Episode 158: Reality 2025: Bridging AI, Security, and Open Source Challenges

Speaker A: Hey everyone. Welcome back to reality 2.0. It's been a minute. Um, it's been actually kind of a while since we've released an episode. I, I hope you're still there out there on the Internet.

Speaker B: Are we at 2.4 or 5 by now?

Speaker A: Maybe. I think we, I think we may need to kick up to a uh, new major Release.

Speaker B: Maybe reality 3.0, reality 2025 because it's our first one this year.

Speaker A: It is, yeah. And there is, you know, we've been busy. Apologies for abandoning you all and. But we're back because there's always so much to talk about and we, we miss talking about it. So, so here we are, we got a few things going on. Depending on when this thing goes out. I will either be in the middle of Open Source Summit or having just completed it.

Speaker B: But that's, you know where, where is it this time?

Speaker A: It is in Denver. In Denver, Colorado. It is June 23rd through the 26th ish or the, the Colo. Things on the 26th there are things like Open SSF Day which I will

Speaker B: be a part of and who puts

Speaker A: it on the Linux Foundation.

Speaker B: That's what I thought. Okay.

Speaker A: Linux foundation and it's uh, it's sort of subsidiaries. So OpenSSF for example open Open Source Security foundation has a day on Thursday and Linux uh, Foundation AI and Data has a AI mini summit also on Thursday where I am also speaking. I am a little over committed as usual. But um, yeah there's, there's, there's going to be some really good stuff. A uh, lot of interesting people coming together to solve the tough problems in the open source world and many of them are about AI and security right now. So I feel like I'm in the right place.

Speaker B: Excellent, excellent. So what is okay, it is because I haven't looked at, I mean to me the Linux foundation just keeps growing and growing and growing and budding off with new foundations within foundations and stuff like that. And um, what I'm wondering is uh, is it devoted to AI in one special way like it was say to uh, Cloud or to you know, one of its other specialties? Or is AI just in everything now? And so it's just so both.

Speaker A: So the Linux Foundation AI and Data is sort of the dedicated organization to issues and projects real specific to AI and data. There are a few things I'm involved in there. One of them actually is a kind of a new conversation working group around MLOps.

Speaker B: Right?

Speaker A: How huh, what is the best way to do this thing right? And then deploy It, So, so that's, that's part of it. So that's kind of the dedicated topic focused org for AI. But there are relevant conversations happening in other Linux foundation groups. Right. So Cloud Native Computing foundation just. They had.

Speaker B: Yeah, the cncf. That's what I was trying to think of.

Speaker A: Yeah, so CNCF Big One. That's, that's, you know, that's with their

Speaker B: landscape, which is always interesting.

Speaker A: Yes. And it's massive. 208 projects last I counted. I give a presentation about that a lot too by the way, about how overwhelming that is and how to even navigate it. But um, but yeah, so the CNCF has, they just announced, I think it was in China at the event, big event there, a new AI conformance initiative. So, so they're, they're, they're looking at it from, just uh, from the, the cloud native part.

Speaker B: Right.

Speaker A: Obviously that's their purview and so, so that part of the conversation is happening there. And then for example, also OpenSSF has an AI and ML working group. Right. So they're talking about specifically from a security angle. Right. What's the conversation there? And they're coming out with some resources and, and you know, expanding the conversation to include how to secure AI applications and then also the security implications of generative AI applications. Right. Because they're, that introduces new concerns. Right. It, it shortens the path for people sometimes to do nefarious crappy things. So there are a lot of smaller conversations happening. But, but they do. The AI and data is kind of the dedicated group. So. So yeah, it's both.

