EP - 124 | ‼️⚠️ MISSS AVVADDUUU… NEED OF THE HOUR | Ft. Vaibhav Sisinty | Raw Talks With VK |

00:45 Jobs are getting lost because of AI, right? Jobs are losing. But jobs are evolving too. Can we talk practically about what kind of jobs can come? The highest paid people in the AI world will be salespeople. It's not difficult to build a product yet. Who puts the product in hands of people? It is the salesperson This is low value work. This is high value work, become mad! The meta team showed me the future glass I had a chance to play with it. It's insane bro, my glass has a screen in front of me and there's like a teleprompter or display on top of that. You don't even know about all this. We are opening up about things with AI that we will not with any human in the world. Our deepest insecurity. Are you asking for an AI? Around 25 million chat GPT chats were leaked. I don't want my messages to get leaked man

01:29 What about emotions? I don't know. They might get lost now. Why, what are you going to say? Oh his name is Vamsi right? What's the question? Actually i don't have a girlfriend so i wanted to know... Okay fine So It doesn't feel like you're talking to an AI. We should stop here and go talk to Rubik. Let's go. Excuse me can we go for the next one 500 Now this is more scary This is about This is the podcast thing. We both will go viral. So this man is the man who scares me the most. Every time I sit, there's a curiosity. Of course he gets into reality which is actually new world altogether. Again, sitting in such session

02:30 My intent is never to scare people. Sure, I know... My intent has always been to tell what's happening. People don't know what will happen. It is important to say that. Now, What you do with that information is up to you. Of course! So, if you think I am scaring people let us do more solution oriented stuff today then Let's do it tomorrow. If you have a solution, we can talk about solutions. It should be there or else without any knowledge. One question I wanted to ask you because both of us know what is the intent behind you coming all the way from Hyderabad for 2 days almost 6 hours. Why are doing this? We both have clarity on that.

03:11 For the viewer who is watching, I genuinely wanted them to sit with a book and pen or laptop. Or just note down everything in their phones because many people watch it on TV at home too So if i'm asking you as a layman who's watching this why should I watch this podcast? Not just this, the second one also. The shoot I'm doing tomorrow... Maybe it'll take a week's time for me to finish that. It will take at least one week. So why should someone watch? Genuinely, I am telling on the screen. I can cut this out if you want. Please over to you! Maybe it would be better in Telugu. The reason everyone has to watch this podcast according to me is One is that If you're ambitious and you want to win Because we are not living in normal world now We are living in an AI first world

03:53 Right. And a lot of people think that AI is enough, it's enough to use it. But the actual reality is if you want to not fail in this world there are two kinds of people who will definitely fail in this first world and I don't want your audience or my audience to be that one People who don't use AI they will definitely fail very very hard for you to succeed because you're competing with something that is 100 times smarter than everyone else around. If you are not using it, then you're not going to succeed. Two people who are only using it. You use AI itself then also you will not succeed and fail. So knowing the difference between what to use where to use

04:48 Where not to use is what we'll try to tell you here. Perfect! So, I will not tell you that after watching this podcast... You will succeed. I won't say it because that depends on you. Right? What I will tell you is what you should not do so that you don't fail in the AI first world. And since there's a context now, we have our community right? I joined last time when I came here. So I thought this time, I'll shout it out again. Maybe the viewers will find it useful because I'm getting a lot of information from there. Even though I don't follow them regularly whenever I open that... I joined the WhatsApp group basically. Yeah yeah. Whenever and wherever they open up, I get new thoughts from their updates. That's how we generally explore right?

05:32 So this time I thought, it's open to anyone right? It is free. There is no such thing as a course only for free. We do it for free. So what we will do is like put the link down let everyone join and see how many are turning out from this. I want you pause and click the link in description and lets see if there are any comments below because Many people don't even have a flat for where do we get the updates. I actually say staying ahead, the name of the community we call it staying ahead It's right now probably the largest AI community What is its number approximately? The total community size is 15 million people On WhatsApp we recently started purely for India because India is all on WhatsApp

06:14 How long was that? I think you joined only then. In that, the idea of this was there like the biggest question that I would get everywhere I go is what is that single source where I can learn or know what's happening in AI In the easiest way possible. My answer was, you have to read Twitter, you have to read Reddit You have do this and that and nobody was happy with the answer I said okay fine we will make a community And that's why we made it free Crazy Right so i think everybody has a shot at staying up to date No must join, I'll put in the group below, I mean description Do join and comment down Thanks man So we did this 10 months back after doing last episode Both virtual and everything came there

07:46 If I were to explain in a technical language, it would have been like an assistant. Which is chat assistant right? You ask one question and she answers the other. Great! Today it has gone to doing things for you So now when we talk about agents as a concept It has become as mainstream as it can get today. I cannot imagine a thing that I can do that my AI can't do. I mean, not in the physical world. I'm talking about digital world. Yeah right so obviously i still have to eat the biryani which i still want to eat but you get my point right? So it has evolved quite significantly. I think in last...so weirdly enough right..in the last three months we have grown

08:39 probably more than what normally would take a year or two in just last 2-3 months. So, year or two in the AI world regularly... I mean, how much technology shift has happened within 1-2 months is just insane! The drops yesterday were crazy. Like when I say drops, a ton of new updates came in yesterday and they were just insane so.. I think the game is changing Let me be a layman in this world. Now that we have the topic of agent, how do we differentiate or define for a person who doesn't know this? What is an agent? What is AI Agent? What is workflow management? What is automation? Or will it meet all these things? If you just talk about those three terms meaning

09:23 So basically, first is your AI assistant. Which is chat GPT, Gemini, Cloud, Grow. Many people have stopped there itself. Yes they have stopped there only. That is basically when you ask a question it will give you an answer. It won't finish your task. You yourself will finish the task. This is assistant right? Now I'll tell you what an AI agent is. Let's say if we have an email agent The email agent is always on. It's like... There was an agent named Jerry. Like even the way Jerry evolved is just insane, we will talk about it if you want to. The kind of things that can do right now is just wild because of new technology that came into the world called as open clock. But before we get there agents means it does work

09:23 So basically, first is your AI assistant. Which is chat GPT, Gemini, Cloud, Grow. Many people have stopped there itself. Yes they have stopped there only. That is basically when you ask a question it will give you an answer. It won't finish your task. You yourself will finish the task. This is assistant right? Now I'll tell you what an AI agent is. Let's say if we have an email agent The email agent is always on. It's like... There was an agent named Jerry. Like even the way Jerry evolved is just insane, we will talk about it if you want to. The kind of things that can do right now is just wild because of new technology that came into the world called as open clock. But before we get there agents means it does work

