The Future of AI: Personal Agents, Taste & Private Data | Lin Qiao & Demi Guo | E9

Added: 2026-05-05

[00:00:00]People in America obviously have a very negative view of AI.

[00:00:03]>> Gen Z's are cynical about AI because they don't feel it's authentic. They grow up with AI around them, all the tools, the combination of human creativity. I do not think I do not believe that be replaced.

[00:00:15]>> The human value in the future will come from the judgment and taste. The mental model of the agent should not be a tool but actually just be another human.

[00:00:23]People choose who they work with not for only for productivity. They just enjoy working with this person. every two weeks or sometime every week there's a new model released. Open, close, doesn't matter. And they're oscillating with each other on the leaderboard. Whenever things are oscillating, it's a clear signal they're converging.

[00:00:40]>> Thanks to our friends at PayPal, the exclusive sponsor for this week in AI.

[00:00:45]Try the payment and growth platform that's trusted by millions of customers worldwide. PayPal open. Start growing today at paypalopen.com. All right, everybody. Welcome back. It's this week in AI. This is the new weekly podcast I am hosting. You get this week in startups Monday, Wednesday, Friday you get all in on Friday we tape it on Thursday and on Wednesday we drop this week in AI. Why did I start this podcast? AI is moving so quickly that I need to every week meet the people who are building the future and understand what they're building and talk about the week's news. This is how I get smarter as an investor, as a human being on planet Earth and we are off to the races. This is episode 9 and we do enhanced show notes on this program.

[00:01:30]What does it mean enhanced show notes?

[00:01:31]We give a lot of the details that you might write in your book with your Zebra G750 pen, uh, if you're using the same as me. All those notes that you might take on your Pluto recorder, etc. We've already done those and we put them in the show notes. Lots of links so you can get smarter. Podcasts with smart people is the way to get smart very quick. This is your weekly assignment. Every Wednesday, listen to the pod. Take notes, look at the show notes. And we have two amazing guests today. Lynn Chow is with us. She's the co-founder and CEO of Fireworks AI. They are a frontier inference platform, uh, cloud platform for developers to run, fine-tune, and scale open-source generative AI models for production use. Uh, and they process tens of trillions of tokens per day. Uh, and Lynn, uh, you worked at Meta for, uh, a number of years, maybe close to a decade. uh seven years. Tell us seven years. Okay. And um >> you worked on um the inference layer or you built an inference layer that lets companies run llama. Tell us a little bit about why you built this, why it's important and and who the customers are.

[00:02:40]>> First of all, fun fact about uh fireworks. Uh we started with seven co-founders and uh >> very big funding team. uh after three and a half years um we are all working in the company uh pushing really cutting edge technologies we know each other from meta uh that's a time when we joined mada is finishing mobile first transition moving its application from the Facebook the messenger from desktop to mobile it was a huge bat at that time if you remember uh that's the time I was having my first iPhone my first app on iPhone was a flashlight. Uh so that's how kind of um early it was.

[00:03:22]>> It's a long time ago.

[00:03:23]>> It was a long time ago. Um and that was a really successful transition and interesting thing is um it significantly pro uh pushed the product engagement to the next level and from there generate a lot of data interesting data and that data became the fuel of AI. We joined we all joined during that time and uh we were bootstrapping AI infrastructure from ground up when there's no AI hardware everything is running on CPU we were running tiny ML algorithms on CPU there's no AI software there's no AI team so it has been fun almost like working as a startup founded by Meta inside Meta uh we build AI infrastructure from ground up powering both massive massive training and the massive inference. At the same time, we also built PyTorch. Uh PyTorch is the now the dominating AI framework um and took us many many years to get to today's stage. Especially now when it comes to geni almost all models are written in PyTorch and deploying production in PyTorch. We started company because we saw after a few years working at Meta the whole entire industry is also moving to AI first.

[00:04:37]Many other companies reach out to PyTorch team asking us for help because um they want to do AI first transition but there's no AI hardware there's no AI software there's no AI team and we know exactly how to solve that problem for the industry and that's why uh we got started >> so explain the product who is the customer that you're selling into right now and what does the product do for them >> we sell to so-called AI practitioners um so the definition is very interesting it started as um you know the startup founders, the CTO's um the machine learning engineering team and now the definition of AI practitioner is actually much broader.

[00:05:20]Um and so that's kind of really fun part uh of of the um of development especially starting this year as we heavily focus on developing tuned models and the fast and cost efficient inference um of those uh tune model to get to frontier level quality speed and cost.

[00:05:41]It becomes the quality becomes so good.

[00:05:44]We expand our ideal customer profile beyond the techsavvy tech focused people to many other professionals. For example, our head of fness use fireworks to massage her spreadsheet and do finance forecasting in the planning. Um, our legal team.

[00:06:03]>> So, who does it compete with? Is it just for like running a frontier model and I would just replace claude or I would replace my usage of perplexity or grock or open AAI with this as a rank and file employee at an enterprise company or is it for you know more on the developer side and APIs?

[00:06:22]>> There's multiple reasons. Uh first of all when you hit the product market fit and you need to scale your product to massive scale millions of consumers or billions of consumers then you need to become real time in your interactiveness especially in the agentic world and second is to scale quickly to your to all your customers without uh bankruptcy without running to bankruptcy that's literally is a problem today. Uh so we help them control their cost but also get to real time while getting to frontier quality. So so that's kind of where we operate is we focus >> so if somebody like Uber or Door Dash some of your companies if they want to test and start using frontier models well they're using frontier models from the large language model companies the proprietary ones but if they want to do open source ones if they want to try Quinn if they want to try Kimmy if they want to try gamma um they would use fireworks. Yeah, >> there are also many other great open model like Neotron. Yeah. And uh Mistro and many others.

[00:07:24]>> And what's their motivation? Is it to save money? Is it to not give their data to Open Eye and Sam Waltman which obviously has some reputation issues at the current moment that are quite acute and I keep hearing from enterprises like oo we're very concerned about our privacy. We're very concerned about proprietary data going into say an open AI and then they would use that for reinforcement learning. Obviously they say they won't but people still have that concern. Yeah, >> it's kind of deeper um than you know just not trusting a company. Uh so I I think let's take a look at the world's data right think about intelligence as a reflection of uh the collection of data.

[00:08:08]>> The data that goes into foundation models are the public internet and the labeling companies label data. That's the data at essence go going go going to the foundation models it is actually a very small fraction of war data less than 5%.