Speaker B: Yeah, there, there's um, uh, now I'm looking for it all of a sudden in a hurry and I can't find it yet. But uh. Because I had this conversation with um, uh, chatgpt this morning about, because I was in a meeting where somebody's going over all the stuff that Apple announced at its WWDC last week. Um, part of which is, and I have no expertise or direct knowledge about this, um, but they're, the idea is that their approach to AI is going to be a personal AI anyway is that you will have um, on device AI that you can do and then in your cloud AI you can do and then, and both of those are like supposedly at least completely, completely private. And then there's a door in the second one that opens out to ChatGPT and uh, or, or the, or the uh, LLM of your choice. So the, the, the question I have about that is okay, when you open that door and you're doing personal stuff out there. Here are my health, here's my health data, here's my finance data, here's my travel data. Here are my contacts and calendars. Here's my work history. Here's what I'm doing at work right now. I mean all this stuff is really very, inherently very private stuff. Um, and can you do it in such a way that a ChatGPT can't use that for anybody but you? Now I imagine if Apple's got to deal with them, that would be it. They'll want to have some sort of secure firewall ish kind of boundary around that that limits what Chachi Picotee could do with it. But how can that possibly be audited? How can that possibly be uh, fully understood by, you know, I mean we don't even have, you know, the, the, the white coaters working for some government agency that can walk in and say how are you doing this stuff? Right. We even had, we, we never, we never had that for search. We never had that for anything else yet. Right. We have it for nuclear plants, but we don't have it for, for big tech. So, but it's a, it's an, it's an important question because we're, we're going to need that stuff. And, and ideally to me, and this is what interests me most about this is that I, you know, my, my health data, my financial data, I think like, and I probably have more than most because I'm old, but they're, I've accumulated more. But I think anybody um, has a fair amount of this stuff. But if you take, if you pull in a model that can help you understand and organize and catalog and, and then put in a form that you can get bring to another broker or I mean like a financial, not even financial broker, like a, a financial advisor or something like that. You don't need big compute for that. You could possibly bring it, bring in a. Yeah, you could probably bring in a model to do this. So I think, I think there is a big potential category. It might even be something like um, you have your own box, you have your own AI box. Or maybe you, this is what Kawaii is doing and there are others doing it as well. You can mesh up a whole bunch of boxes where you encrypt it and shard it out or whatever. I'm not sure how you do it, you know, but you know, that's what coin that's doing like I, on, on a shelf behind me over there. You can see it. But yeah, so I know that's an old laptop.

Speaker A: It's very interesting and it's. Yeah, uh, yeah.

Speaker B: And by the way, right now we are talking over my new for 59amonth. I've got a 2.1 gigabit symmetrical fiber connection here.

Speaker A: Wow.

Speaker B: No data limits on it. Do what you want. Yeah, Pretty damn cool.

Speaker A: That is very cool.

Speaker B: It's really good. It's really good.

Speaker A: Um, but yeah, so the, the you bring up the, the quiet thing. That's, that is a very interesting thing this idea. I mean I guess my understanding in a nutshell, aside from the fact that it's a distribute, it's distributed, right. It's using compute power from all over, well the world probably at this point. Um, there's also it, it basically functions as like a rag app builder as a service. Right. So you, you know, you pull in some large model via API but then you train it. For example, as I understand there is a doc bot that is basically.

Speaker B: Yeah, somebody threw a bunch of my writing in there and there is a

Speaker A: doc bot using the RAG method. So instead of retraining or you use the base model and then you augment with this data set which is Linux journals. So that's very cool, right? That you have this ability to easily spin up something like that that you can then augment with whatever is specific to you. In your case Linux journals or your, all of your blog ent or anything like that. And I think that's very cool. And there's a lot of new potential with other methods. I am um, part of a graph rag tutorial thing next week actually which is a whole other thing in terms of identifying these relationships between data. Uh, again I'm not actually a super expert on this but somebody else will be there who is. But given all of the progress that's being made in all of these areas. Areas, there is so much potential. But like you say there it. There are issues of reproducibility, there are issues of understanding how all this stuff works especially in, in certain industries, right. Like finance, you talk about health and stuff like that. You need, uh, especially in regulated industries, you need to be able to just describe how you made the thing. Yeah, for so many reasons. Right. Safety, privacy. So yeah, so, so you know, part of the answer to that is guardrails. Part of it is, you know, other things, security, best practices. But a great opportunity to plug. Uh, one of the conversations that uh, one of the presentations I'm doing next week is about this exact problem is why, why we need standardization and why we need to kind of simplify and, and codify the best practices and building these things because you have to be able to explain how you did it. And yes, there, there's so much of this. It's such a big a black box. But, but at least if you can have some explainability in the way that you build the app, build and deploy the applications, you improve the experience for all everyone who's working on building them. But then you also, you kind of start to solve that explainability and reproducibility problem. So in it. But you help it become a little bit less of a black box like the, for certain parts of it anyway. I mean the LLM itself is a little bit of a. That's a different conversation.