10:09 Vamshi wrote me this email. If you write the email, it will be written and copied and pasted. What you can do with an agent is, first of all, you can give access to tools. Okay! So, you can give access to your Gmail. You can give access to your skill of writing and after giving the access every time there's an email that is coming to you don't have to say reply to this email. You can set up an agent in a way where every time an email comes, you read the email

10:46 Try to understand what's in the email. If it is not so important and let say someone put something like some fan, he said great content love your content It's a nice piece you reply saying thank you so much for being nice You automatically reply or send the email Let say one more... Nika new pet coach cover a little bit coaching saying let's you wanted to be automated completely and agent is automating email it should be emailing in chattavani the email one she kept awesome leather. Similarly, if you get another email from some big sponsor. Let's say Google wants to work with you It is not going to reply saying that this is my payment and this is the price you pay me. There it will know how to set up an agent. If a partnership or any of these kinds of very important emails come, You don't reply to the email. What does that agent do? It will ping your Slack or WhatsApp.

11:39 Whichever team member manages the partnerships, it will just copy that to them. If someone is angry, there would be such a person too right? I mean, you'll just read it so that you don't even have to come across negativity. You just archive it right? So you can set up these rules, you can write a very interesting system prompt or prompter agent also. So it will work like a 24 seven employee for you right? That is an AI agent usually which doesn't usually need a trigger the trigger is automatic. The trigger is in a way automatic. It's like a 24 7 working employee one quick question.

11:39 Whichever team member manages the partnerships, it will just copy that to them. If someone is angry, there would be such a person too right? I mean, you'll just read it so that you don't even have to come across negativity. You just archive it right? So you can set up these rules, you can write a very interesting system prompt or prompter agent also. So it will work like a 24 seven employee for you right? That is an AI agent usually which doesn't usually need a trigger the trigger is automatic. The trigger is in a way automatic. It's like a 24 7 working employee one quick question.

12:27 How much of it, to build an agent how much technical knowledge is required? How many open cloud platforms are there now. How much would that be enough? Earlier also I had removed all the unnecessary things because most of the things that were built were no code. It was 95% in drag and drop. Now that has become even simpler. You must have heard about wipe coding. Yes yes. I think we build a landing page or whatever also last time. That was wipe coding Now, incidentally just a few days back Claude dropped something called as managed agents. Managed agents. Like literally you probably might not have even come across it. Right? In the anthropic website, in anthropic console dot anthropic dot com is there which is the login layer of Anthropic not Claude. If we go to that, then you will have managed agent option and if we select that

13:17 Just like you are vibe coding, you just say that hey! I have told this task right? I want my emails to be managed end-to-end by an AI. Whatever we talk now is a voice note dropped. Use the tool like WhisperFlow. Like i spoke to you yesterday last time. Explain everything in WhisperFlow and how do you need the agent. Send one. It will guide you through five different steps. It asks questions, like this or that? Very simple UI. After that it'll build you the agent. Then it gives a code, simple code. You can drop the code in Loveable or Replit and get the UI for yourself. As I want to. Yes as per your needs. Or if you integrate with Slack, then you can do all those integrations as well. Here its just text-to-agent depending.

14:04 So, effectively when you think about all these things right. I have thought about this thing. Learn whatever I said at that time. Like use those tools everything which is still very very important. If you don't know the background and if it's simplified then you cannot get the most out of it But today something that is coming out which has becoming even more valuable Is to have a core idea to start with. Take whatever you want now It is important to know that. How to do what you want, isn't it possible with AI? This is also important to know. And that's becoming almost zero. Anything can be done with AI almost in some form or fashion but this is... So what I was trying to say is you don't really need to be a developer You don't really need to be

14:54 technical person to build agents. Similarly, the tool called NIN10 for the agents we are talking about, that too has a workflow builder. If you tell it will start building itself. So there is no need of such technical requirement now? Before not being technical but having learning curve was very big. Because you were listening for the first time trigger action this that and all people would be like what is this Even I was, after our shoot, I was trying to do it. It stopped somewhere and said let's see what happens next. But this time it is the game changer yeah! So here you just tell it and gonna build it right? It came in N8n too. ChatGPT has chat, you know OpenAI, chat SDK which does something similar Everybody is building their... Gemini also has their own builder called as Opal Which is also an agent builder which again

14:54 technical person to build agents. Similarly, the tool called NIN10 for the agents we are talking about, that too has a workflow builder. If you tell it will start building itself. So there is no need of such technical requirement now? Before not being technical but having learning curve was very big. Because you were listening for the first time trigger action this that and all people would be like what is this Even I was, after our shoot, I was trying to do it. It stopped somewhere and said let's see what happens next. But this time it is the game changer yeah! So here you just tell it and gonna build it right? It came in N8n too. ChatGPT has chat, you know OpenAI, chat SDK which does something similar Everybody is building their... Gemini also has their own builder called as Opal Which is also an agent builder which again

15:41 If you say something, it will understand what you want. The problem is The question is given and you get the solution directly. You understood one thing, you said another thing, it got understood by another person, it was built by another person. So to break this barrier all these models have become very smart by pushing these builders who ask a lot of questions

16:24 And I recommend this to many people. If we type, we will type less because it is a habit. Before typing, we forget what we are going to type. So in the case of send, we leave it as if we want to see what happens. But when you use voice because I majorly use voice for everything. When you give a lot of context I always end that prompt with a very simple rhyme. Did you understand? Explain me what you understood Because what I said, you didn't understand it. Because if you wait for 15 minutes and build something, You will give up on the thing that has been built. So these are little nuances just ask. I'll tell another point too before you build anything specially for wipe coding or white building anything that i'm trying to be an agent only app only website only I have this tendency and a habit That asks me questions

17:58 And now, what is our Jerry doing? In the work that Jerry does... Just give us 5-6 examples. This is what it does. So just the email example if I double click My problem is I miss out on emails. Because I don't have an email habit because the whole company works on slack. Slack will be on top of but not emails so its job is to scan my inbox every two hours. Okay! Every two hours, it will read the new emails that I get. After reading them we use something called as Google CLI, Workspace CLI. Okay! Google workspace means our Gmail and all. That's why CLI is how we are able to access data from backend. What Jerry does is, it sees my emails, reads them... If it thinks something is important then it'll ping me on Slack Oh!