[00:08:24]Majority of wars data more than 95% are the private data locked inside application locked inside enterprises and those private data will never get shared with uh with venture labs to train a foundation model because those private data are individual companies IP.

[00:08:43]>> Yeah. So, so then if you look at that it's very interesting majority of the intelligence reflected in the private data were not being activated in represented in the uh foundation model.

[00:08:55]So we want to work on the next phase of frontier intelligence which is activating the private data.

[00:09:01]>> Got it.

[00:09:01]>> Customize and tune the model and to make it smarter. And do they do that customization themselves or are they relying on you to send you know uh forward deployed you know enterprise uh developers into their enterprise and like work with them on fine-tuning and do they have their own instance then on their own hardware your hardware how does that go down >> at early phase of a new technology adoption it goes through waves the first adopters are all hackers power users uh so they want to control everything because they are capable of diving into very deep part of the technical stack.

[00:09:38]So we have the lowest level of abstraction for them to control everything. For example, cursor recently re released uh composer 2 uh and that's their own tune model on the open model and they use our lowest level abstraction because they're very capable of driving all kind of uh tweaking and adjustment. And the next level objection is we package a lot of defaults but we still give meaningful um parameters for a machine learning team who are comfortable in massaging the training process um and move forward. So it's kind of media level and the highest level are the application developers.

[00:10:17]They do not have deep a expertise where they can easily tune by many things being automated. So those three level obstructions are the um platform interaction we provide and we expect adoption will go over uh time more more towards the upper level.

[00:10:34]>> Let me introduce our second guest today and then we'll get into all the news of the week. Uh Demi Goa is here and she is the co-founder and CEO of PA agents for creative. They launched back in uh they launched AI selves back in February of 2026. was a persistent digital twin uh that would learn my voice, my style, my personality for better or worse. Um and then uh you've evolved a bit since then.

[00:11:00]Uh tell us a little bit about what you're building Demi and who is the customer and why are they buying it?

[00:11:06]>> So we started a company around like two 11 more two or 10 years ago. So um you know something that's I'm always you know very excited about you know personally I'm always engineer. So I've been starting coding for like since since elementary school but at the same time like I also you know really wish I can be an artist and really you know and the the goal for the company democra like how do we help more people to be able to stuff or to enable more creativity from like more regular people. So we started a company by building you know um a web tool for to help people to create videos uh and we we explored you know different direction like know and then we we realized like for the the web tool it's still mostly for proumers and then it's still very hard for you know like a lot our team members cannot even use the web tool to to create videos so it's really hard to promp and to edit and blah blah blah and we explore like different interfaces we explore wo and such and then we we found out that actually maybe the best interface and the most most accessible interface for people to actually create stuff whether it's create videos or even create you know um uh animation uh it's um through a humanized agent like by talking like more like a humanized interface where you don't you don't actually need to like um like need to learn all the complicated UI or all the editing skill or all the prompting skill um you just do it just like you're talking to a human which is like more like a humanized agent Um so that's why we recently pivoted to um focus on you know using a more more humanized agent for people to be able to create stuff to create videos create uh whether it's like social media videos or films or um or even potentially create um you know um uh like design poster design or or uh or like short film short drama or like vlog or whatever. So, a creative person, they're an influencer on Instagram. They want to start creating versions of themselves or content. It knows their persona and it just makes it. Do we have a a video here? I understand we might have a a demo video and uh we could show it and then you could talk over it.

[00:13:14]>> I want to clarify a little bit. It's it doesn't have to be I I know there the the like the the ex may be a little confusing. It doesn't have to be like a digital to it doesn't have the user influencer. It's really just anyone who need to uh to create stuff uh create videos create create multimedia uh uh artifacts. So yeah we can play video. So this is an example of like you know um and yeah so yeah the the motivation we talk about which is really like you know um what is the best interface for people to to have more a more accessible way for them to create create stuff right to uh to create videos create image or whatever creation it is we we realized actually the best interface we try web mobile the best interface is actually through a humanized agent just like you're talking to We're like it's like a creative assistant, right? So, we're like telling the agent what to do. The agent will brief you. Telling a human what what you want and they'll brief you. You can even like it's more hum it's very humanized that you can even like video call agent. You will like a humanized agent which will like even like screen share what's working on.

[00:14:24]It's like really conversational really really just like you're working with a human. Um and yeah, so that's >> so you can create an image. I could see like some of the personas have different animation styles. One of them is a little anime, one of them is a little Disney Pixar. You create that avatar. Uh you create that representation of either yourself or just uh any persona and then it goes on a Zoom call or books a meeting and is like a customer support agent, a sales agent or is it for entertainment or you're just going to let the users decide what they use these agents for?

[00:14:55]>> Yeah, our primary focus is do about creation. Um that's kind of where we started. So it's really to me it's it's more a evolution of the interface of how people should be able to create stuff.

[00:15:06]So the primary use case about creating vers vlog or creating um like short film or creating um like yeah like dancing video or creating singing video or creating short drama or creating ads uh like like you know maybe even beyond im videos or maybe like even like any marketing materials like poster design or or uh slides or whatever but it's it's really just for uh we we just realized that the best interface for people to create stuff uh is not actually a web web tool. It's actually through maybe not even like a commonly like like a commonly perceived agent, but really through like a humanized agent like you feel like you're talking to a person. You can even video call the person. You can tell them what to do. You can screen share what you're working on. They can also screenshot what you're working on and um through that just like you're like hiring another human. All right, so let's get through our docket. The first story we have here is a paper that came out. This is uh the AI layoff trap paper. And here's your summary. Uh if AI displaces human workers faster than the economy economy can reabsorb them, it risks eroding the very consumer demand firms depend on. We show that knowing this is not enough for firms to stop it.

[00:16:26]In a competitive task-based model, demand externalities trap rational firms in an automation arms race. displacing workers well beyond what is collectively optimal. The resulting loss harms both workers and firm owners. More competition and better AI amplify the excess. Wage adjustments and free entry cannot eliminate it. Neither can capital income taxes, worker equity participation, universal basic uh income etc etc. So essentially a prisoner's dilemma. If we don't cooperate then everybody loses. And the proposed fix these uh UPAN and Boston University um academics came up with was a robo tax. Charge companies for the demand they're destroying. Uh and so I guess they call this a piggoian automation tax. A proposed policy designated to address the negative externalities of labor displacement caused by auto artificial intellig artificial intelligence and automation. Uh, and of course there's been a lot of debate about this issue because Block cut half their employees. People say, you know, Dorsey, Jack Dorsey said it was about AI. Other people said it was just a convenient um excuse to do it, doing more with less, and that we had maybe uh overhiring historically in the tech industry. Both of these things might actually uh be true. Uh, and CFOs are probably saying cuts will be nine times bigger than what's being reported.