Speaker B: But yeah, there's almost too many conversations and then they'll all go off in tangents. One of the things that I'm noodling on lately is that we have a model. Model's M the wrong word because the model has multiple meetings there. We have a routine or an expectation that the way we deal with NLM and ah, therefore we deal with AI because that's now most of our experience with AI and is that we. It's. It's Q and A. I have a query, I, I have a prompt, I get a response. But let's see if I'm looking at, I want a list of all the books I've got. Well after I've asked it to put it in a spreadsheet. Well, I'm just looking at a spreadsheet, you know, or I might want to look, look through all my health records and I want to go, I want to look at it back through time. I want to look at it this, I want to look at it, um, uh, categorically, uh, you know, here's hematology, there's oncology, there's gastroenterology, and, and the. I have a feeling that neither the new Q and a prompt and response model nor the old app model with its um, here's, you know, put together a document with Word, put together a spreadsheet with Excel. Neither one of those is adequate. There's some other view that we need on our stuff and also cross referencing stuff. I mean I, I want to look at a map of where I've been. I want to look at um, you know, when I have a question and it's not always a question. It's just like I just want to check to see, um, when was the last time I met with this person. Right. I have another meeting coming up. I was on a call today where I was reminded that I actually spoke to one of the people on the call, like, you know, a year ago. And I don't remember that exactly. And if I'd known they were there, I might have been able to look it up and, and dig on it. But right now, um, when I look it up, I'm going back through email. Is it an email? I mean, you know, five other. So many people are now using signal. Some people use WhatsApp, you know, and I can't cross reference those. They're all in different places. They all amount to a kind of unstructured data that some kinds of AI deal better with. But I have a feeling there's a, there's a navigation model for this, for our personal knowledge. That's not one of the above. And, um, that's just something that's occurred to me in the last several days that we don't have the. It's early days with this. It's really young, the whole thing. Yeah.

Speaker A: And people are, we need that vision of, you know what? You have so many of us nerds kind of in the background trying to figure out how to put the pieces together and build the blocks in the right shape or whatever, but you really ultimately need, you need to have the conversation to have the vision of really. Okay, but, but at the end of the day, you know, regardless of how we build the thing, why and what, what is.

Speaker B: I know it's a, uh, what are

Speaker A: we building and what is the purpose of what we're building?

Speaker B: It's also like the metaphor that. Yeah, as, as you're talking. The metaphor that comes to me is like, okay, we think we're working with Legos, but it turns out it's really. We need cookie dough here. But it's really. Maybe not that. Maybe it's, it's uh, asphalt and concrete with rebar. I mean it. There, there are, there are so many different actual and potential approaches to what we do with gigantic, well, programmed pattern recognition.

Speaker A: Right? Yeah, exactly. Well, given the nearly unlimited potential, maybe, um, we still, we still think of things in very old, um, school terms, especially from a perspective because again, I'm talking about that all the time. But I, I always use the example of like the input, like an input box. Right. A text field. Right. Now it is no longer in the, the Internet of, of X number of years ago. You know, an input field is, was a pretty simple thing, right? It was. You expect and you could test and you expected to get this value and you would, you know, and Then you get this value out. And it was very predictable, deterministic. Um, but now we're, we're in this totally different world and yet we're still using just an input text field. It's a very complicated and convoluted input field, but it's the same thing. Yeah, because there's a lot of uh, let's say it's a broadened threat attack surface or whatever but because again it is a far more complex input field. But at the end of the day we're still working in the same construct that we used to, which is just this kind of very little simple box and you type text into it and then you get, you know.

Speaker B: Right. Yeah. With it. With a QWERTY keyboard that was invented in 18 something.

Speaker A: Exactly. Yes. That is exactly where I was going. We're still thinking of things in kind of the same way. Way. Yeah, I don't know and who knows in five years if we still will be or not.

