The Terrible Ideas He Keeps Saying No To

Hey everyone. Welcome back to TwoSet AI. Today I'm sitting down with Dara, CEO and co-founder of Delphi. This is the platform that turns experts into digital minds that can scale one-on-one conversations. I hope my interpretation is correct, but let me know if not. So Dara, awesome. Dara, thanks for joining. Can you tell us a little bit about yourself and where you're building?

That was perfect, distinct. Yeah, so I've been in the AI space since 2014 and started my first company in 2021 when I first got access to the GPT-3 developer list. And I was like, oh my god, this is. This is going to change the world. And during the time I was a solo founder, didn't really have any mentors, but was reading a book about my grandfather and really wanted to get his advice on what I was doing. So I created a digital version of him that I could talk to and learn from. And that experience just got me really interested in how we access other people. You you can read books, can skim a person's personal website. There's no way to ask someone a question to truly understand what they know and how they think about things. And so Delphi lets people create a digital version of their mind as a way to almost like always on AMA, allows people to ask them questions when they're not in the room, learn about their perspective. And yeah, we've been at it for three years now.

It's different from reading a book, isn't it?

Yeah, it's interactive knowledge transfer. A book is passive knowledge transfer. This is interactive knowledge transfer where it's immediate to your questions and it's personalized to you.

I'll ask you more questions around that. Maybe this is the future, but I would love to know more about it. Do you have any New Year resolutions? Both.

For myself or for the company? I think for myself, we're in a very, as an AI company, you can be pulled in a lot of directions. know, people want to use your product for all these different things. And my resolution is to just say no to everything and just continue to only do what we're doing. Um, cause I think as a company, you can just be, people want to use your product for use cases that it's not intended for, it can be tempting.

Mm-hmm.

especially if it's someone famous, to be like, okay, I'm gonna do this. But I think it's really important as a company to just really focus and be okay saying no.

I feel like you're facing maybe a lot more temptation than an average AI founder, because you're talking with the celebrities, right?

So many temptations. So many temptations and it was initially hard to say no but I think I've grown the muscle now.

Can you give an example?

Well, I think you'll have a famous person who will be like, you need to build this custom app, and it needs to generate videos for the user, and it needs to ⁓ call them through a phone number at home. And we can all do these things separately, but we can't be building these custom workflows for people. The product works the way it is, either you're happy with it or you're not.

Right, right. And I feel like that's one of the hardest thing being a founder and, you know, product manager, right.

especially in AI, because a lot of people have not conceptualized this technology before. So it's causing people to have more ideas than, you know, pre-AI. If ⁓ just a regular software company starts, you know what a software company is. You're not going to project your ideas onto them.

Right. There is a gap, isn't it? Because it sparks our imagination. Everything should be possible, especially.

Everything's possible. Everything's possible. Exactly. And eventually, like in five years, there will have been a lot of companies to satiate your curiosity. And then it starts to become the norm. It's no longer this novel new technology where the world is your oyster.

What do you say to them to reject their requests?

I mean, I think I'd just say, hey, this is what we're focused on. You can connect all your data. We create a high fidelity version of your mind that understands what you know and how you think. And you can allow other people to learn from you and talk to you, whether for free or you want to monetize it.

Hmm. Interesting. Okay. So you're basically saying, no, this is what we do. You can have what you want maybe in the next few years, somebody else will build it. It's not me.

Yeah, I just like, we'll get around to that eventually. Though there are people who will try to like create anime characters or fake characters with Delphi. And that's one where we're just like, we are human only, human verified, no fake characters. That's never going to be us. You should go to another platform.

I love starting with this question, although I already asked you a bunch of questions, but I'm going to ask you this. So this is something that really helps our audience to understand what you're building and you already explained a little bit, but can you, can you, ⁓ try, ⁓ explain what Delphi does at a slightly different angle? Like for instance, how would you explain it to an elementary schooler? It makes sense. Like my daughter will talk with like one of your ⁓ avatars. Makes sense. If she knows. if you have Taylor Swift. Yeah, my daughter is a Swifty. So I'm sure she's interested. What about a college student and a professional? Yeah.

Yeah, I think for elementary school, imagine if you could have Taylor Swift in your pocket all the time answering 24-7 personalized to you. ⁓ And I think for a college student, I think product marketing, there's a lot of ways to talk about Delphi. And I've learned over time the things that make sense. College students, like digital office hours, 24-7 office hours for the person that you want to access or learn from. ⁓

I love this explanation.

Yeah, so that we've slowly learned the more concise ways to talk about digital mind that knows everything about you.

Yes. What about professionals or maybe your customer, your potential customer?

Yeah, for our customers, we just say it's an always on version of you that can handle repetitive questions. So we didn't even use AI. We just talk about the utility.

That makes sense. Great contrast between three different layers. And what is the name Delphi coming from?