[00:17:52]That's from a Fortune take. Uh HBR's take companies are firing based on AI's potential, not what you can actually do yet. So Lynn, what are you seeing through your customers and how they're deploying? Is it creating more jobs? Uh, and are people um or is your belief that we're going to see more people displaced or we're going to see, you know, more people hired because people are going to start more companies, people are going to find more problems to solve and they'll be inspired by the technology. This is the debate of the of 2026. Clearly, >> absolutely. Yes. Um, so one thing that's clear to us is we have seen a boom of ideiation to production. Never have been faster. Um in the past a prototype that can reach into many hands of the people uh to test product market fit it will take multiple quarters and now just take multiple days. Um and that empower a lot of people have great design taste uh and great ideas. They they um they can access so many tools for them to um realize their dream quickly without a large team without creating without have the organizational skill set to to hire people to kind of organize them to deliver tasks together because now you can orchestrate a a fleet agent to solve those problems collaboratively. So I think uh the creativity is off the chart right now and we do see a lot of a ton of startup with brilliant ideas. So the most fun operating in our space is we are the AI index. We see a lot of great use cases, crazy ideas emerge on top of build on top of fireworks and a lot of experimentation and many of them just take off um like um you know it just es escaped the gravity and uh reach a lot of people uh quickly. So so that's a fun part. With that said, our company, we only have 150 people. Uh, and uh, like you mentioned earlier, we're tens of trillion um, tokens a day, but we probably that it's extremely high traffic just as a reference. Um, based on our understanding, we may be mistaken. Um, that traffic is bigger than open eyes API traffic. So um so we reached that um as a reflection of the direction we're heading towards is we happily operate on customized model that is not offtheshelf model inference at all. It is using private data to tune the model and then bring the best quality customized focus for your application and speed and cost to your application. Um so so that's what we see obviously um that is what we see on the starland. We have a many enterprise customers especially digital natives.

[00:20:48]Interestingly those dig natives are they were startup decades ago um one decade two decades ago and they are the survivors they are the winners um winning over the competition um they have a huge amount of traffic and uh they have a lot of people and they're rethink they're rethinking how to convert themsel into AI native company again how to reinvent themsel so as a as part of process of in this reinvention it's not just a product uh reinvention. It's also organizational reinvention because in order to survive this wave of heavy competition, they have to revive their velocity and that velocity got buried uh through layers and layers of hierarchy in the organization. Um and uh many companies start to kind of reduce that layer especially mid management um where for example uh from my experience in the past um there there were like hierarchy of binary tree organization that kind of probably doesn't make any sense uh these days and now we're talking about a manager uh not just be responsible for seven to 10 people in a layer probably they can handle 20 to 50 people because >> so this is I Think is such a key point, Lynn, is if these tools are so great, one manager can manage two, three times as many people because all of the check-ins, all of the knowledge is already surfaced by AI and you don't have to be a warden or a babysitter of employees, which is what, let's face it, middle management was in a lot of these organizations. You probably saw that acutely, Lynn, at Facebook. Yeah, there was a middle management layer that was coordinating and writing notes and doing meetings and doing standups. In your experience, now that's all automated, correct? By AI.

[00:22:43]>> Yeah. So, I think by and large it can be automated because as you said, um information is much more discoverable and because of that, so imagine a manager's job in the past was collecting information and uh relay that and make sure everyone's aligned. But these can be much more efficiently done. Uh for example, at fireworks uh we don't do a lot of one-on- ones because a lot of context is shared uh with a group of people. If we make decision, we just quickly make a decision by checking and discuss that and done. Uh because um we are very chatty on Slack. lot of information is on Slack and uh um people we all share similar context and also we can summarize what's happening on Slack per individual's um needs. So that is extremely streamlined. Um that help us stay on high velocity decision making execution. Um and we're very flat uh organization as a as a startup. Uh but interestingly uh I also saw many large public companies start to transition their organization structure flattening out because that smooth out information flow up and down. Um and the reason they can do that is the information discovery, data discovery is much easier. Uh it's not just limiting to individual's work and uh people management. It's also about data scientist for example product analytics, right? So how do you reach a party decision uh and you need to understand a lot of data and in the past we have uh large layers and layers of people trying to do that. But if you mquify your data access um and then you can write an agent you can build agent to be able to extract that information and uh um and summarize it and uh uh synchronize across different fronts and get very precise analysis assessment of the health of the business and >> and interestingly Lynn I don't know if you saw the report but Ruof Bofa and Jack did a podcast this week and they've and they did a blog post two weeks ago Ruof from Sequoia uh and Jack Dorsey is basically saying he wants all 6,000 block employees to report directly to him and that with AI he can manage that.

[00:25:00]It that's a fantastical vision, but it's not ridiculous if you're a hardworking CEO and you work 12 hours a day and every 30 minutes you have 24 segments a day to work. 24 segments of 20 workers.

[00:25:17]you start doing the math on that 400 workers per segments you know information you know uh coming in you could actually make it work you could actually get through in 10 days 15 days every single workers output uh Demi what's your take on you know the combination of new structures and then what's happening in old organizations and then maybe even how you're running your organization as an AI first organization >> what I really believe the human value in the future will come from the judgment and taste uh or personality or like unique identity of the human. Uh I think it's I I actually feel like it's ultimately not not even like um necessarily the most most productive thing but but but like what it really value for human is like for example like um you you um like you you the only thing that human matter is you have your own judgment right like because it's you because it's your company for example or like you know you what you're doing so and then you just you can just have one AI agent that really reflects your judgment your taste, your your thinking style and your decision- making style or >> or like whatever like your personality and then that agent can just you know either can just do everything for execute everything for you right whether through or orchestra a lot of agents together or through um just like working on its own so I think in the future um that's kind of what Jack Dorso is doing is is he's trying to really amplifying his own judgment and taste u for for the company uh and I think future um we'll see more than that and the value is more about and there will be more people who are going to create their own unique agent that really reflects their taste and per judgment and then uh there will be more freelancers in the future and >> and many more freelancers who can just come in and be that human in the loop and maybe even work with an agent.

[00:27:12]Here's a clip of um uh Peter from OpenClaw talking about you need to have a human right now in the loop with agents. That's definitely been my first inexperience because it doesn't have taste perfectly yet. Uh but here it is.