Speaker B: But, and I've noticed and I've you know, read um, a little bit about this, you'll probably know a lot more that um, the more, the more casual and colloquial and in some ways even uncertain. Your language in a prompt sometimes brings better results than as some sort of carefully worded thing that say a lawyer would come up with or interesting, you know, that you've run it through Grammarly and um, uh, and it's, and it's taken out you know, your two complex sentence structure or made it complex for you. It just seems to me when I've, when I've noodled with chat, GPT or Claude or any of them, I'm getting somewhat better responses than I get when I, than I got early on when I very carefully construct a prompt. But again, you know, we, you know, when you're driving a car, you know exactly what's going on. Right. You steer it this way and that and it goes exactly where you're thinking it's going to go. It's very, I guess deterministic would be the right word. It's mechanical almost. But we don't have, I mean we certainly had that with old fashioned programs like Excel and you know, any of the apps that we use, you know, a map app or something, they're, that's another thing I think part of, part of the way that the ad tech world works. We don't have to segue into that but the sub, the, the suppositions behind it are, and I think that AI has inherited some of these that it's good that it's guessing about you at all times. And the more, the more information it gathers. I wouldn't say it knows, but the more information it gathers, the better its guesswork is going to be. Now, ad tech is all angled toward making you buy something, which AI is not. It's just angled toward, um, giving you better information each time you ask it something, um, or giving you what you're asking for. And it doesn't necessarily lead into a sale. Which by the way, I think is an enormous relief from the search world. I mean the search world is so inside now on Google. I mean I, you know, the, the first I, I swear to God it was something I looked for the other day. Like the first three pages. Responsive results. Like, where was the result I was looking for, which might have been a Wikipedia page. Right. And one of the reasons I think we like having a conversation with an LLM is that it's not trying to sell us something. It's not, it's not. It doesn't have an ad in it for something or other which automatically biases the results. I would think I've kind of drifted there.

Speaker A: But it's like, I mean, well, I mean it all goes back to, I mean it's the same. It goes back to the practicality of the thing and the why of why we are building any of this stuff in the first place. At the end of the day, some of it's really handy if you, if you dig in and find, you can find uses for it. Maybe it has to do with productivity, maybe it doesn't. Maybe it's just fun. You know what? I, you know what, I use chat GPT a lot for, um, maybe semi embarrassing but, but I'll talk about it anyway because it's fun. So, um, what I, I try on purses and like, try on outfits and I go get like, like what colors might look good on me and what. So I've, I. You can upload like a. Some number of photos of yourself that maybe you like and you can just. And then I describe. I'm like, well, I'm this height and you know, and my, my skin color is like this and pretty pink and um, like I have like super pink cheeks. Anyway, um, and you know, I'm this height. I wear this size and, and you know, I kind of like these colors and this brand. And then you know, and, and here's a picture of a bag. What would that look like on me? And it's actually so helpful.

Speaker B: Really? That's interesting.

Speaker A: Yes, it's Almost creepy. Um, in fact, I will. I will plug. I spent some time last weekend rebuilding my personal website because. For fun, because I wanted to try a static. We've talked about static site generators in the past. We had a whole episode on Hugo once with Sean Power. So I tried a. I wanted to try this one. It's astro. And it was lovely. Hugo also lovely. Anyway, so I made myself a new website. Um, if you were to look at this website, you would see a photo of. Not me. It is based on photos of me. Although frankly, it looks a little better. It's the me I could be but trying on. Anyway. It came from one of these, like, chat GPT sessions where I'm like, what if I'm wearing Wore m. A blue shirt with. You know. And, and anyway, um, it's just kind of a fun, weird thing that I do. And it turns out it's actually incredibly useful. It's. It's incredibly useful for a lot of other things too. Um, you know, I, I ask it to create, you know, again, make a spreadsheet template or make a template for something that I want to put together, even if it's some like, just simple personal project, but if it can be really useful at the end of the day. And, and so, so what does that mean? Maybe the answer is that our, our old way of just having an input text field is the right one. I don't know.

Speaker B: Well, it might. I mean, it's an interesting thought because, um, you know, Marshall McLuhan, who's one of my heroes, says that all of our technology is an extension of ourselves. I have a cup here, and it's an extension of my hand. I don't have to cup my hands in, in a pond and pull up water. I, I have a cup that I could use instead of that. It's a. You know, a bicycle is an extension of your foot. You know, a pencil is an extension of your fingers. And, and, um, you're. When you're in a car, you're, you're, you're. You're a machine. You're partly a machine. And those are, those are my fenders and my wheels. You know, there's. That Your feeling extends out to it. But we are also talking animals, right? We talk to each other. So why wouldn't we talk to machines? You know, especially if the machine is coming, is able to come back with, with what we would call an intelligent response. I mean, with a rational response based on information that it's gathered. And, and that is, It's a very Human thing to do and a human thing to want. They are. They are extensions of ourselves. The part that creeps me out is like, my car is an extension of me, but I'm not. I'm not in a. A giant carscape where I don't know what's going on. I mean, it's the black boxedness of it all.

