarce. We already see a ton of investment in live music and entertainment. That, of course, is one of them. How do you find the things that are hard to do and create more value in that? Would that be a fair takeaway, you'd say?
- [23:06]
A
Yeah. I mean, I think the things that are very uniquely human and that connect the human artist to their fan, they've been valuable since long before Elvis was shaking his hips on TV or whatever, and they will continue to be valuable. And so then it's a question of just what are the mechanisms by which people do that? And I think some of those exist today, like live performance. And I think there are other ones that might surprise us in Kind of quirky ways. Like my son, who's 11, asked for an MP3 player for his birthday.
- [23:39]
B
MP3 player? Like an ipod?
- [23:41]
A
Yeah, yeah. And I was like, but okay, but we've got this account. You can just use this account and you've got your Chromebook. And he's like, I want. Like he wanted to hold something and he wanted this thing that was his and he doesn't have a phone yet. And so it's interesting, right? We like to touch things. We like physical things. I think there are things that we will anticipate will continue to be important again. Live performance, fan connection, fandom, those sorts of things which can exist in all kinds of different flavors. And there will be things that are either very much from the past or that are futuristic that we haven't anticipated that will emerge as well.
- [24:10]
B
And from a business perspective, I often wonder, you know, what's the ceiling for that? Right. We can see it in the data. Vinyl sales in the US have topped $1 billion, I think, for the first time since. I don't even know however many years. But like, how far does that go? Right? Because I don't. We probably aren't going to get to the point where, okay, this is going to eclipse streaming. This is going to be even a quarter of streaming.
- [24:28]
A
Yeah.
- [24:28]
B
But it feels like in general, there could be some way to look at it across mediums, not just in music, but you take the thing that is dominant but abundant, and then you look at the thing that is scarce, but maybe a bit more analog or old school. What is the ceiling of that relative to this? And is there a similar ratio or a similar comparison across industries?
- [24:48]
A
Right. And then how do you triangulate that with actual value of the good? Right, right. Of how much people are willing to pay and what's the demographic that's most interested in paying for all of that? I mean, it is interesting when you look at historical data. I think we would all have expected physical goods to just disappear, but they didn't. Right. They bottomed out, but then they continue. And then maybe that continues as vinyl, but then somehow CDs and DVDs grow. And again, I think that it's really useful to take a step back and look at broader societal trends. I look at my kids generation and I look at how they talk about the Internet, and I look at people that are a little bit older than my kids, like my nieces and nephews that are like early 20s or that are teenagers. There does seem to be a little bit of backlash towards digital things. Even though they all are still fully digitally native and can navigate those platforms and so forth. But there's a real value placed on kind of community. Any generation takes things in one extreme and the next generation is going to kind of want to go in the other direction. I haven't owned a DVD in how long have you owned a DVD?
- [25:46]
B
Oh, mid 2000s.
- [25:48]
A
Right. And so, like someone 10 years younger than you, like, probably would not want to touch that because that's like old Dan stuff. But maybe someone 20, 25 years younger than you, maybe that becomes interesting, right?
- [25:59]
B
And then they're like, oh, you have all these hidden scenes here and this like, second side of the disc. I didn't realize that was a thing. We just completely overlooked that.
- [26:07]
A
Yeah, it was the same with the vinyl and my parent, like, they were so dumbfounded. Why would you want to have vinyl when we fought for you not to have vinyl? Jessica?
- [26:15]
B
You know, and going back to coding, this does make me think, is vibe coding an element of this for the engineer that still feels like, oh, let's say the main thing that they were doing for a big company is now been outsourced because the tools can do it for me, they need less of me, but I can go vibe code my own things purely for my own personal enjoyment, not necessarily for utility. This inefficient thing that still gives them joy.
- [26:40]
A
I would actually bet more that the engineer that's feeling displaced somehow or that's feeling existential about things is actually just going to go and take up gardening. I think all of us that spend a lot of time in computers and coding and so forth, like there's a level of abstraction sometimes in the job that you sometimes wish there was something more tangible. We still need humans that have good curation and decision making skills. I mean, babysitting isn't even the right term, right? Like, they need a lot of direction. I think what does change is that maybe perhaps the nature of engineering, right? To get something really good out of all of these tools, you do have to spend a lot of time building the prompt and the instructions. And the more time that you put in up front, the more value you get back. That's a fundamentally different task, right. It puts less into the discovery process and more into the recipe building, if you will. This past weekend, I was playing around with an idea I had and I was so irritated that the answer really to what I wanted to do was just to giving very clear instructions at the start because I wanted this very iterative process and wasn't getting what I wanted out of it. And I was sitting there trying to go in and then tweak the code, and finally I just, like, gave up and I just wrote incredibly clear instructions. And Claude did everything for me and did it exactly the way I wanted it. And it was both a triumph, but also just kind of like, why did I even bother? Right? Like, why did I ever even learn these things in the first place?
- [28:05]
B
Right. That goes back to what we're saying in the beginning part, where the people, at least at this stage, the people that can get the most value out of it already have some heightened level of expertise with the underlying product.
- [28:15]
A
Yeah. And I see that across sectors. Like, I don't think that's just an engineering thing. I think that people that know their craft really well can do a lot more thanks to these tools.
- [28:25]
B
So five, ten years from now, if everything that you laid out in this essay rings true, everyone in this industry still has jobs. Their jobs have changed, but everyone still has jobs.
- [28:35]
A
I think the thing that I get a little tired about hearing AI people say, or tech people say, that I've had to hear for the past 10 years is everyone in tech loves the story of the atm. So the story of the ATM is we all thought that the ATM was going to get rid of bank tellers and that banks were going to be all ATMs and no one would need to interact with humans. Turned out, no, they gave different jobs to humans is what happened. And that when you look across, you know, history and so forth, new jobs are created. And so that's like the optimistic look, which is to say we will have new kinds of jobs. They will be different jobs, but there will be jobs. And I think, particularly in the creative industries, we like to create, and there's a lot of scaffolding, I think, that needs to be created to help people that really love to create, to help to get their creations out into the world. But I think none of this gets rid of what that transition looks like. So many people love taking Uber and Lyft, and no one feels, it seems particularly bad about what's happened with taxis. But to that taxi driver and their family, Uber and Lyft were incredibly disruptive. And to the Uber and Lyft drivers who will then be automated by the Waymos and the Zoox and the own automated versions that a Lyft or an Uber will eventually have, or they'll pivot entirely to food delivery or whatever it will be, that will be incredibly disruptive. And so even if as a society, we net out fine because some jobs go away and new jobs are created for that transition generation. I think that's an incredibly painful process. And at least sitting in the US where you don't have much of a safety net for those people, I think that's something that we don't focus enough on.
- [30:12]
B
And I think with music, there's aspects of both of those that you can kind of see how it'll play out. But Jessica, appreciate you coming on. Really insightful.
- [30:19]
A
Can we end on something like positive? Geez, AGI. Jessica, on that note, I'm gonna go catch an Uber.
- [30:29]
B
And that is a wrap. Thank you again to Jessica Powell from AudioShake. Thank you to Gian Eric, our audio and video producers, for everything that you do to help make Trapital possible. Thank you to Rwana on our team for everything you do behind the scenes. And most importantly, thank you for watching. If there's one person you know that would really enjoy Trapital and get a lot out of the show, then send them a link. Whether it's this episode on AI Music or any of the conversations that we have, word of mouth is still the best way to grow. And if you have a few minutes, leave a comment, leave a review, make sure you're following the show. That helps make sure that trapital is reaching the right people and that the algorithm does what it does so that it can reach others as well. Thanks again. Talk to you next time.