[00:27:28]>> It can create code and run all night and then you have like the ultimate slop because what those what those agents don't really do yet is have taste. They they are really they are spiky smart and and and they're really good at things.

[00:27:44]But if you don't if you don't navigate them well, if you don't have a vision of what you're going to build, it's still going to be slop. If you don't ask the right questions, >> it's still going to be slop. When I start a project, I have like this this very rough idea what it could be. And as I built it and as I play with it and as I darously feel it, I my my vision gets more clear and like I I get like I try out things some things don't work and I evolve my idea into into what it will become and that's that's like my next prompt depends on what I see and feel and think about the current state of the project. Yeah. Yeah, >> but if you try to put everything into this kind of like human machine loop and then I don't know how something good can come out without having having feelings in the loop almost like like taste, >> right? So, Demi, I guess that is a critically important thing for when you're running your company and people are creating agents. Doing everything up front seems like the right idea. Hey, I want to create this website. I want to create this piece of content. I needed to be entertaining. I wanted to have these themes. I wanted to accomplish these tasks. But you can only kind of um frontload so much in the prompt or so much in the skill. And it's really about I think iterating on the scale over and over and over again and pointing the agent in the right direction so that it does the task for you. Is that your experience as well, Demi?

[00:29:13]>> Yeah. The reason we transition from like a web interface or prompting or mobile to agent is because we realize in the future what matter is you train your agent to have your own taste basically.

[00:29:25]So by using your agent or like iterating your agent by giving feedback agent you're gradually making your agent to have your own unique taste and your agent can do a lot more things than you.

[00:29:36]So it's it's almost like the goal is not about like okay you're creating individual website. Um but actually about like you're using your agent maybe whether it's just telling the agent what the taste should be or just the throughout like using your agent to create website you're like training the agent to understand your taste and then your agent will in the future when your agent have your taste it can just create infinite websites right >> so I think what matters is really just to um that's why we transition from the just like prompt interface to more like age interface >> Lynn this seems to be I think a hardfought lesson I thought setting up my open claw and then saying I want to solve this specific task on reporting was a one-time scale. I create it. I run the crown job. I never have to touch it again. It turns out that's wrong. You you have to be a bit more interactive.

[00:30:25]And a lot of times I'm finding these agents drift from what I told them explicitly to do. And because they're so sickopantic or uh they're they're very inconsistent. I I don't find they have the consistency. So maybe some thoughts on when these will anticipate a little bit better and when they'll be more consistent, Lynn, in your >> So I I I have a I have a lot of conversation with my daughter. Um she's in high school about AI. So it's very interesting. I'm surprised. I feel like her generation will be AI native. They grow up uh with AI around them, all the tools. they will just kind of be very deeply embedded in using those tools natively.

[00:31:13]Uh but I'm surprised her reaction is um they are like Gen Z's are cynical about AI because they don't feel it's authentic.

[00:31:23]>> Ah >> they don't feel it's creative. they feel those are all the repetitive mundane things. If you use AI is not being thought highly of for example they have this um school magazine and if you use it to generate uh pictures they like it's better if you draw it yourself and kind of shows authenticity. So I I I feel like AI is getting really really good actually. If we look at a gener images, it's actually really good because it's it's trained using uh the the human intelligence and and kind of it it's able to simulate that. But I feel like we as a human as part of our soul, we we we need we need a creativity as a satisfaction uh fundamentally. I I'm wondering if there's a future that we can integrate deeply always have human in the loop to kind of to be part of the soul of of this new world building. I think that would be fascinating. Uh and I like the demo from uh from Demi the new um PA agents. I I feel like if I can embed my uh part of the creativity into this uh this humanoids agent that can uh represent me and uh give a little bit surprise here and there because there there's a little bit of impromptu of how we react and so on and give it that um it would be really fun.

[00:33:00]>> Yeah, it's it yeah go ahead. Oh yeah, really agree about like the creativity and also the human the loop perspective and that's kind of what we are really like leaning towards because I really think right now the reason people people are really treating I think we need to change our mental model about what AI is. Uh I think a lot of people in the Silicon Valley are really treating agents as a tool and as a productivity tool and to people choose who they work with not for only for productivity but also for you know they just enjoy working with this person. So that's why we really feel like the mental model of the agent should not be a tool but actually just be another human and then you should really treat like having your own agent as having your own child. So it's a constant feedback and iteration.

[00:33:49]So human should always be in the loop.

[00:33:51]It's not like you one click everything done. It's more about okay you're raising a child you're like constantly teaching it and gradually you will grow up and you like can do things for you.

[00:34:01]Um and it's more this like like iteration process with your ch children and it's also this emotional attachment beyond productivity.

[00:34:09]>> Yeah. And if we just recap this previous segment, Lynn, um the the take on uh how organizations are changing, well, if even if people get laid off and organizations become smaller and flatter, there's been this Cambrian explosion, I guess, which happened hundreds of millions of years ago, uh you know, where life just suddenly emerged and in a very violent many people competing because of oxygen and you know, this perfect uh ecological soup that occurred. I think that's kind of what's happening now with startups or just individuals being able to say, "Hey, I could just make a startup in a weekend, test it next week." As you were pointing out, it used to be a two or three/arter journey to get your product out and test it and you know, beta testers and then you know, close beta and then open beta, etc. Now you're talking about doing that in three or four days potentially with just one or two people. That means many more ideas and you know um we could see instead of 10 20,000 startups getting funded every year you could maybe have a 10020,000 or a million or two million created and the startup then doesn't have to be venture backable and it could just be enough to pay somebody's salary or maybe even half of their Facebook salary. If they were making $300,000 at Facebook, they might be completely happy to make 150,000 but live, you know, at a ski resort half the year and live by the beach half a year and, you know, check out a bit. Yeah.

[00:35:41]Lynn.

[00:35:42]>> Yeah. So that's what I mean is I feel ultimately I see if this continue to play out um we will I'm actually pretty optimistic about the future is uh we human will focus on the most creative part because that's how we evolve over thousands so many like tens of thousands of years is we find creative way um to form new structure um in new invention um and new technology uh new economics all the time um and every time it's there's some wave powering it so uh for us to leap forward um but we always resort back to being creative uh we never stand still and circling on the same spot I feel like AI is actually um if we do it right it's going to power the next level creativity. I couldn't imagine what will come out of it. Uh we may be able to do spo space exploration much faster. We may be able to reach the uh planet light years away much sooner.