Speaker A: Yes. It even works at all, frankly.

Speaker B: I know, I know. It's. I mean, it is miraculous. And the mistakes it makes are. Are, uh. You know, it could be quite amazing actually, sometimes. But. But it does make this kind of best effort. But I feel like, you know. You know, what are we. You know, there's the. There's the. The famous, uh. I think it was a Twilight zone from, like 60 years ago. Um, To Serve man was the name of it. Do you remember this one? Have you ever seen it? Um, all the Twilight zones are like 20 minutes long.

Speaker A: One I always think of is the one where the woman wakes up and everybody has the face of a. Like a pig or something. It's really creepy.

Speaker B: Oh, yeah, that's a. That's a creepy one. Yeah. There was also another one where, you know, it. Everybody used to look the same, and so everybody with a normal face as the pig is one of them. The, um. Yeah, another one is that everybody, you know, at a certain age, they have to convert to having one of the 10 or 15 standardized looks that a human being can have because they get altered that way. But this one, um, was, uh. Uh, you know, the aliens come to Earth and they bring a book with them called To Serve Man. And they seem to be very beneficial and very beneficial and everybody loves them. And everybody starts. They said, you're welcome to our home planet. And this guy's lined up, getting onto the spaceship, and then this woman comes running to say, we found out what To Serve man means. It's a cookbook. And, uh, that's a little. I wonder whether or not there is something to AI. Whether we intend it that way or not, that it's a cookbook.

Speaker A: Well, there are always unintended consequences. Not to go off on this tangent, but if you read some of the articles lately about people forming, like, unhealthy relationships with chatbots.

Speaker B: Oh, sure, of course.

Speaker A: And going in there and, like, feeding their delusions or. I. I can't remember the details, but, I mean, none of that surprises me. Whatever we build will be misused.

Speaker B: Right? Um, yeah. And, well, it's also. I mean, this is one of the. To me, one of the more annoying things about it is that there's, they're programmed to be flattering to you, right? You know, they're, That's a great, that's a great question. Oh, that's so insightful. Here's what I think about that now. You know, okay, um, quit blowing. I mean, you always think it's blowing smoke up your ass a little bit, you know, totally. Um, but, but that's, you know, a kind of flattering, low grade courtesies are sort of part of human nature as well. It's part of the, the, you know, the protocols that we have for dealing with each other. So why should our machines do the same thing?

Speaker A: There's, there's so much potential there, there. You know, we talk about, we talk about a lot of risks and we talk about, you know, the new threat landscapes and how are we going to address the, you know, security issues that have resulted from all of these things or privacy issues or, you know. But I would like to remain optimistic. I think that there is a lot of use for this technology. Helping to solve the easy problems, helping to make things a lot like uh, again, I'm always thinking about from m, the nerdy developer angle. But if I can use an AI based tool to save me time having to do easy security checks, for example, or um, anything that's time consuming and not creative or rewarding, then that's fantastic. So you know, I, I'm optimistic. I think, I think there are more good uses for this kind of stuff than there are bad. And um.

Speaker B: But yeah, that's, well that's, you know, that's part of human nature as well. I think human, um, beings are incredibly constructive. I mean we're always building things all the time, everywhere. This is what we like to do. And we make messes, we tend to clean them up and then we make a business out of cleaning them up and then we invent recycling and then with, there's a separate truck for that. I mean there's, there's so, so many things that are, that are, you know, we're very enterprising animals and, and now we throw AI in the middle of that. So I'd like to jump the rails a little bit toward something that is just really hot in my, my life right now. Because we've, we've been working since 2017 on a standard IEEE P7012, now known as my terms, which completely flips the cookie consent thing. Get rid of the cookies. Just, you know, we, we have the terms every place we go. We're going to assert our own terms and they can agree to them or not. We record whether they did. We both have an identical copy of, of the. At least the fact that we agreed and what the term was that was sitting at Customer Commons or wherever else it might want to sit. But and the terms could be friendly. They can say things like okay, you can use data about me anonymously to train your AI um or you can go ahead and show me advertising, just make sure it's not based on tracking me outside of this site or hey, I know you already and I belong to your club and um, uh, and I, and, and and I want you to share this data with me and I'll share this data with you. There's a whole bunch of possibilities there. Now if we have an. Now this has been so anathema to the status quo but in the meantime um, Europe is quite fed up with it from everybody I've talked to over there and they're ready for the next thing and we're getting many many queries about this and it's starting to feel exciting and, and the, and I'm even thinking geez, you know some of the bigs especially Apple should be able to use this in as we go about our um, dealings with companies if you know like okay, they have great ads right now I don't think sure how accurate they are about Safari. I don't know if you'll watch the, the basketball playoffs but they're on every time. Um, a person sitting there where they're working you know and, and they're reading, they're looking at their tablet and Ace a surveillance camera with wing is flies up and like sits on their shoulder and stares at what they're doing. And then more and more security cameras with wings like this whole species of security cameras starts following people around and then uh, and people get more and more creeped out by it. And then this guy, uh, it's called flock by the way. Look at, look it up. Fl o c k it's on YouTube. Apple flock because there are flocks of surveillance cameras and this woman, you know on her phone clicks on a Safari uh Chiclet and um, uses Safari to blow up all these things.