Yes. So the Oracle in Delphi in Greek history was the place you would go to when you wanted to learn about your future. so with, and we call it Delphi, not Delphi, but it just sounds cooler. But you can turn your mind into an Oracle that other people can ask questions of. So I think there's just been a lot of ⁓ inspiration from Greek history with our company. Our office rooms are called Athens and Sparta. So that's kind where the name comes from.

That's so cool. you know, it kind of, you connotates to this sharing of wisdom, like from the experts. This is a modern way of interpreting that, right?

Yeah, and even in ancient Greek times as well, mentorship and apprenticeship was so important. Having someone that can teach you their ways. And then I think that's really what Delphi is about, being able to learn from someone who's been there before.

I would love to learn from the best of the best directly, right? Do you think this will affect education in the future?

Yeah. I mean, think AI is going to affect education, not just Delphi. My hope with Delphi is AI is promising kind of general expertise of the masses, where Delphi is saying like, no, human expertise is still important. We value things that come from lived experience. It's the reason that you may follow someone on Twitter or you may read a book, even though a Wikipedia page on the topic still exists, that it's grounded human experience where you can hear specific perspective versus generalized intelligence.

How does that democratize this learning experience?

Yeah, I I think the experience with my grandfather, it's not that he said anything that was necessarily eye opening. It was the fact that I heard it from someone who had achieved the success that I want to achieve. And so I think being able to access or learn from someone that has been in your shoes and is where you want to be adds a layer of trust and confidence in the information that you're getting. That you actually are more willing to take action.

Yeah, someone who's done it and who you trust. Can you, I gather your grandfather is also an entrepreneur who had significant success. Can you share? What did he do?

Yes. Yes. He ⁓ was one of the largest business owners in Iran before the revolution, you know, worked in textiles, detergents, oil, but then he lost every, he had 30,000 employees at his peak. He lost everything during the revolution, came to this country with nothing. yeah, he amazing entrepreneur in his time.

It's also more than just the success, maybe also the resilience part of you also learning from him.

Definitely the resilience, how he treated other people, all of the above.

You know, people, I assume average people, including myself, hear the idea of AI clearance, it's still like science fiction to me. And the first reaction is like black mirror or kind of dystopian or a little bit creepy, right? How do you actually want people to think about this?

Yeah, I mean, we move away from the word clone and we actually move away from the word AI. Like I think the whole ethos of our company is that AI is not the main character. The human is. Other companies are trying to automate humans. We're saying no, AI is best in the power of the individual where you own your data, you own your identity and AI is being used to scale you and augment you. And it's not replacing you. It's not a conscious being. It's just a new kind of artifact of your mind that people can experience. If you paint a piece of art, that's an artifact of your mind that people can experience. If you write a book, it's an artifact of your mind that people can experience by reading. And now with Delphi, you have a new artifact of your mind that people can experience through conversation.

That's fascinating. It's more than that though. It's also an interactive experience, which means it can give me something that doesn't, you know, if nobody asked this question before, then it could be something new, right?

Yes, I mean, the whole idea is not just capturing what you know, but how you reason. So it can be maybe predictive of what you might do in these new situations.

Hmm, this sounds like more technical and I do have some questions around that. Give me one. I'll ask about that later. I do want to ask you that you could you cause you also mentioned, I think in our last call, you mentioned the comparison between Delphi being something that to let people access what you know versus LinkedIn, like what you've done on resume. That's a, you know, that's a, I didn't think of the comparison between the two. Can you unpack that a little bit?

Yeah, I I think on LinkedIn, ⁓ way we evaluate people or try to learn about someone is, we go, we see where they went to school, we see where they've worked, and you can search based on those things as well. But ultimately, I think some of the best relationships we have in our life and the people that you look up to aren't really just based on what they've done. It's how they think about the world, what they know, their experiences, their values.

Mm-hmm.

That's what made my, like all the people that I look up to, it's not just that they're successful, there's a lot of successful people out there, but how they did it. What is the way they did it? And those are not things that a LinkedIn profile can really show, where a Delphi profile shows what you know and how you think and allows people to explore it further based on their curiosity.

Right. That's very interesting. you envision people? I mean, I can interview an employee just based on their Delphi profile or the interactive chat interface. Is that reliable? Is that a use case?

Yeah. Yes, no, I speak to people's Delphi if I'm going to meet them before. A lot of candidates who apply for Delphi will speak to mine. And the great thing is when we hop on a call, they're not asking the questions that we ask over and over again. And the conversation is actually a lot more in depth.

Hmm, I see. Yeah. I'm asking some of the repetitive questions again, because we're sharing the information with the audience. Yes, but I did talk with your Adelphi. So it's very interesting. And I'll address a little bit of blurb about like talking with your colloquial versus talking with you directly in the talk. Later, not right now. You know, can you tell me who should use this?

Nice.

I know I opened it with that experts. That's what you told me last time. But you have Arno Schwarzenegger, you have a lot of the celebrities. If I look at your explore page, I don't feel I qualify.