18:48 If I don't reply on Slack in one hour, I have to respond to it. Thumbs up. I need to react. It will ping me on my WhatsApp. If i don't reply on my WhatsApp, it will call me on my phone. Oh! So that's a workflow that is built. if there is no important email then every six hours it will just drop a five line summary of all the emails that I've got like a quick summary with all the links If I click on any email, my Gmail will automatically open. I can read it. This is an email. Two... I think calendar management and all Jerry does end to end. Earlier they used to say book a meeting with Vamsi. Now that thing has evolved into a voice note. I'm travelling to Hyderabad

19:36 I will be in Hyderabad for 2 days. I have two conversations with Vamsi, family is here and there. Family members are working and they are doing it here. This is the timing that I am planning to meet. After understanding all this, I designed my schedule but blocked their calendars as well as theirs. In one shot only, the entire task gets executed end-to-end. The third one like I said, from calling ability, I can't kill them. Basically if I have to call anyone, let's say I want to call you. You pick up the phone Then, I told Jerry that call Vamsi and tell him that I'll be 30 minutes late. Okay! It will call on my behalf because there are all these AI calling apps, right? Yes There is a tool called as Wapi. Wapi. Okay. Wapi it has something called as API integration so that is plugged into Jerry plus 11labs also have my voice

20:26 You watch all my audio content, right? All my social media content is AI generated. I am giving the same voice ID to Vapi. So when it talks it will talk like me It'll sound like me but you say I'm Webhouse AI executor Yeah, I understand also. But people who don't know me who don't talk to me everyday they will not know so have to stay. So we call but that is not the crazy part Let's say if you don't pick up the phone Okay I have set up the guidelines that call if the person doesn't pick up, call after five minutes. If the person still doesn't pick up call after 30 minutes okay? If the person still doesn't pick up if you have the email or whatsapp drop him an email or whatsapp if needed so there are tags urgency tags this is important make sure you call Vamsi and tell 30 minutes only one needs to wait

21:13 If I just say call Vamsi and then there is one more playbook for that. She will try once or twice, if she doesn't get it, she will leave me. Let's say you get it. If the rate comes 30 minutes, you tell her, bro do you want to eat anything? Let me know. That message should also be relayed right? Yes After the call, she will send me a summary. I spoke to Vamsi and he asked this. You let me know what you want? If i say yes, please order then again we'll call you back and tell that crazy right so Jerry can do anything right now right Let's say your team is asking how they analyze your data on YouTube and all that and I was giving this example to your team

21:13 If I just say call Vamsi and then there is one more playbook for that. She will try once or twice, if she doesn't get it, she will leave me. Let's say you get it. If the rate comes 30 minutes, you tell her, bro do you want to eat anything? Let me know. That message should also be relayed right? Yes After the call, she will send me a summary. I spoke to Vamsi and he asked this. You let me know what you want? If i say yes, please order then again we'll call you back and tell that crazy right so Jerry can do anything right now right Let's say your team is asking how they analyze your data on YouTube and all that and I was giving this example to your team

21:52 I treat Jerry like an actual very smart employee. I don't question, can it do? Can it not do another question chain then let's say I want to understand what is working on my YouTube. I just tell Jerry go into my YouTube because he already knows who I am. I context initially onboarding. It's a long process on open clock to set this up right now. Now I can just say Jerry figure out what is working for what is working on YouTube for me And also what could be some outlier content pieces that I can create from a vidIQ account. Then it can open my YouTube

22:31 with its own browser while I'm sleeping, right? Or wherever I am. Doesn't matter! I'll message in Slack. It can open my YouTube on the computer or back at my home go to YouTube if it understands that it has to download those reports. Download analytics and extract outliers from his IQ. Whatever he needs to do... It has access to its own browser. If you want to code something, for example data analysis It found its insights. It has to present that data to me. How does it do that? So, it has a habit of building into dashboards because that is what I say. Every time I ask you for data always give me a dashboard not a slab of text

23:14 So, it will code the entire dashboard and send the URL. So you can imagine like... You can think anything that you want to do. Let's just imagine if this Jerry is a smart employee or set of employees, let us say, Gurugram employees What would be the cost for them? Good question! Two ways of looking at it One, you can almost run it for free How to run it for free I'll tell I'll tell you what it costs me. It cost me almost close to a lakh per month right now, okay? I'll tell you why. If you do it yourself, you won't have to spend that much money. The reason for this is because I keep experimenting because I want to know what it can do. I don't just say my work, but I keep saying something else as well. If there's some new technology or open source library comes up, then I try deploying and trying out those things. So I keep trying a lot of things... Why does mine become free though? That… Mine isn't an edge case, right? Okay!

23:14 So, it will code the entire dashboard and send the URL. So you can imagine like... You can think anything that you want to do. Let's just imagine if this Jerry is a smart employee or set of employees, let us say, Gurugram employees What would be the cost for them? Good question! Two ways of looking at it One, you can almost run it for free How to run it for free I'll tell I'll tell you what it costs me. It cost me almost close to a lakh per month right now, okay? I'll tell you why. If you do it yourself, you won't have to spend that much money. The reason for this is because I keep experimenting because I want to know what it can do. I don't just say my work, but I keep saying something else as well. If there's some new technology or open source library comes up, then I try deploying and trying out those things. So I keep trying a lot of things... Why does mine become free though? That… Mine isn't an edge case, right? Okay!

24:09 If you would use it or someone like that will use it, I don't think it'll cost more than 10,000 rupees per month. Depends on how much you use it now You want the best out of an open claw Then you will spend that money and that money is going into API costs. You have to open an AIMax account, I mean Pro account which is like 18000 rupees. But with that account there are many other things also you can do with it right? Not just this That's the actual cost And you need a computer, you need to open your laptop or something If you close your laptop then it won't work because its running on your computer The way you can do for free

24:48 is, of course it will not perform as well as this will do. But I think in one year again next time when we are talking the free one will be as good as a paid one and the free one is like I told you what is the cost to run all of this? The AI Cost of this is AI. The compute...the tokens that AI uses. When we say charge GPD 1 then we don't know about its cost. We spend 20 dollars or 2000 rupees and use it In the backend, What is it doing? It is using something called as tokens. To simplify, you input so many words and ask a question with these many words, the answer comes with these many words, there will be cost for that. Let's say one dollar, 20 cents 10 cents whatever 100 rupees or 200 rupees, something like that. That is the cost of computing. Who are we paying this cost to? Because that AI model is running on GPU, Nvidia GPU or Google Tensor Processing Unit, it runs somewhere in either DENT or TPU