[00:36:52]Um and um and it's just kind of I think our limitation is our uh imagination is our limitation. Uh so I that I'm very optimistic towards that direction.

[00:37:05]>> Yeah.

[00:37:05]>> But in in >> continue. Yeah.

[00:37:08]>> Yeah. In the face of doing that uh even step by step um before we we were like hobbyist is hobbyist they they tinker and just as a hobby right so um but now we're like hobbyist could be the next phase of uh inventors um because the weekend a weekend project uh could really hit uh something fundamental fundamentally deep and hit our biggest pain point because everyone is now empowered to to create to think to imagine uh with their tools. So um so that's the part I I really I really believe is the combination of human creativity. I do not think I do not believe that we replaced um and uh our brain is only a few percentage activated also at the same time there's a lot more we can derive out of it. Uh so we actually as a platform well frontier inference platform as a platform we pay a lot more attention to hobbyist these days because uh we give them the tools for them to uh test uh invent experiment and once they hit something interesting they can quickly scale and they do not need to worry about scaling because scaling is a complex system problem. uh especially scaling um your deployment of your tune model. First of all, how you tune a model um it involves a large amount of GPU fleet or a lot of kind of data tinkling, a lot of parameter to set and experiment and once we want to scale, you want to scale globally across many regions with low uh latency, high reliability, robustness and a lot of people using it and so on. So how are those uh open source models doing on a you know how many months behind in human years not dog years or AI years just actual human years uh because everybody knows dog years are seven years to a human year I I would say a month is seven u months each month is a year in AI time right now so are they six months behind >> or three months behind nine months hop behind Opus 4.6 whatever.

[00:39:29]>> Yeah. So it's a very interesting race that's happening right now. Uh but I want to put them against each other. I just to us we we saw like every two weeks or sometime every week there's a new model released open close doesn't matter and they're oscillating with each other on the leaderboard. Whenever things are oscillating it's a clear signal they're converging. Um so last year I think like one big uh sticker shock is last year's Deepseek um Deepseek V3 release and that's the first time open model got very close to um Frontier Labs model and since then Frontier Labs continue to uh leap forward and then open model catching up.

[00:40:10]So it's kind of um but overall we see the quality gap start to converge and that's great news for for application who has a lot of data that means they tune if they can activate their private data they can actually forward and really get to next level of intelligence specific design for their application.

[00:40:31]Um and uh and we see since like since last year we start to see a lot of adoption of this customization especially led by more frontier uh thinkers more um application they are pushing the boundaries.

[00:40:47]>> So they're getting closer but still frontier models are ahead. So how far ahead I guess is the the question I'm asking and still people so I get it's oscillating. There is definitely convergence but one group seems to be leading that's the frontier proprietary models how far behind are the open source models in your estimation and then it depending on how many months or years you think they're behind what is the road map to this convergence >> for complex task uh for example most complicated hey uh if you want to generate the most intelligent agent to um to build a distributed um system P2P P uh system for eventual consistency or whatever. Uh that is really hard. So um I think frontier model is absolutely leading there. Um they're probably six month ahead, six months to one year ahead. Um but for many day-to-day task for example um managing spreadsheet um managing calendar um having a router or classifier of um um doing some kind of routing logic um or um doing um writing uh improvement or um doing some good very good kind of document processing uh and so There are many tasks varying very with uh less complexity but actually cover a lot of our day-to-day they're very close I would say even on par uh to something >> okay so simple tasks couple of months on par complex tasks six to 12 months behind I think would be what I'm interpreting uh from your comments there which I think is super helpful for folks because you do have this issue Demi with the cost of these models and people with agents seem to very quickly and I I don't know this is your lived experience right now >> I would say but uh this is offtheshelf model quality but with our private data we have so many cases after tuning on the complex task it can be on point or even better than frontier labs model >> demi what's your experience with your customers when they start embracing this technology what is their usage profile in terms of token usage and how does it change when somebody goes from you know using like a search engine or a researcher and just asking one question versus hey, I'm going to give you a complex task. I'm going to give you uh you know an agent uh an agentic kind of uh existence, dare I use the word existence, but we're going to will this agent, this replicant to exist and I'm going to interact with it every day. How does that change? How does that token usage change on a multiple?

[00:43:36]>> For sure. I think we we do do see that really depends on like what kind of like models we're using. We're trying to maybe get more user more customization because it will really value a lot based on like to L's point like if it's open source like much cheaper versus like the more the frontier lab it's more uh expensive like it could be up to like 10k we have like user who have like 10k per months or something >> 120k a year to empower their agent and do they make back 120k a year or are they just they're so their their business is so great they don't mind losing 120 to be on cutting edge.

[00:44:12]>> I think the the reason is we should really not compete like really like treat obviously there the agent when the model is better where you always like open source or like with tuning what link says you will probably be cheaper over time but but also just generally I feel like we should not really treat agent as a tool that okay you cannot pay like 10k per month for a tool but you should really treat agent as a human so whether it's your child your race or or your employee right your assistant you're like you're it's like it's like whether you hire hire a junior creative assistant or you use for agent or like you know you either you raise your own pat or you use agent. So it's like that should be the comparison like people spend a lot like it cost a lot to have a junior creator assistant or it might cost people are willing to pay a lot for their pack. So it's not really a compar a fair comparison your comparison with agent was like a um like a tool but but we should really think about agent as like a a human or like a life form and you're like you know you're you know you're you're like um keeping it right whether it's you're hiring or or or like raising it. In technology, there are some folks who were early on in in the 70s and 80s who very much looked at what are the hobbyists doing because the hobbyists quickly become if they figure something out the entrepreneurs. And then they would look for the toys, the tinkering and the toys, and those would become the tools and the services in the future. And so a lot of what we're working on right now feels like toys. It feels like tinkering. It feels like a hobbyist and creatives and agents certainly fills that kind of uh description, but then they become actual real tools and services in the future.

[00:45:57]And it's a great way to either build a company or invest money in companies is to just watch what are people tinkering with, what are the tools and that's why I became obsessed with OpenClaw the moment I saw it. That's why I became obsessed with Tao and the Bit Tensor subnets. Um or even app development and you know content uh in the form of apps back during that revolution. Lynn, when you were at Facebook, it just was like oh the flashlight. Oh, a little Flappy Birds became like you know critical uh a critical economy very quickly in in the the future. Lynn, question for you.