I think we have two kinds of users. One is the Arnold Schwarzenegger, the Jay Shetty, the people who have a lot of content and people are learning from them actively. And Delphi is a tool of leverage where they can scale themselves in this new way and even monetize their expertise. And then the majority of our users are actually the high inbound professional that maybe they get inbound, people are asking them questions, whether it's about their business. ⁓ whether it's follow-up questions from online content, and now they have a page they can send people to. Ask My Delphi. You know, I may not be able to answer an email. You may not want to search through all of my content. Ask My Delphi if you want to learn about me before our meeting, after our meeting, after you interact with me online. So these could be founders, investors, consultants, domain experts, content creators.

That makes sense. see. you you probably the first ⁓ Top use cases people who are having a lot of inbound so maybe some influencer

Yeah, never repeat yourself again. I mean, it doesn't have to be just an influencer. It's anyone who posts online that is, again, a founder, an investor, someone who needs to hire, someone who needs to talk to a bunch of candidates, someone who needs to talk to a bunch of customers. Yes, yes, I'm definitely my own user. It's very, very, very good.

Are you one of those? Are you the only one? What are you using? Who's using your Delphi and what are they talking to you about?

I mean, Delphi is my main profile. you know, I'll post on Twitter, goes viral. If anyone, you know, wants to ask a question, they go to my Delphi. Customers talk to my Delphi. Potential customers talk to my Delphi. People who want to work for our company talk to my Delphi. Journalists who maybe don't have time to get on a call with me, they'll interview my Delphi. And so it allows me to scale my leverage and then also be notified of the people that I should actually hop on a call with.

Mm.

because our incentives are mutually aligned.

Right, do you ⁓ sift through that data? Because I got to believe that's very rich data.

I don't even have to sift through it. It will proactively recommend ⁓ based on what it knows about me.

Oh, okay. So it will based on like 100 or maybe 10,000 people talked with you in the past week, and you'll be able to know who's potentially your, your leads, or whoever.

Yeah, we're the best people. Leads, customers are just like interesting people that would be good to meet.

Yeah. that's very interesting. That's a great use case. It could be a cells gen tool, part of the cell cell strain pipeline. Yeah, that's very interesting. I'm discovering a new use cases because at first glance, I feel like I have to be a celebrity to use this.

Totally. Yeah. that's part of the, we're doing a rebrand because I think these big experts that we brought into the platform, you know, that is the obvious first use case with every startup, there's a wedge. and these are the people that have the most knowledge, but ultimately it's, it's similar. It's like an always on availability. And whether that always on availability is for like your followers or in your audience, or it's for your customers or leads, same use case, different audience.

That makes total sense. ⁓ How are you different from like character AI or like story file? ⁓ What do you think like technically or philosophically? How are you different?

I mean, think a character AI allows people to create characters for entertainment. It's not really utility. I would say Delphi is utility, and it's verified and human first. StoryFile, I'm not too familiar with. I think they're focused on legacy. Well, again, I view Delphi as a personal profile and a utility, a new destination for the human being. And our focus is really on verified human intelligence and representing the mind.

How do you verify the intelligence? With a XAML.

⁓ Verified as in, know, sign up with LinkedIn or submit a photo of you holding your ID. you know, any Delphi you see, you know, is owned by the person. It's not fake. And that's kind of where you get the trust from.

Okay. So I cannot create a Delphi for somebody else, basically.

No, you cannot. And then in terms of intelligence, every Delphi you talk to is trained on a subset of data that is curated by the owner. It's not just open internet.

Okay, so it's a fine-tuned model for understanding the owner, him or herself. Got it. ⁓ Is there any surprising use cases that you've seen? I've heard customer service and then maybe sales gen and talking with people, building relationship.

Yes. I view Delphi the same way that you might view a personal website. You have a personal website and you can think about your personal website as a form of customer service. It's where people get to learn about you. You can view it as a form of sales gen, but ultimately it is the representation of you when you're not in the room. What that results in is, you know, leads, support, but it's representation. ⁓

Mm.

And that's kind of the main use case we want to settle around.

I love that analogy. So if I were to build my own website, of like I have some blurbs about me and my past experience, so people can just look at it. And also, I can also have embedded a component of Delphi of myself so that anybody can talk with me directly as well. And you will supply that data to me to give me insights. What kind of insights can I have?

Exactly. It'll tell you interesting people, trending topics and questions, and it will notify you when it doesn't know how to answer a question. So you'll see where the gaps are, and you can actually hop into conversation and answer that question yourself.

Awesome. feel like I'm very tempted to use one and now like just really because I was thinking what is the new business cards these days, right?

Exactly. That's a great way to frame it. It's almost like the new business card. So I'll get you set up. I'll get you set up.

Awesome. Yeah, I'm curious, how does that work? And how should we present ourselves in a very, very succinct and also very ⁓ memorable way? Right. It's really interesting. It's a very interesting angle. ⁓ technical side of things, maybe a little bit. Just design thinking. ⁓ Because you mentioned last time about mind architecture. So I'm curious, what does it really mean, technically?