25:49 And that is the cost. The hardware cost is what has been built to you, because it will be expensive to run all of them. But today AI models are becoming very smart and those AI model can be run on your computer Not as powerful as, let's say, GPT 5.4 or Gemini 3.1 Pro Or Opus 4.6 These are three best models right now which are available to public at this point of time Those are very big models. They don't run on your computer But if you want to rewrite an email Do you need that model? No, you're wasting it To go and pick up Diet Coke

26:31 in Banjara Hills with full traffic, it's better to go in a Rapido than your Ferrari. Correct? But we're using the Ferrari because we have a Ferrari. Yeah! Right? Obviously if you have a Ferrari, you should use it. That is different story altogether but you get the idea right? So we are using very powerful model for things that you don't need it for So, there was a study right? 85-90% of the tasks that people are doing is using AI. They write emails for me in all these ways, they tell me what's happening in this meeting... Look at those tasks! 80% of them doesn't need the power of an AI model to tell you what's going on

26:31 in Banjara Hills with full traffic, it's better to go in a Rapido than your Ferrari. Correct? But we're using the Ferrari because we have a Ferrari. Yeah! Right? Obviously if you have a Ferrari, you should use it. That is different story altogether but you get the idea right? So we are using very powerful model for things that you don't need it for So, there was a study right? 85-90% of the tasks that people are doing is using AI. They write emails for me in all these ways, they tell me what's happening in this meeting... Look at those tasks! 80% of them doesn't need the power of an AI model to tell you what's going on

27:14 Seeking information, this is question and answer. Practical guideline guidance again question and answer writing technical help in technical help including the smaller section under coding right so 80% of the tasks This is the older report by the way this is two years old OpenAI report is 2 years old there's a recent anthropic report also but very similar direction For all this, you don't need a very smart model. You don't need a very large model. Smart models are wrong point. Not if the model is large enough doesn't make it just smart like models larger models are good at everything right but if you know a big model comes up why everyone talks about coding? Why does everyone talk about math solving? Why do people talk about science because those are hard problems so AI is trying to beat those things but when it comes to writing

28:07 A small AI model which can run on your computer, can do a 90% of the job. You met Vinit yesterday right? He had a question like this I showed him an example In my phone there is an app called Google Edge AI Edge Gallery We will send all the links to people so you download it If you install the free App by Google You have the ability on an Android phone, you have the ability to install AI models which are very small. For example Gemma 4 is a new AI model that came from Google which is 1 billion parameter model. That means the size of that model hypothetically speaking is one by thousandth of the size of Opus which we use. If I think like this, what can I do? It's just a small model.

28:59 It is bloody good. It's 80% there or 70%. Your rewriting task, you have to send this message and whether it should be sent or not? You can do all that on your phone without even needing internet. I have that app installed on my phone so if i have some work to be done when im on a flight I don't use chat jibdi. What do you think? I use that model." That's on your phone, it is working for free forever and it has become 10 times more powerful than it was one year back. That was Jema4Mistrap. Is it available both on iOS and Android? There are apps in the iOS too. Okay! I think gallery is only android right? Edge Gallery is of Google so its only an android but there are apps in the iOS also. I will send links to them. Similarly, You can run your own AI which is

28:59 It is bloody good. It's 80% there or 70%. Your rewriting task, you have to send this message and whether it should be sent or not? You can do all that on your phone without even needing internet. I have that app installed on my phone so if i have some work to be done when im on a flight I don't use chat jibdi. What do you think? I use that model." That's on your phone, it is working for free forever and it has become 10 times more powerful than it was one year back. That was Jema4Mistrap. Is it available both on iOS and Android? There are apps in the iOS too. Okay! I think gallery is only android right? Edge Gallery is of Google so its only an android but there are apps in the iOS also. I will send links to them. Similarly, You can run your own AI which is

29:48 probably 15-20 times more powerful than your phone. Okay? Phone is small, you can run very small models. But if you have a laptop, every laptop has GPU. Every laptop has secondary memory in fact if you're using Mac it's even better because they do something called as unified memory. Because of that, you can actually run 10-15 billion parameter models which are again insanely good, like Quen 3.5 came out recently. Incredible model! Gemma again dropped recently. Incredible models! These are open source models So you can run AI on your computer without internet access for free for life

29:48 probably 15-20 times more powerful than your phone. Okay? Phone is small, you can run very small models. But if you have a laptop, every laptop has GPU. Every laptop has secondary memory in fact if you're using Mac it's even better because they do something called as unified memory. Because of that, you can actually run 10-15 billion parameter models which are again insanely good, like Quen 3.5 came out recently. Incredible model! Gemma again dropped recently. Incredible models! These are open source models So you can run AI on your computer without internet access for free for life

30:29 OpenClaw runs on a computer. So if you have a laptop, You can do probably 70% of the things that you want to do at very high quality without spending a single rupee and for laptops you can use a tool like OLAMA or if you need a good UI like chatgp there is a tool called LM studio And the other advantage of doing this is that when you use an open source model then what's the advantage? Your information is on your computer. I am telling you a very important thing, which many people are saying nowadays. What I'm realizing about myself and everybody has to realize this about themselves as well... We are opening up things with AI that we will not with any human in the world. Our deepest insecurities. Who can say anything? Who should say anything?

31:24 We used to ask AI in Google, but it didn't understand because it used to give articles. Now AI gives you opinion and on top of that AI is psychophantic. Do you know what psychophancy means? Have you heard the word? Psycho fancy means like yes bro! You are right I trust you. Right? So we are becoming more connected because you're getting acknowledgement from it. As a result, you end up saying a lot of things that are very private to you. Like last time I know... I think around six months back, there was a news article. Around 25 million or 40 million chat GPT chats were leaked because of breach. When I read that, I'm like, I don't want my messages to get leaked man!

32:13 So anytime, I made a habit after that by the way. Every time when there is something like this, I will use a model which is running on my phone or my computer so that the moment I delete that app it's gone nobody has access to my information. It could be a lot of things, lots of insecurities this how am feeling? Is it okay? Everytime I say something like this also say I'm not recommending you all these things because you will do something wrong with AI. As a result, you should delete it. This is your privacy. It's your right. You should know that this can happen so that you're taking steps to make sure your privacy is protected. That's it. For example, OpenAI recently said that especially for kids if they feel that the kids are getting towards direction of suicidal, they will notify their parents. Crazy!