[00:46:32]I I believe you went to school um in uh Shanghai. Yeah. Um yeah and >> well >> that's >> yeah so that's the second I think city in terms of the density of large language models Beijing is number one still in terms of where all the computer scientists are working on large language then maybe Beijing and Shenzhen I guess because of the hardware footprint what is the movement there like I'm sure you have friends and family and colleagues that you went to school with what is their perception of the race in terms of open source frontier models and just overall what this technology will do for society.

[00:47:15]>> My understanding is um I think because of population consumerfacing product are much more popular >> um and uh much more polished. So many of the models has been heavily used to power newer consumerf facing experiences. Um for example I heard open cloud at um at China is is becoming very popular and various different company uh offer open cloud as hosted managed service uh and get a lot of traction. Uh so so that's kind of I I think um I think it's just because of the population advantage uh and the focus is more like consumerf facing product um and how to bake those um new way of surfacing user experiences uh through LM and more like multimodal gen models not just LMS >> um are the kind of primary focus >> interesting and what is the general perception of AI I we just saw some studies come out. People in America obviously have a very negative view of AI. We had this New Yorker story come out last week about Sam Alman. It was quite negative. Uh and then we had these horrible attacks on his home, a molotov cocktail, and then somebody shot it.

[00:48:40]This is absolutely terrible. But there's an incredibly negative perception of AI in America today. And then in China, my understanding is it's extremely positive. people think that this is going to be amazing for society. Is is that true? And why in in your estimation, Lynn?

[00:48:56]>> I do not get uh I guess I I'm living in the Silicon Valley bubble.

[00:49:01]>> Um I feel extreme optimism here about AI. Um but my observation is I feel like China is probably more advanced on the robotic side because um it's day-to-day.

[00:49:15]It's in like delivery, food delivery or robots. Um and uh people can order ice cream. It got delivered in a couple minutes through uh via robots. Uh and that's kind of the limit test of how fast things are is you order ice cream, it's still frozen.

[00:49:32]>> Is your ice cream melting or not? It's the ice cream test, >> right? Um so um that's my sense.

[00:49:39]Obviously, robotics habit depends on AI.

[00:49:42]Um and uh that's just one reflection uh of uh the consumerf facing focus >> and um um yeah I I again on the AI sentiment uh I shared earlier even from my daughter so they she told me like they are Gen Z is uh cynical I mean she also grew up in Sic Valley. Maybe the Silicon Valley teenagers are cynical about AI and they they want to preserve the or authenticity of creativity uh from their heart as their identity. Uh I I do think that's a human need we need to address. Uh we shouldn't lose that along the way. So I think it's more a societal um homework for us to figure out together is you know as we co-evolve with this new technology uh and how to focus on maximize uh our um our fundamental needs as a human to express ourself um over time. So I I think that's just we're at such early stage of uh this um revolution and and we're going to figure it out. I'm very optimistic about it. want to go on is what like links to me because that's actually something that we're trying to really solve on is really like what we really believe is I think Silicon Valley is really crazy about how productive the agent are and how it's like replacing work blah blah blah but something we really care about is like I think probably people a lot of people are very scared that okay the agent are replacing my job blah blah blah right and something that we really care about is like enable people to create their own unique agent um like their own agent that they could customize and it agent should not feel like a tool the co-hearted thing, the productivity thing. It should really feel like a a child you race. Uh, and it should really feel like um like um like you are uh your own like you you own it and it's your own agent and it should really be like working for you. uh and it should really be something beyond just productivity and there's also the emotional connection because like you know it's like there's like identity creation of it and it's I think it's very important actually beyond this the productivity value of agent obviously your agent will be productive has utility but like the personality or the human perspective and identity creation on agent I think is also very important and I think it's almost a sense of it's almost a form of self-expression in some sense where uh you could imagine like okay like um like now Picasso's art really inspire us but maybe a 100 years later like your your own agent which with unique personality and unique voice and face really inspire people like really touch people's heart 100 years later uh and like among other corporate agent >> yeah I don't if you guys saw this this was one of the you know Shenzhen comes out with all these incredible uh toys uh and again back toys becoming tools. Um, this one is called Bubble Pal. Um, and I don't know if you've seen it, but it kind of went viral last year. Uh, I just gave the link to the team. They'll pull it up here. Essentially, you put this device. It's like a little Alexa sort of uh, digital assistant, but you put it on any toy and then the toy gets a persona and it gets a personality and then here's here's a quick video. We'll play with a little bit of sound. Um, but it's definitely having a lot more playfulness and then beauty filters and AI filters and AI chat bots and making short videos with AI becoming incredibly popular in China. Um, with consumers. So, here it is. Um, you you pick any um here's a little girl getting left out of playing soccer with her friends and then an little girl who wants to be an artist. And you bring your toy, whatever toy your kid has. Like I could give this to one of my daughters and their teddy bear. And then I just strap this on and it creates a persona for the toy and it becomes like a best friend, an imaginary best friend. But it actually does talk back to you. I'm not sure if this is dystopian or just incredibly engaging. Your thoughts, Lynn?

[00:53:56]>> I think we all we all when we were little, we all have an imaginary friend.

[00:54:01]>> Yeah.

[00:54:01]>> Somewhere. I think it's fulfilling our need of having this imaginary friend in some form. Um I I I could see my, you know, it goes back 10 15 years. My my kids would would love playing with it.

[00:54:17]>> Pretty pretty interesting stuff. I think it relates a lot to what you're doing at Pika, right? Is uh this uh this concept of creating >> some kind of relationship with digital beings. Yeah.

[00:54:30]>> Yeah. I I think it's for sure like obviously it's I mean agent is not just a toy in the sense that like everyone will have agent in the future, right?

[00:54:38]It's we all like kind of feel like agent is not only the next interface for web and mobile but actually replace the next computer because agent like has storage and compute and everything, right? So everyone have agent but it's really about like okay do you want a generic agent or do you want your own agent that has more humanlike and more emot has more emotional beyond just productivity.

[00:54:57]>> Yeah. And here's uh just to put a final end cap on it. Here's the US perceptions of AI societal impact. These are the general public versus experts. In other words, the bubble you responded you you referenced Lynn. We're we're living in the AI bubble. We see all these incredible things uh as possible. Here's US adults versus AI experts. Here's the chart. Um, and if we look at uh medical care, if you're an expert, if you're in the industry, you think it's 84% um is your percentage uh that AI will have a positive impact over the next 20 years. The public 44%. K through 12 education, elementary school, 61% people in our industry think it's going to have a great impact. 24% of people who are not in the industry, just US adults helping people do their jobs. This one is the one we all see in the economy.