Little bit, little bit. Yeah. I mean, I think in 2014, I read a book called How to Create a Mind by Ray Kurzweil, where he said that one day AI would be able to recreate someone's mind patterns. And he talks about that the mind is a hierarchy of pattern recognizers. And so in 2021, as I was playing around with LLMs, you realize an LLM is a pattern recognizer. You know, it recognizes the patterns and words and it produces an output based on those patterns. And so the mind is almost like a hierarchy of LLMs working together. If you can represent the mind in the right way. And so for us, we, yes, you can fine tune a model, but the problem with fine tuning models is you lose explainability and you lose control. Where I think if you're representing a human explainability and control is really important because this is all about trust. This is the difference between us and Chachi between Claude is like, this is how this human mind works. so I think the

Mm-hmm.

The architecture of the data is really important. And so we are developing a temporal knowledge graph, which is good because the mind is temporal. It changes over time. You need to be able to track how someone's beliefs and opinions change over time and the recency bias of new things. And the knowledge graph has different primitives, like events, entities, heuristics, and how those things are connected help it learn how it behaves in different conversations. ⁓ And then we have an agent that knows how to traverse and use that mind in conversation.

Hmm. So you are doing a knowledge graph based kind of like a retrieval mechanism, you know, the grounded retrieval, like RAG systems, retrieval, augmented generation. Okay. Yeah. the, when, do you feel when it's, cause there's a generation component, right? Do you feel when it's generating an answer to a new question, is your results still representing the person, the owner, him or herself?

So we have two settings. When it's on strict, it will only answer questions that it knows the answer to directly. And otherwise, it will say, I don't know. Then we have adaptive, where it will answer the question if you have data where it can infer what you might say in a new situation. And we're building evals for that right now. an example would be.

Mmm.

If someone were to ask my Delphi, you know, I am a lobbyist. How can I use your product? I have no data around lobbying, but you know, given that it knows how I speak about Delphi for other professions, it might be able to infer what I might say for that profession.

Got it. And how do you set up the evaluation? Do you contact the owner and say, you verify this is what you're going to say?

Well, I think right now with eval is we're actually just creating data sets. We're just creating, we're creating a lot of data sets on like, what does it mean to reason? What are, ⁓ it's really a heuristic. Like you have, if I put in a piece of data that shows how I think a podcaster should use Delphi, there is almost a heuristic there of like how I map profession to Delphi use case. Take that heuristic and apply it to something new.

Okay. Okay, yeah, that's interesting. And then that's the baseline. I mean, can always do more, a lot of stuff you can go into the eval piece, right? And you can get as sophisticated as you want, depending on your metrics and what you want to Or your customers' crazy requirements. ⁓ If I were to make a clone of myself, what do I need to give you as input?

Totally. Yeah. Yep. Well, now with onboarding, you sign up with LinkedIn, ⁓ where we find everything about you online, or almost everything. So we'll already find some data about you. And then you can add more if you want. can add podcasts, YouTube, Notion, Google Drive. We have this interviewer feature as well that will ask you questions about any topic that you want to embed expertise in.

How long does it take?

It takes any time, depending on how much content you have, it could take anywhere from 15 minutes to one hour. Yeah, the interview time is pretty fast. The interviewer time is very fast. The thing that probably takes the longest right now is YouTube, ⁓ because YouTube ingestion, you transcribe, you dialize the speaker, same with podcasts. Those are probably the two that take the longest.

That's not that long, including the interval time? Right. Yeah, that makes sense. So if I have like more than 100 videos and if I just pass that to you, am I not consuming a lot of your tokens? Am I going to be a high cost user?

We've cost optimized, not too worried about that. A lot of iteration and seeing what's out there in the market and how we can use caching and, you know, model routing to make sure costs are down.

Really? Okay, how do you do that? Yeah, yeah. I mean, we can go with the YouTube transcripts, First, that's the first batch. So that doesn't go through a transcription model and things like that, which is

It's actually better for us to just take the YouTube videos directly because then we can get the time. Yeah, it's better if we do that.

Really? I didn't know that. OK, I thought you were just going after layers of approach, getting from YouTube API first and then, know. Hmm, interesting. Yeah, I know the quality from them is not very good. Yeah. If you're using LLM along the pipeline, there is the classic problem of hallucination, right? Can you tell us a little bit more about your opinion on this and how bad can it be? What's the risk for the user?

I mean, this was kind of the first thing we really had to care about because no one's going to use our product if they're afraid it's going to make them look bad. know, humans care a lot about their reputation, how they are perceived. Delphi could, cannot work without solving this problem. ⁓ So for the first two years we had kind of solved it. We didn't really have e-valves. was more like we tested it a lot and we couldn't really find ways to break it. Now we're finally doing e-valves. And like I said, on the strictness setting, when it's strict,

Mm.

and you ask a question, like, what are your political stances on XYZ? It will say, I don't know, but let's talk about Delphi. So it can like intelligently weave it back to what it knows about and avoid the things it doesn't want to talk about.