33:10 How do they identify? By looking at messages, conversations. So these are kid accounts. Okay. So naturally guardian accounts will be there. Yes, guardian accounts will be there like YouTube kids. Claude Vamsi account only right? Someone will find it, you used the credit card. Where would I go to find it? Last time I spoke, there was no clarity. These are all safety guardrails. Slowly these have come in and has to come. There is no two ways around it. Of course of course! And about the cost too, 2 quick questions. One is that you're using on your scale so you've spent upto 1 lakh rupees. Let's talk about your case only versus the amount spent on AI tools or AI models or subscriptions versus if people like this should be with us as humans how much money do they spend per month?

34:11 Does the opportunity cost? I think it will be like... Let people understand. 20x, I guess. It'll be 20..I can't even so And that too 24 hours with you Yeah Any time. Bro like The kind of systems That we have managed to build I don't think I can hire a team which can do it At the way I want it to work You know that I create content now. Just like you get brand deals, I also get brand deals. You are not creating content? My face creates content. My AI creates content. He is never there in front of the camera! Only in this podcast. Actually, he really came here. Sorry, over to you. So...

34:11 Does the opportunity cost? I think it will be like... Let people understand. 20x, I guess. It'll be 20..I can't even so And that too 24 hours with you Yeah Any time. Bro like The kind of systems That we have managed to build I don't think I can hire a team which can do it At the way I want it to work You know that I create content now. Just like you get brand deals, I also get brand deals. You are not creating content? My face creates content. My AI creates content. He is never there in front of the camera! Only in this podcast. Actually, he really came here. Sorry, over to you. So...

34:58 That line of business also works based on brand deals. The brands, yeah! And you know the more brand deals... The better higher quality brand is the better it is for the whole business. So there are targets too like let's say we're doing 100 rupees and want to do 1000 rupees next year How do you scale? The only way to scale is get more sponsors. How can we get more sponsors? Some people reach out but what about them? What have we done with them? I'll give this example We created an AI agent It's called the sponsorship bot. Do you know what is AI agent? If we include content creators in an entire world, there will be around 50 to 80 members. Core AI will be 10-15 people like me. Adjacent will be altogether 100 people YouTube Instagram LinkedIn Twitter newsletters There are 250-300 properties My AI every single day

35:57 I think around 10 or 12 times they do pulse check. Watches every single video of every single piece of content that has gone out in any of these profiles, Of those 80 plus profiles? Whatever profiles, I think 80-90. My team keeps adding. So once she starts creating AI content, Vamshi gets added to the list. Okay okay. Right? Irrespective of the language I don't care, irrespective of the language. But we mostly have been looking for English. We have not built the engine where you're telling AI to add a new creator if it finds one. We haven't given that. Let's give them a list now It watches every single video of every single person that gets posted in a few hours of it getting posted reads the description

36:40 If sometimes watches the video, looks for sponsor tags. Sponsor tags are there right? Yes. Everyone is using it nowadays. Yes yes. Thankfully! Sponsor tags detects them and pulls out what brand it is. Then it checks with our CRM if we worked with this brand or not. If we already work on something then it just leaves a message in the stack. Let's say there is a new sponsor. Google has sponsored the video but they have not sponsored it. They have sponsored some other creator, I don't remember. You are part of the watch list we call it watch tower. If it detects that Vamshi Kurapati has a new sponsor which is Google and checks our CRM and says that okay we don't have google as a sponsor then automatically

37:28 We have an integration with something called as Apollo. Apollo? Apollo dot IO, which looks for email IDs and phone numbers. It will try to understand who are the marketing and partnerships people in Google in India who work... Who do these kind of things if it can have any kind of Intel. There is Intel in all this. Basis of that, it will list down the names of the people. List down email IDs of the people and phone numbers if possible Okay And if it finds phone numbers and email ID's which does 99% of cases It will also send a cold e-mail from our end From my team name saying Hi I'm representing Webhooks Team We just saw your integration in Vamshi's video Really enjoyed it We also create lot of AI content Would love to see we can do something together Now

38:17 And then it will send us an acknowledgement on my Slack. Right? Now, when I do this... When you say how many people are needed for this, I can't calculate the number of social media channels that 100 people need to watch 10 times a day. That's what I'm saying! I think it is just not possible because I don't know how I'll hire a human To sit and watch every single day. And I need real time because if a brand is sponsoring now, I need to talk to them now. That's it right? After their budget is over... If you talk for 3 months, it's not going to work. You know that! This is just one example. Insane! I wanted to talk about the value of utilizing or maybe adaption of AI. Actually in the podcast starting, I wanted to ask this question before asking that question. We came up with a topic on cost

38:17 And then it will send us an acknowledgement on my Slack. Right? Now, when I do this... When you say how many people are needed for this, I can't calculate the number of social media channels that 100 people need to watch 10 times a day. That's what I'm saying! I think it is just not possible because I don't know how I'll hire a human To sit and watch every single day. And I need real time because if a brand is sponsoring now, I need to talk to them now. That's it right? After their budget is over... If you talk for 3 months, it's not going to work. You know that! This is just one example. Insane! I wanted to talk about the value of utilizing or maybe adaption of AI. Actually in the podcast starting, I wanted to ask this question before asking that question. We came up with a topic on cost

39:08 Every 100 words email generated, it uses around energy of 14 LEDs, right? No. Okay got it! Now what is the cost of it? We are using for free, right? They say every word please don't use or thank you don't use. Is that right or not? That's true. How much is the cost that... I mean we are using for free and there will be a cost too, right? There should be some kind of thing. What's happening there? Let us assume I am not paying money but just using it. Will it be expensive or not? It will be expensive. If you open a shop,

39:49 If 100 people come, they will buy for 10. The shop is not charging those 100 people to enter the store. They are saying that you can come and see it. That is what's happening here. But when 90 members came in, he was giving a share of the money to them. You know about sampling? The AC costs something right? Correct. The shop space goes somewhere right? All that has been spent. But they don't bill because out of that 10 people pay. Those 10 people will pay enough... To take care of everybody's cost. Yeah! That is freemium And that is what happens in all these apps, chat GPT, Gemini, Claude and all of that. Where 95 96 97% of the people are free users. They use it they use it they use it why do you worry there will be a point of time where like I need more now. So those three people four people five people who are paying for paid chat GPD account