[00:55:51]73% uh and 69% of people in the industry think it'll help people do their jobs and will help the economy. But 23% and 21 for the public um and uh for personal relationships, people do not think this is going to help personal relationships at all. uh 22% of uh AI experts, one in five of us believe this will help your relationship.

[00:56:17]7% of the public thinks so we have some alignment here for the elections and for news and relationships pretty good consensus it's not going to go well but for everything else we we do have this massive swing uh your thoughts Lynn just generally looking at this maybe how do we change the perception I guess is the bigger issue we we if we accept that this is the case >> so for example in medical care right so we have customer we have many customer from medical care they are doing amazing things um I now see doctor um and by default they will ask me if they got my permission to turn on recording because they can use medical scriber to automatically take notes. uh we power uh another medical company that they are doing building preventive care software and as in uh as I uh going through um you know see my doctor and then they quickly can pull out this software can pull out my uh medical history and suggest to doctor what kind of preventive check I need to have and then they will give me a list I will work on that. The idea is great because um they putting in more preventive uh either exercise or tracks or exams and what not is going to help make us healthier and reduce the also the medical bill. Uh so they build their business out of that.

[00:57:40]But the the barrier is actually software integration with hundreds of um patients record systems where there's no standard how to integrate. uh and they want to roll out this great preventive care idea, they are going to hit the system record barrier. However, AI is here to the rescue because there is one universal interface >> that is screen >> screen is one universal interface. So it's very interesting right in the past few decades we have we have per make the perfect interaction between human and computer HDI through various different kind of software development >> in order for us to kind of um make the long horizon human-driven workflow more automated then the next phase will be automate HCI so um so that just already started piloting through like authentic creativity from people in your chart the first row of expert working in medical care uh system that's why I I I think that makes sense they are very optimistic about the impact of AI um I just feel the more people they see the outcome of AI the more we share the stories uh share like um day-to-day stories uh what are the benefits and uh um And you know we can get out of the end product. Yeah. AI.

[00:59:11]>> I think our our industry is doing a terrible job of this. We have to show all the wins for people. Yeah. Demi. And maybe a little less of the fear that everybody's losing their jobs and uh it's the end of the world. There's so much joy and fun and cost savings and life extension that could come from this. So where do you think the industry is getting it right? Where do you think the industry needs to improve? Demi, >> my feeling is there is like really a differentiation between like pure like maybe like productivity AI and versus like some AI that's more like diverse or like more uniquely yours in some sense which is like maybe not not even have to be the most productive but but it's uniquely your own agent and and maybe like more human more creative or emotion whatever it is I think what Silicon Valley is really care about is like okay is the agent productive is it smart enough or is it going to solve this productive issue like enterprise work or developer or developer issue and I think something people like and then like that really the narrative really scares like regular people because it feels like okay like you know like okay like human like agents are really like AI really real place in my job. Uh and I think something that is really missing is like um how do we like use AI to enable people to create their own AI and to empower themsel instead of you know use like how the angle of like okay using your AI to replace versus like empower yourself uh instead of like you know like a like you know AI is like replace like for productivity and really um like you know really um like really good and I actually think what what is like to your question about it is like I feel like there will be two type agent like like I mentioned one is more like your own agent which is more diverse it's really for diversity like you're really really unique your unique taste and judgment even though it might not be the smartest judgment and taste and p personality and there's another agent which is like purely for productivity and actually for productive agent I actually think it should have no personality it should be um maybe you should not have human judgment because human are not the most productive people I Because human has flaws like human has emotions, people has like strong personality and human are has ego or whatever right like human and I think like so there is this two kind of AI in my opinion like the the the to have to achieve the best productivity you should have like no human perspect like no human component to it. Um and then there's like the other AI which is more human and more personal like more personality. It's it's really like you know like for diversity is really important for human because we you know it's like it's there is value for human that we we we might not just want like the most productive thing. We might want something that empower us and make us feel good.

[01:02:02]>> So something I think we're like kind of ignoring right now.

[01:02:06]>> Perfect uh segue. Meta introduced um just last week Muse Spark Mus ML MSL's first model purpose-built to prioritize people. Uh so this is uh I guess the first uh language model that Facebook has produced that's closed sourced. This is I guess their frontier model.

[01:02:29]Interestingly, when you look at the specific um examples they gave, and I I had a perplexity here do a little summary of this incredibly long announcement.

[01:02:42]They gave uh a bunch of what I would call domestic uh Mr. Mom or mom focused benefits. So, planning a family trip to Florida. Uh, one agent drafts the itinerary, another complares Orlando versus the Florida Keys. A third finds kid-friendly activities all in parallel.

[01:03:04]So, they're not doing here's the enterprise and how we kill jobs. They're doing here's how we make um, whoever the homemaker is, mom or dad, here's how they're going to uh, do ask health questions and take a picture of all the snacks. They keep going to travel. all the snacks at an airport shelf and say which one has the most protein in it. Uh or shopping mode, get ideas of what to wear, how to style a room, or what to buy for somebody drawing from inspiration across meta apps. And this is using all of the information on Instagram, WhatsApp, and Facebook.

[01:03:40]Really interesting uh I think very human forward approach. So your thoughts, Lynn, on two things. one meta going uh from open source to closed source here disappointing interesting what are your thoughts and how good were their open source models um and then two their human first approach and if what you think the uh public reaction to it should be >> yeah yeah we hoping meta will open source their avocado model for sure u but I also understand like um their main business is a consumer product and this next generation model need to power their consumer product and uh MA is really good at product model code design very very good at that they have been doing that for decade um and I'm pretty sure this model uh powering uh WhatsApp and messenger and all this um consumerf facing product is not a static model it's going to evolve over time uh it's going to be smarter uh and learn from meta's uh private data uh and so on So that's kind of the the typical pattern I think all application developers in the future all AI native application developers should be able to adopt and today there's not such kind of tool it's a vacuum in the space it should be as easy as uh product analytics it should be as easy as everyone's doing AB testing uh and our dream is kind of give this tool to everyone's hand and it should be turned on by default uh so um I on my side I really hope there's more a lot more US-based open models. I'm uh we work very closely with Nvidia. Nvidia is putting a lot of um focus on driving the Neimotra model. Very very uh bullish about that. Um and uh I think they're constantly improving the model quality and there are many other providers uh in on the US soil on the uh >> tell me about Neotron. It's it's Neotron. Nemo O T R O N. Not a great name. Nemo is a Pixar movie. I No, or Disney movie, I'm not sure. Um, and Tron is a Disney movie as well. Neotron. Two Disney movies put together. Very strange name. Uh, but how good is it and are people using it yet, Lynn?