Mm-hmm. Hmm. You think do you think? Tragedy BT even tragedy BT or Claude and perplexity are not doing as good Of a job because because this is high stake for you right less of a high stake for them

I think it's not as high stakes for chat between Claude, be honest, they risk, the beauty of those products is that you can talk about anything. There are no rules. And I think that's actually one of its, that's not a bug, it's a feature. And so by like spending too much on hallucinations, I think that they could potentially worsen the product. ⁓ Perplexity, obviously this is very important for because they're the fact engine.

Right. Right. Right, true. So would you say that you are the best company that tackling hallucination because this is high stake for you?

I wouldn't make the claim that we are the best company yet, but we need to be the best company, we want to be the best company, and so soon we'll have ⁓ our benchmarks for evals and hopefully we can release that.

You have a you have somebody who's working on this I'm happy to have them talking about this as well on my podcast if you can make an intro

When we release the benchmarks, absolutely, that'd be great.

And then we can tell everybody that we are the best company. I'm kind of your evangelist now. So I'm part of your company. Accept me. Yeah. Awesome. Right. We can make that claim. I would love to learn more about how you do these things. Cause this is a big topic everywhere across all industries, right? Agents making decisions for people. And you're like representing people, there's their voice, what they're saying.

I would love that. I'm writing that down.

I remember last time you also mentioned something interesting. You said you are very bearish on consumer video interactions. I was really surprised. I thought people want to have video chat.

Yep. I don't think so. And it's either two reasons. One, the state of technology is not good enough. And we've seen this, we used to have video. I think it's like a cool, shiny object. Like people will try it once, but they don't come back because once the novelty wears off, you start to pay attention to like the weird little quirks and it becomes a little creepy. And specifically for us, voice is very popular and text is popular. When you see the video, almost... It doesn't leave room for imagination anymore. I think human leaving room for imagination with AI tools is important because you almost have to suspend disbelief for a bit where video does the opposite. So maybe we get to a world where video is so realistic that you can't even tell. ⁓ Even then, I think there will be some weird psychological doubt in the mind of consumers. ⁓

Mm.

that they want to sit there, see their own face, and also see the face of a fake character.

Yeah, it's hard. It's really hard for me to imagine. But if I think from the book's perspective, like our book analogy, right? I'm looking at a book. There's a lot of room for imagination. And now you're making the book talking back to me. I still have the room for imagination. if you're thinking the book becomes the person and then starting. Yeah. Yeah. ⁓

Exactly. There's no more of them. Yeah, no more guessing.

It does sound a little bit weird, although I never tried that. What do you think of the, you know, the star wars, maybe the holographic imaging? So it's not a reality, but it's a little bit realistic with the image. What do you think about side of the world? Okay.

Yeah, I mean, think that's entertainment. like Delphi's utility where, you know, the holograms and character eye, this is around entertainment. And I think that's kind of where we see the division. Should we invest in technology? Is there a utility behind it? No, it's actually just shiny objects and maybe good marketing moment, but maybe not high utility.

I can tell you're a very disciplined founder. Like no shiny features, it must be hard.

only through trial and error that I've learned that shiny features are the way to a company's death.

my goodness, this is so hard because as you said, for AI projects, AI products, everybody has 100 ideas and those ideas compounds. That's a path.

Yeah, I mean, since the beginning of the company, all people want to talk about when they want to talk about Delphi is, hey, you should do this, you should do this, you should do this, you should do this. And I'm just like, you got to let it all go.

What else did you learn while building this?

I learned, ⁓ yeah, probably the number one thing is what I told you, focus is the most important thing. And people have ideas with technology that sound good in practice in theory, but in practice, they just don't actually work. Like one other thing that I feel very strongly about, which is probably the most recommended idea we've gotten to Delphi is you should build a business around legacy. People are going to pay to bring their grandparents back to life and it's going to be amazing. And. One, I just think it's a horrible strategy for a company because you don't want to be a company that's associated with monetizing grief, where ⁓ if you stop paying us, you're going to lose your grandfather. That's just horrible. ⁓ So yeah, I think that that's an example of an idea that sounds like it's a great idea in theory, but I actually don't think people will pay for it and they'll hate your company. ⁓ Another is just how with certain ideas,

yes, terrible, terrible.

Most people are just going to think it's a bad idea. And that's actually not a bad thing. There's a book called Crossing the Chasm, where it talks about, you know, the technological adoption curve and how you have people who really, really get it in the early days. And then there are groups of people who only get it once people they trust start to use it. And it starts to become a domino effect. And so, you know, by starting with these high status people, slowly more people are starting to see, Delphi is something that I can use. And we're still at a point where a lot of people think our company is weird and they don't understand it. And I've just seen so many people change sides. And I think it's just a matter of time and patience. You just gotta let it play out.