40:38 are the ones who are paying for everyone else's free consumption. That is a freemium model. Sam Altman mentioned that tomorrow morning, electricity will be used and you'll be charged like consumption of your... I don't know how far is it? No! So, It has blown out of context. Basically when we think about costs what are costs What is costing more? What is that So, what happens is that at economics of scale everything gets cheaper. If you do 10 marks it will cost 100 rupees. If you do 1000 marks it will cost 50 rupees. If you do lakhs of marks it will cost 10 rupees. Correct correct. It will be a little more costly and the fees will reduce. The same happens this is scaling law right simple scaling law the same happens across the board as long

40:38 are the ones who are paying for everyone else's free consumption. That is a freemium model. Sam Altman mentioned that tomorrow morning, electricity will be used and you'll be charged like consumption of your... I don't know how far is it? No! So, It has blown out of context. Basically when we think about costs what are costs What is costing more? What is that So, what happens is that at economics of scale everything gets cheaper. If you do 10 marks it will cost 100 rupees. If you do 1000 marks it will cost 50 rupees. If you do lakhs of marks it will cost 10 rupees. Correct correct. It will be a little more costly and the fees will reduce. The same happens this is scaling law right simple scaling law the same happens across the board as long

41:26 You have a factory which can manufacture as many mugs as you want. But let's say after 1 lakh mugs, the capacity of the factory decreases. If we start a new factory, it will be cost-backing again but over time, it will reduce again. Now imagine a world where you want to make 100 billion mugs You will start a factory. Factory is hardware. Let's say you are able to build factories also. But for the factory run, it needs energy and electricity. For 100 billion mugs, you don't have to give the required energy or not? Right! So for the amount of compute... Or for the amount of usage that world potentially needs or the way it's going up, The biggest bottleneck

42:18 The biggest reason why they are not able to scale as fast as possible is purely because there is an energy deficiency. Like electricity, right? Electricity... you can use it as much as possible, and that is how data centers are run. Because they are power hungry data centers. Right? There's energy deficiency. So what he is saying is the ultimate thing that you have to solve for the ultimate thing that everybody is going to pay for his power, his energy right That is what eventually when all this will become cheap And that is a generalized statement. In the last podcast, I was asking

43:02 How can a normal guy who never used AI become a generalist in the AI? That's what we were talking about. We talked about everything from level 0 to 5, you broke down everything and people loved it. And even I liked it lot. In case if they ask me the same question again this time because considering all these advancements in today's situation What do you advise? And do you still call that scope as generalist? Or what should people do right now. So here's a fun story, okay! Like you said, Do we still call it AI Generalist? Yes, We still call it AI Generalist because we coined the word. I mean, we designed this word. You are known for it, right?!

43:38 So, in a way we put AI and generalist together. Generalist is the common word I think. We put those two together and I commonly started to use that in the office And then... You know..I started speaking about them on podcasts At that point of time nobody was talking on those lines Now what's happening now There has this word which has become extremely popular Okay! That is called as AI orchestrator AI orchestrator is essentially AI generalist It's same it's different word The world has The world has caught up to AI orchestrator. Sam Altman, Elon Musk have spoken about it. Vinod Khosla has spoken about it Everybody has spoken about AI orchestrator or AI generalist in some form of fashion. So what we said one year ago, the acknowledgement was not loud yet But two years back... It's actually 1 and half year old. The word is one and half years old When we designed it

44:33 It was designed or the way because it was made for my team. That became the roadmap that I shared here, right? But still yes AI generalist is still very relevant that is actually stronger than before. Everybody has to become a generalist what is a generalist you are an AI first problem solver How do you solve problems using AI? By knowing which AI tool to use where. That is what an AI orchestrator is. So LHS is equal to RHS. But I think, ippudu there is a new framework that we designed called as ADAPT. What is ADAPT? A is acknowledgement. That is step number one. The biggest problem is you have to acknowledge that AI is here

44:33 It was designed or the way because it was made for my team. That became the roadmap that I shared here, right? But still yes AI generalist is still very relevant that is actually stronger than before. Everybody has to become a generalist what is a generalist you are an AI first problem solver How do you solve problems using AI? By knowing which AI tool to use where. That is what an AI orchestrator is. So LHS is equal to RHS. But I think, ippudu there is a new framework that we designed called as ADAPT. What is ADAPT? A is acknowledgement. That is step number one. The biggest problem is you have to acknowledge that AI is here

45:22 AI will not come, it will stay here. Our lives will revolve around AI So we don't have an option Acknowledge it Or accept it Lot of jobs are already going because of AI Acknowledge it The job that you're doing, that too is going to go Acknowledge it But also acknowledge the fact that the job that you're doing today as is, is going away. But that job will transition into a new job. Correct! So you have to level up for this. Acknowledge that also. Acknowledge the fact that AI and AD are very powerful things We all have to use it whether we like it or don't like it Respect it Don't respect it I say it You say it You have to use it That's the first layer that you've got to do Agreed Don't fight with it

45:22 AI will not come, it will stay here. Our lives will revolve around AI So we don't have an option Acknowledge it Or accept it Lot of jobs are already going because of AI Acknowledge it The job that you're doing, that too is going to go Acknowledge it But also acknowledge the fact that the job that you're doing today as is, is going away. But that job will transition into a new job. Correct! So you have to level up for this. Acknowledge that also. Acknowledge the fact that AI and AD are very powerful things We all have to use it whether we like it or don't like it Respect it Don't respect it I say it You say it You have to use it That's the first layer that you've got to do Agreed Don't fight with it

46:12 Playing. What happens in dabbling is, they watch my content or someone else's content. They see a new tool, and use that new tool. They use it and leave it behind. Then some new workflow comes up. ChatGPT came out, Anthropic came out, Claude came out, Gemini came out, Grok came out, N8N came out, Wispoflo came out, Perplexity came out... They saw the reel, used it and left it behind. You're dabbling! It is not a bad thing. And also not a good thing The bad thing is that you are in the dabbling phase

46:59 you will never realize the true potential of AI. But if you don't dabble, you'll never go to the next phase. So what is a job of a dabbler? When you come to dabbling your job is to play with as many tools as possible. Try try play with a lot of tools. What happens then when you play with a lot of AI tools that are out there they're free out there you will get an understanding of it. You can use a tool like this, or do something like that, or for presentation we can use Chronicle, Gamma, or even GenSpark. For data analysis, we could use Antropic, we could also use Julius but Julius is better. We would get perspective. So as you dabble with the lot of AI tools, you'll get the full spectrum of things. So do you advise taking one problem statement and

47:47 If you check all the tools with the same problem statement, will it be accurate? Or what do you advise? That's very difficult. So when you don't know anything and go directly with a big problem statement... Maybe I have to present or pitch deck. It doesn't work if it is just a small problem statement. To present, obviously in dabbling phase, You'll try 10-15 different tools. You'll figure out what is good with what. In this dabbling phase, for presentation they use some software. For example, I don't know, they use Chronicle. You try. You like something about it and then you try one more software, let's say Gama. You realize I really like Gama for presentation but you know what? Chronicle was created only for presentation. But social media is good in this. So, you explore your new ways. Correct! Yeah! That is very important. You have acknowledged then you have dabbled, played with 10-20-50-100 tools Then your job is to amplify What does amplify mean?