Speaker A: That sounds vaguely.

Speaker B: And uh, this is privacy, that's Safari. And then the Apple has a little lock on the top and it closes and then one more of the things blows up in the sky. Um but that's going to work best if other entities are agreeing to our terms to begin with. And ah, what Apple has right now is you can globally actually on the phone Turn off tracking or at least as they put it, if you do it on a one by one basis, which you'll do on Apple TV if you have Netflix and the rest of them on your Apple TV 4K box. Every app will start with ask App not to track or allow app to track. But it's ask, it's not tell. So my terms will tell rather than just ask. So there's just a lot of interest in this all of a sudden. And so I'm kind of excited about this because I think there's, I think we can actually blow up the entire cookie consent thing with this. I think there's a real risk that we might make that happen. And having your AI agent on your side that knows, knows when you go to um, the Home Depot site, you already have relationship there. Here's what it looks like. Here are the terms we agreed to. And it's not just I've agreed to your terms that you've agreed to mine, but if they're friendly enough terms that we can make secondary agreements that will open up business. So that's what's going on in my life right now.

Speaker A: I think before we sign off though, I think we probably should, we should um, talk about that. We're definitely going to keep putting out new episodes

Speaker B: on more frequently than on a yearly quarterly basis.

Speaker A: Yes. We will increase our publishing frequency.

Speaker B: I would be glad to.

Speaker A: More than one per year. Uh, yeah. I would like to just kind of open it up to people listening and we, we welcome. Any thoughts on what you would like to hear about.

Speaker B: Yeah. And I mean we can bring in some. An important fact about, about why, um, we slacked off is that we're busy, you know.

Speaker A: Daring. Yeah.

Speaker B: You're working hard. I'm working hard. We're doing stuff. You get paid for what you do. I don't get paid for what I do. It's. I mean at my age, which I'm, uh, is embarrassing, but I won't say what it is, but it's uh. I'm at the point in life where it's interesting if you're still alive and doing well and are engaged, nobody wants to pay you for that. They think you're, you're having too much fun. You're at the volunteer age. Right. And also I'm at the no age. I'm not going to do anything that's a waste of time. So. And most things are so, so. But the interesting thing about it is that I, you know, I'm feeling more effect right now than I've felt my whole life, like there's stuff happening and,

Speaker A: um, you know, you live in interesting times.

Speaker B: You live. You know, we look for effects. I mean, that's what people do. They look for effects and they're doing things. Um, you know, so anyway, it's the opposite of angel from Montgomery, the John prime song, you know, John Prine. How the hell can you, I say, go to work in the morning, come home in the evening and have nothing to say, you know, that that's. Well, that's a life of quiet desperation, which a lot of people live and, ah. Which at times in my life I have lived, but I'm not right now.

Speaker A: So anyway, we have a lot to say, I think is.

Speaker B: Yeah, we have a lot to say because we're doing a lot and doing a lot.

Speaker A: We have a lot to talk about, so we're going to keep doing it.

Speaker B: Pearls to make an instrument string from species of oyster that don't exist yet.

Speaker A: Well, thank you for everyone who listens to us. And that is actually quite a few based on, uh, the last time I checked those downloads, dart. So thank you for that. Um, so yeah, so you know what? We will, we will, we will pick this back up next time and see what kind of interesting stuff we can dig up. That is indeed. Okay.