Where do you stand right now on the curve of the chasing chasm?

I think we are in the middle of crossing the chasm, actually crossing the chasm to the mass market right now, where we have like a bunch of proven use cases. People, a lot of people still think it's like an only an influencer tool, even though there are people that are not using it as ⁓ they're not influencers. So I think part of that is gonna be like simplifying onboarding and updating the brand of our company.

Mm-hmm. Can you say one more time what's your top three use cases?

Yeah, I think I would actually just say it's one use case, which is just always on availability, handling repetitive questions.

What is it? You've probably seen a lot of conversations between users and their AI clones. I'm still calling it AI clones, but they're Delphi. ⁓ But Delphi isn't just like a GPT as we talked about. It's not just a chat bot. It represents a real person. Do you see users, you know, based on your observation, do you see users behave differently than they would with chat GPT? How you chat with chat GPT and how you chat with somebody else Delphi. Does that behavior... different? ⁓

People are more respectful for sure. mean, this is an entity of the person. ⁓ But then you have people who are like trolls on social media. It's very similar to social media behavior where you can comment something really mean on a tweet and people still see it's you. Or you're more thoughtful of your response. Though I do think some consumers have said they like talking to the Delphi of the person.

Right.

because they feel more comfortable asking dumb questions that they wouldn't have ⁓ felt comfortable asking with the actual person. Yeah.

Thank Right, yeah, that's a good point. I would feel the same way. Like sometimes going to conference and then if Elon Musk is there and I want to ask question, I would never raise my hand. was like, okay, good to know that you wouldn't either. I want to though, but maybe you should get his Delphi.

I would never again.

One day? Okay, not yet. You're in a Bay Area? But you're less than six degrees away from him. Good luck.

Yeah, that's the plan one day. In Bay Area. We'll get him at the right time. think with people like Elon, you only really got one shot, so we want to make sure it's the right timing.

⁓ yeah, that makes sense. Now you sound like my CMO. My CMO says exactly the same thing. Yeah, this company will have one shot. We've got to get it right. Yeah. When you're talking to users, potential customers, is there any objection that you hear a lot?

Yeah. the right time.

What else?

Totally, all the time. ⁓ They'll say, what happens to my data? Am I giving away my likeness? Which no, we've opted out of all data sharing. We don't train models on your data. You own your data. They'll say, what if it hallucinates and makes me look bad? That's probably the biggest thing we need to overcome. We're very confident in our product. And you could tell someone as much as possible, hey, of the million permutations of questions someone could ask, you're going to be good.

Right?

How do we teach that in a very short amount of time? It's difficult. And then number three is kind of the crossing the chasm thing. Like people can't quite see how to use it until they see other people using it. And LinkedIn had a very similar situation. The early days of LinkedIn, like what do I use this profile for? Like what even is this? Why do I need to share this? They had the same exact problems that we're kind of overcoming right now.

Right.

And it just happens through memeticism. Like once enough people are using the product in a certain way, then only then do the mass majority of people change their mind about something and they're okay starting to use it.

Yeah, as a new category, mean, you're different and it does have a cold star problem. Right. And do you envision yourself, you might become a network, ⁓ know, just like LinkedIn or even marketplace in the future.

Totally. Totally. Yeah. mean, I think, no, no, I don't think we're planning to replace LinkedIn because I think there are two different kinds of network. LinkedIn is a professional network where the nodes are connected by who you know, and who you've shared work experiences with. Where Delphi is a knowledge-based network where the nodes are actually connected by similarity of thought or similarity of knowledge, or, know, someone's knowledge actually filling a gap in your own, in your own mind. So.

As you scale, you face this problem, right? Because right now, I mean, I see a lot of celebrities there and I'm interested in this person who's expert in this area and then I'm going to go chat with this person. Try it out, right? But as you scale and you have millions of people on your platform, how am I going to know who I can talk with? And I'm going to think there must be a lot of wisdom in this library of people, experts on your platform. But who am I connected with?

It's a different kind of network.

and who should I talk to?

That's the magic of the knowledge-based network where you have a Delphi, you have a digital mind, and based on that, we can say, here are five people who think similarly to you. Here are five people who can help you achieve your goals. Here are five people whose opinions actually contradict your own, and we think it would be good for you to explore the other viewpoint.

So you will do some recommendation as well. Nice. Are you doing it right now?

Yeah, I think it's the only way to manage the thing. No, no, got, I think we're not going to focus too much on the network until probably like March or April of next year.

Okay, okay. Later on. then this is this is a great. This is a great add on like later on after I know as you scale, right? Yeah. And then if you recommend somebody I can try with other new things, maybe I can think of ways like I'm going to reach out to them on LinkedIn directly.

Yep. Exactly. Exactly. So I think LinkedIn, if anything, could be complimentary to Delphi.

Yeah, okay. That's the reason you're not replacing LinkedIn. Although I kind of hate LinkedIn these days. Right? ⁓ So much noise. Do you think you will have ads one day? Too early?