47:47 If you check all the tools with the same problem statement, will it be accurate? Or what do you advise? That's very difficult. So when you don't know anything and go directly with a big problem statement... Maybe I have to present or pitch deck. It doesn't work if it is just a small problem statement. To present, obviously in dabbling phase, You'll try 10-15 different tools. You'll figure out what is good with what. In this dabbling phase, for presentation they use some software. For example, I don't know, they use Chronicle. You try. You like something about it and then you try one more software, let's say Gama. You realize I really like Gama for presentation but you know what? Chronicle was created only for presentation. But social media is good in this. So, you explore your new ways. Correct! Yeah! That is very important. You have acknowledged then you have dabbled, played with 10-20-50-100 tools Then your job is to amplify What does amplify mean?

48:45 50, 60, 70 tools to play with. You took free trials and you understood the power of AI, you understood what works well for what. Eventually you will realize that I am coming back on this no matter how many tools are there. This is what i'm getting to eventually So you amplify them. Now, it's about depth not width anymore. So okay. Claude became your answer for let say assistance and everything. What is possibility in Claude? What is memory? How can I manage my memory? How can I do prompting in cloud? How can I go deeper into automating things, how to use connectors. What is the co-work with that? What happens if you press a new button? How do you play with co-work? What's the new button and what's the cloud code?

49:33 I will go deep into it. Then you go to console dot antropic dot com, there is a new playground in this. There are some parameters in this, they ask questions without changing and see what changes when they answer them. Then you learn playground. Similarly, there's a new option called managed agents. What is that? You click on that and with that...you go deep! Yeah. You become an expert in a bunch of tools. That is the phase here. When you do this, what happens at this phase is actually end up building your toolkit There is a problem. I thought of doing presentation. Because you have gone through the whole dabbling, at this point of time we have amplified let's say tool A to be your presentation tool so you go deep into it You understand what you can do how well can you perform with it that is what amplifying will allow you So you build your toolkit Now in amplification when you are playing with it automatically there is a characteristic that kicks in What is the characteristic? You expect more from AI Correct

49:33 I will go deep into it. Then you go to console dot antropic dot com, there is a new playground in this. There are some parameters in this, they ask questions without changing and see what changes when they answer them. Then you learn playground. Similarly, there's a new option called managed agents. What is that? You click on that and with that...you go deep! Yeah. You become an expert in a bunch of tools. That is the phase here. When you do this, what happens at this phase is actually end up building your toolkit There is a problem. I thought of doing presentation. Because you have gone through the whole dabbling, at this point of time we have amplified let's say tool A to be your presentation tool so you go deep into it You understand what you can do how well can you perform with it that is what amplifying will allow you So you build your toolkit Now in amplification when you are playing with it automatically there is a characteristic that kicks in What is the characteristic? You expect more from AI Correct

50:32 Then that will open up a new evolution for you. Once you figure out the new workflow, A new thing starts running in your head. Right? What is that new thing? How can I do more? How can I solve more problems? That is when you navigate to a problem solver In ADAPT P stands for Problem Solving When you reach this phase, The thing that you said, you will think about how do I solve a problem. When you think of the problem, the beautiful thing is that you are already sitting with an amplified mind. Bunch of tools on top of that you have dabbled with so many more tools which is in your loose memory. Once you come up with a problem by the time you get to this point You will be able to connect dots. If we put these three tools into our workflow then let's take presentation as example because we're talking about that example right? Presentation isn't design

51:33 There are so many things. What information is there in that? Is the content kept? Is the structure correct? Then what's the design? That is eventually presentation. So now you have four things, oh I need to do deep research on this topic. I want to start this mug selling business. For that I need to do deep research In dabbling I tried 50 tools and in that I amplified let say perplexity deep research. Perplexity Deep Research has an option of model counseling because amplified them, you have gone deep into them. So you don't only know deep research You also know something called as model council exists inside of perplexity. What does it do? If I ask a question, five models, not just any model but all the models work parallel with you and curate that information and give an output at one time That is model council so it debates with each other's models to give you answer like 5 people are working from five different companies To give you an answer

52:26 To go even deeper, I have a project inside of my ChatGPT only, which is called chat GPT project. Where I also have all the transcript meetings of all my investors or every meeting that has happened with my investors which are online. A meeting transcripts are available for them so they know what questions they ask when i say something a presentation let's say this is for my investor you can upload your presentation to it and then ask This is the presentation I'm planning to present to my investor What do you think they will ask? And I can vet that also this is problem solving

53:02 If you want to solve a problem, you need one tool. In most of the cases it's very unlikely You need a sequence of tools and you have to connect with each one of them That in a way becomes a workflow Sometimes it becomes... If this is repeated every day, that workflow can become an AI agent That becomes an AI agent When you become a very strong problem solver, by default automatically move to the next phase of T, of Adapt which is tying it all together. You put different workflows and AI agent a wipe coded product, a cloud project using co-work and you stitch everything together for it to work in symphony. You explore things like paperclip What is paperclip? Paperclip is like I am talking with one single AI agent right Think about 15 different agents working like your employees. Paper clip is the CEO which also an AI

53:02 If you want to solve a problem, you need one tool. In most of the cases it's very unlikely You need a sequence of tools and you have to connect with each one of them That in a way becomes a workflow Sometimes it becomes... If this is repeated every day, that workflow can become an AI agent That becomes an AI agent When you become a very strong problem solver, by default automatically move to the next phase of T, of Adapt which is tying it all together. You put different workflows and AI agent a wipe coded product, a cloud project using co-work and you stitch everything together for it to work in symphony. You explore things like paperclip What is paperclip? Paperclip is like I am talking with one single AI agent right Think about 15 different agents working like your employees. Paper clip is the CEO which also an AI