[01:06:05]>> Uh, so I think it's um it's getting better and better. Uh, so one thing about model training is it h it you cannot jump ahead of time. Uh, you have to go through steps. uh you have to train from a smalls size model, generate synthesized data, train mediumsiz model, use the synthetic data, blend with uh whatever data you have and then generate more synthetic data and then it's so there a kind of process you have to go through um and uh and then during that process and with very complex um post mid-training post training of SFT RL um and then the model becomes a lot better.

[01:06:42]So imagine kind of the the process is like um first you need to build a base IQ like we human when we're born we come with IQ that doesn't change over time but that IQ accumulation takes time of uh of learning basics uh and then once you have base IQ then we human take years to be trained as a domain expert right whether we are a dentist or heart um surgeon or a a lawyer um across the board it's kind of we takes us years to get to certain specialty and that's RL training and RL training is basically narrow down the focus um and uh and kind of really specializ in certain area solving certain kind of problem really well uh so so that just take time um and uh I think I think all these models as long as there is substantial amount of research effort behind that a lot of GPUs uh it's a matter of time they will get there. So, so we >> what do you think about the data drought? Lynn, I'm curious and then we'll go to you Demiier just on where you're sourcing data to make better models.

[01:07:51]>> There will be a combination of everyone's sourcing from public internet. There's no secret sauce.

[01:07:56]Everyone's kind of saying >> and there's nothing left, right? There's nothing left.

[01:07:59]>> Um and again, a lot of data is locked inside the vertical private applications. Uh you just don't have access to and only those data owner has access to. Um and then there's labeling companies label data almost all companies all foundation frontier labs use the same labeling companies. So, so then what is could be differentiating is how they generate inside the data uh and the mixture a blend of the data could be differentiation and how the and even the algorithm of training pre-training u mid training post training are converging I think there's unless um I think we're are overdue across the industry across the research community to have a new model architecture transformer is uh way old usually there every three years there's a leap of new model architecture uh I think this is um probably seven years in the making more than seven years uh so um there could be likely that there there's a leap forward new model architecture it will learn knowledge in complete different way then this phase of convergence model convergence may we may have a breakthrough so far we haven't seen that yet uh so if we stay in current Of course, I think um I think whoever owns data owns unique amount of data is going to >> uh it's going to bring up a a new level of intelligence.

[01:09:26]>> Demi, your thoughts on where people are getting data from or and has that reached your customers yet where they say, "Hey, I want to bring some data or I need you to go find me this data or I want to hire a data labeling data sourcing company." These dark pools of data seem to be the next frontier. uh seem to be the next opportunity. We have an investment in micro one. We had Ali on this week in AI in one of the first uh pilot episodes. So maybe you could talk a little bit about where you think the next data is coming from.

[01:09:55]>> Yeah, I actually feel like kind of what we really believe in the future is people own their like potentially maybe their own their own data in some sense, which is like the when they're using agent is techally like they're like kind of feeding their data to their own agent. Um and and I think and then like they have the ownership of their own agent, right? what's important in the future that differentiate different people is actually your own taste and judgment, right? So, and then you want to use AI to really amplify your judgment taste to maximize your value.

[01:10:25]So, it's almost like you're trying to train your AI with your own taste, which would can be feedback iteration or it can be data and to train your AI to be like to the the taste whatever judgment you have, right? That's like kind of like a data feeding process and then like that it could be like okay I'm I'm a CEO that's kind of what Mark Zuckerberg or like a radally or a lot of people are doing which is like they train agent AI that's like agent that's like themsself and then like they use it to talk to like Jack Dorsey kind of also want to do it right like to do user owner judgment and to really amplify by like managing the company right so and then like probably they want to own their own data and then because that's like the value acrew in the future. So like people people are will probably like that might like in the future when we like don't even like work everyone just going to train their own agent and to inject your own data or case and judgment to your agent and then like the agent will help you to do stuff. So in some sense that like that kind of is also like a maybe in the future like people are like kind of owning their private data in some sense from from that perspective.

[01:11:31]>> Okay. Uh I think that's a great place to start. Two amazing guests this week on the program. Uh, I know you're both or I'm assuming you're both hiring and growing quickly. So, >> Lynn, tell us a little bit about what positions you're hiring for and how folks can get in touch with you directly if they're a genius or where they can go to jo, you know, join the fireworks team.

[01:11:54]>> Yeah, we are enter um exceptionally fast growth phase. Uh so we're hiring across the board um and uh from product engineering to marketing um sales um and DNA across the board.

[01:12:11]Even recruiting team were hiring. Um so we >> so recruiters are back to work. We had this whole like recruiter apocalypse where everybody thought they didn't need recruiters and now it's back.

[01:12:23]>> Yeah. But we love people who uh who love using air tools uh because that that does bring us extra amount of uh productivity and we aspire to be the smallest b biggest company in the future and it's possible. Um and if you are uh dreaming big and you want to drive the max amount of creativity, it doesn't matter which uh domain you work on. Uh please talk to us and uh we we love to um have a conversation with you.

[01:12:51]>> Awesome. Uh okay. Uh Demi, who are you hiring for? Who do you need? How do they get in touch?

[01:12:56]>> Yeah, for sure. We're we're also like uh we're still a very small team. We're hiring like a I would say designer like engineers and researchers. Um, and yeah, so we're we're trying to like, you know, really like really build like a a plat like for people to create their own unique agent that's more humanized through like the multi model like like research we're doing and and yeah, so like we're hiring like we're designer who is really interested in in the vision where like to help to to help people where people to build agent that has unique tastes not like a dry agent.

[01:13:30]And we are hiring like engineers who are interested in like the backend problem and more agent problem and we're hiring researcher who are interested in multimodel research. Yeah.

[01:13:39]>> Great. Awesome. Uh this has been another amazing episode of this week in AI. We drop every Wednesday. We record on Tuesday. We drop on Wednesday. Please go ahead and visit us this week in aai.ai and sign up with your email and you'll get our research. We're doing proprietary research on the AI space and who's getting funded and who are the next unicorns in the space. On the show, you get to meet the companies that have already reached that unicorn status and have vibrant businesses in the research department that we've created this week in AI.AI. Go sign up for the email and you'll find out about the next cohort of companies that are just three, four, five people and that are growing and building interesting things. We'll see you all next time. And bye-bye.