Yeah. Too early to tell. I mean, I, again, want to do what's best for the user experience.

Got it. Is your darkest day building this?

so many. ⁓ I'm trying to see if I can talk about any of them. I think there was one time last year where we got banned ⁓ from OpenAI and Anthropic. And that was a tough day because I thought the company was literally over. So it's just like everything in the product is breaking. I can't get in contact with anyone from OpenAI and Anthropic. We're done. Because we helped a political figure create a digital version of themselves, which is against the terms of use of a lot of these companies, I think. Now we don't support political figures. That was tough. I actually thought the company was going to

What?

die and we were done. ⁓ But we found a lot of clever ways to get back up and eventually get unbanned from those platforms.

Hmm. Yeah, there must be a lot of these type of moments, right? It's unexpected, isn't it?

It's expected. mean, at this point, you know, it happens one time and you're like, I hope this doesn't happen again. Then it happens another, and then it happens another. And you're like, actually, this is actually the norm. And this is going to happen many more times. Yeah, now I'm prepared as you know, knock on wood.

Yeah, but you're prepared. Yeah, knock on wood. Okay. Good luck for the new year, by the way. Yeah. And last time you mentioned you have something new launching around this time. Do you want to share?

Thank you. We'll be launching our free tier soon. Not now, but probably end of January. This new onboarding thing I talked about where anyone can sign up with LinkedIn, we find everything about you, and now you have a new beautiful profile that can show what you know.

What does free tier mean?

You can set it up for free. So right now you can only use Delphi if you're paying. And now we have a free tier where anyone who has knowledge to share and they want to set up one of these profiles, ⁓ which you should link my profile so people know what a profile looks like, they get one up.

Okay. Yes. Yes, I will do that. And you're just limiting the usage.

Training data usage integrations, like lot of the enterprise paid tiers are like, I want to add it on my website. I want intelligent analytics. I want like a full CRM, you know, all these like additional things.

Mm-hmm. I see that makes sense. The question, sorry, I lost that one. I'll ask the tougher question now. ⁓ You I mean, during our prep call, you mentioned something that stuck with me. You mentioned about Delphi, how Delphi exists in a post AGI world. You want to share more about that?

Yeah. So I think there's a thesis that we're kind of operating on that if it's not true, then Delphi might not work out. But the thesis is, you know, you project ahead AI is, you know, so smart, really fast, pretty much free. And there's an abundance of access and information. You can get anything you want in the world. What ends up happening is in my opinion, is humans become the scarce resource.

Mm-hmm.

humans become the things that we actually start to appreciate, things with lived experiences. And so our bet is that humans actually matter more in a post-AGI world, because we're going to crave things that are real. You know, for the next five, 10 years, maybe there's shiny object syndrome and novelty of all these deep fakes and AI generated videos. But at certain point, the content becomes so much that you're like, I don't even know what to do anymore. There's too many things to consume. There's too many options. Can someone just tell me what to do? Someone who's actually lived it before, can they tell me what to do? And you see this with chess right now, for example, or in boxing, you have robots that box, but it's not as engaging as seeing humans that box. You have AI that has beaten the best chess player in the world, but you still watch human chess players, things with lived experience.

Mm-hmm. Right. Yeah. I resonate greatly with that thought. ⁓ We've seen musicians like robot, violin players, you know, it's fun to watch them one time. And then I don't care about them. If my daughter does it, then I feel like, I'm going to watch it. Right. Even though how crappy she is, right. It's still amazing to me. And I feel it's more of a role model thing because we're the living thing. I'm living my life and I want role models. I want to see other people achieving and having the discipline, having the resilience and not from a robot. They beat us, but so what, right?

I think there's an unspoken, almost spiritual connection that we have with other humans. there's no way to quantify it, but when a real human does something, there's a different, almost contract that is being made.

Yeah, we don't have that data to feed into any AI user, right? And what is that data?

Yeah.

The data you're talking about, it's like a subcontract.

which data. The unquantified, I don't know. It's a combination of almost like evolution. we, 300,000 years of edge evolution, we have been in communities, we've taken care of our family members, we've made friends, we've made relationships. Like this is primal in our nature and technology can change super fast, but you know, that primal instinct, I don't think changes.

Right. How is our life going to change? Are we going to stop using the technologies?

I mean, I'm a bit of an idealist. you know, in my world, TikTok and all these things are like smoking a cigarette on an airplane used to be where it's like, oh, we used to do that. That's crazy. And that's the optimist point of view. The negative point of view is actually, you know, social media algorithms get so good and people get hooked and...

you Yes.

you know, if generation alpha is already getting hooked, then how are they going to know to raise their kids any differently? So I'm an idealist. I will hope for the optimistic side of things.

Me too. hope for the optimistic side of things as well. And I already, I can tell you, I don't know how to raise my daughter in a different way because she has access to so many, so much more than how we grow up. So it's so different and hard. It's hard. ⁓

It's hard.

Do you think it's possible that one day just AI just get so good that anyone can set up something like a clone similar to your yours with a few, you know, a few simple prompts, like let's say Chad's LGBT does it.

Yeah, I mean, think our, ⁓ it's very similar to the argument of like pre AI. you know, if you start this company, Facebook or Google, they're gonna, they're gonna beat it. And I think that's why focus is so important, which is like, let's obsess over like, what does it mean to have a human mind? ⁓ let's make that useful and then let's build a big enough network such that it's like a distribution mode where, you know, LinkedIn has not been able to be disrupted because they built such a strong network.

Yes.

⁓ And so I think if we're able to build a similar kind of network, then I'm not too worried.

Okay, I'm looking forward to learn more from your platform because I feel you're going to have a lot. The content is constructed in totally different ways, which is very interesting to me. The way, know, we historically learn is from, from university, from schools, from books, from watching YouTube, YouTube channels, right? It's very different from interactive interaction with, a, with like a wise object, a wise human being.

I mean, we'll see. think, you know, before AI, there's Bloom's two sigma, which shows that having a private tutor improves your ability to learn by two standard deviations. So it's like a positive thing. But then there are more recent studies that are like kids who have been using chat GBT through college are actually not as well off because they haven't been forced to think harder. ⁓ so I think we'll see whether it's net positive or negative.

Yeah, that's a fair point. I feel like it takes a curious mind and the human nature to really continue to work. Yes, good point. Do you have Steven Bartlett on your platform? My role model, so I would love to. Yeah. Awesome, right? ⁓ Are you hiring by the way?

How do you teach curiosity? I think that's probably the thing you want to think about with kids. Not yet, but hopefully soon. He's awesome. He's great. We are infrastructure engineer, design engineer, and consumer GTM.

Okay, what kind of profile do you think would be a fit for you, for your team?

⁓ So culturally high level, think ⁓ every single person in the company who works here has moved to San Francisco for the company. know, everyone's very passionate. So I think someone who is like passionate about what they do is kind of a common factor. ⁓ Someone who's creative infrastructure engineer is kind of obvious. You know, they can build scalable infrastructure systems. Design engineer is pretty much a designer that can code. Consumer GTM is someone who has scaled markets and networks.

Mm.

So those are the three that we're looking for.

If where can people find you, whether they want to, you know, create their own Delphi or join your team or just learn more.

Yeah, so delphi.ai slash D-A-R-A is where you can ask my Delphi. ⁓ And then I post a lot on xD-A-R-A-L-A-D-J-E.

I'll share that link. I noticed that you don't post as much on LinkedIn. Not as much. Yes, yes. I have a, I am starting a traditional closing question. So if you were to remember a book for a 20 year old or 40 year old or a 60 year old, what would they be and why?

Not as much, not as much.

Yes. But you don't have to, if you don't have three books to recommend, that's fine too. Whatever you can think of. I would love to have some, you know, new book lists. I'm out of books recently, so yeah.

A book that each age needs. I think a 20 year old, ⁓ this is probably like cliche, the Almanac of Naval, Ravikant. I read that I think at 22, whenever it came out and it was so good. So many good life principles in that book for young people. ⁓

How old were you when you read it? Also, you were just in the age range when you read it.

I think like 23. And I studied that book. So I think rather than reading a lot of books, I think find a couple of books that you study. A book that I read recently that I've been studying is called Anti-Fragile by Nassim Taleb. Phenomenal book. there are books that come across that just have such unique ideas. Truly original thought. And anti-fragile is all about, anti-fragility is about getting stronger during stress. How can you become someone or your company or systems

Mm-hmm.

You can build systems that actually benefit from chaos. And chaos is the natural order of things. Like I said, there's always going to be these crazy heart moments in the company. How can you get stronger for them?

and try Yeah, do you have one tip you can share with me?

I mean, it's more of a mindset where if you reflect on all of the hardest things of your life and you think about the period after and who you became as a result, I would say most of the time you're like, wow, I really grew as a person. So this is actually just learning. ⁓ It's like going to the gym, but for your mind so that the next time you experience something hard, you're already used to the 50 pound weight and now you can handle the 60 pound weights.

Mm-hmm. I feel people don't share a lot of those, you know, misery of going through something. Once they pass it, I feel people start having a rosy glass on the past because most of founders won't tell me how, you know, what's the hardest time, how hard it was. They all feel it worth it.

I think it's good to recognize it because then it makes you more confident about future ones.

Right, right. Awesome. Yeah. Thank you so much for your book recommendation. I'll write them down and yeah, I'll get a chance to read them as well. And I really enjoyed chatting with you today, Dara. Thank you so much. Yes, awesome. And I'll share all the information that you give me to the audience. If they're interested in joining your team, they'll reach out.

Thanks for having me. Okay.

Thank you so much. Give me one second.

Cool. Thank you so much. This was fun. All right. Yeah.