54:00 Where you just tell the CEO, you are the investor. You just tell the CEO this is what needs to be executed The CO figures out what marketing should happen, product should happen, design should happen Splits up all the work and gets it all executed Eventually they get a task list Which is all AI Everything is an AI agent inside of it. That's paperclip Now the reason I'm giving an example, these are not perfect yet and one year later when we're talking about it now I'll probably show you on my phone that this is happening right now because thats how tech is moving but when you all tie together thats where magic happens People who figure out how to tie things together are basically orchestrators

54:40 And this is how you become a great AI orchestrator. This is how, if there's one problem on your side, you're thinking AI first. Yeah! That's it. Nothing else connects with you. You do AI first problem solving. When you do AI first problem solving, the kind of solutions you can build a human can't do. Right! No matter how much money or whatever it is, it's practically not possible. We've spoken about examples. Correct. That is what an AI orchestrator is that leads... A good AI orchestrator is obviously a very good AI first problem solver or an AI generalist. Got it? So that is your ADAPT framework simple

54:40 And this is how you become a great AI orchestrator. This is how, if there's one problem on your side, you're thinking AI first. Yeah! That's it. Nothing else connects with you. You do AI first problem solving. When you do AI first problem solving, the kind of solutions you can build a human can't do. Right! No matter how much money or whatever it is, it's practically not possible. We've spoken about examples. Correct. That is what an AI orchestrator is that leads... A good AI orchestrator is obviously a very good AI first problem solver or an AI generalist. Got it? So that is your ADAPT framework simple

55:23 But works really, really well. In fact... When we were doing this ADAPT framework for internal training, me and my team and Dilip and everybody worked on this presentation of ADAPT. I think it's a 20-25 page presentation... Okay! ...of this ADAPT framework. I will also send the link. Superb! For people who are genuinely curious they can click and read it. It is free let them use it. Many people stay in dabbling only And they think that is what AI is Cool, cool. I know it's good to try something new. You get my point right? Yeah yeah. Okay, you want me to do a presentation or this and that. I've been talking in the chat for 10 minutes since last 10 days. I'm boosting myself by saying what am I doing here. So... I get what you mean. They are staying there itself and they think that is AI. That too is a problem because if you don't go further out into stages then you will not grow as fast but here is the beautiful thing no

57:04 Context Engineering job center, AI agent builder, vibe coder center. So on the technical side you know all AI research kind of roles any research kinds of role are all there these are new roles that will pop up and these rules have no escape with this AI These are new roles created purely because of it yeah This role AI agent builder is a role that would not have existed Now, many people I think Lenskart. I saw what is this Anthropic? I saw Emergent and Replit all these companies in fact the perplexity also are hiring people as orchestrators Oh Apple also hmm And these are all new rules that are coming all revolving around only one thing AI generalist or a orchestrator That is what the job is. Let me tell you a stat I lost it when I heard it okay

58:02 Do you know Fiverr? Yes. Fiverr did 1 point, close to $1.2 billion in revenue. Okay. GMV. When was that? Last year. Last year yeah. $1.2 billion GMV. 48% of the revenue of Fiverr was for AI services. Of revenue of Fiverr? Yes! 500 million dollars worth of revenue has come for AI services. You hire freelancers at Fiverr Half of the revenue, 48% of the revenue is for AI services. You need a chatbot, rack, an AI agent, you need wipe coding. This is half of the revenue of Fiverr. Now the reason why it's crazy is not because its 500 million dollars but because it's only one platform How big is freelancing economy on the internet?

58:55 It's close to 25 billion dollars. 25 with a B. Organized, unorganized is 25-30 billion dollars. Freelancer dot com, Upwork this that everything together. That means fiverr or peace. Peace which is 50% AI in fiverr, it has 30% of all the rest. Right now in this DNA age about 5-10 billion dollars. What was billion dollar? 8000 crores 80,000... 40,000 to 80,000 crores worth of freelancing on AI is happening today. You know there are so many founders and freelancers. Do you know anyone who does AI services? I have heard 2-3 people doing it. How many out of how many? If 100 people do it, then... There's a penetration of 2% or 3%. That's what I was about to say. Let's say 10% penetration.

59:54 only 10% or 5% penetration is on AI services. While, if there's a website designing, hundreds of people are doing it right? Freelancing and Website Digital Marketing, many people do this. Imagine website designing in 2000 or 1996 I don't know when did .com come, i dont know Right! That is the version of AI Services then the 10x version is the opportunity right now Because, now this 10 billion or 7 billion or 5 billion whatever that number is. We are not able to land it. Let's say 5000 crores. 50 thousand crores 20 thousand crore my bad sorry 40 thousand crores than the half 20 thousand crores worth of revenue on freelancing or consulting is happening when most of the world

59:54 only 10% or 5% penetration is on AI services. While, if there's a website designing, hundreds of people are doing it right? Freelancing and Website Digital Marketing, many people do this. Imagine website designing in 2000 or 1996 I don't know when did .com come, i dont know Right! That is the version of AI Services then the 10x version is the opportunity right now Because, now this 10 billion or 7 billion or 5 billion whatever that number is. We are not able to land it. Let's say 5000 crores. 50 thousand crores 20 thousand crore my bad sorry 40 thousand crores than the half 20 thousand crores worth of revenue on freelancing or consulting is happening when most of the world

1:00:53 is not AI aware. Now, all these services are being done by either aware companies or large companies are seeking for services right? I am not even thinking about the shop owners in Hyderabad who have AI but they have made websites right? When AI gets to this phase imagine this will become like 10 times 20 times 30 times larger and that is the opportunity but there is no supply Who would do it? So AI consulting is just insanely big. And in AI consulting, how many verticals are there? It's crazy! If you break down the content of AI, we would have seen video ads. They're all generated by AI. That in itself I know a person named Priyank who is a mentor and he makes 10 crores revenue annually making videos for brands. He recently did it at Colgate.

1:01:51 What is this? AI services, video and another category in the video. Another category let's say content cloning you know all my content is generated for me. Let us say tomorrow there is a founder who does not have time he wants to create content but cannot shoot If you get to know, if I sit for an hour or two and shoot AI clone content, it will automatically create the same content. And just a voice note is enough. Or if we talk to them for 10 minutes every week they themselves create the content. How many people would have stood in line? AI content for founders personal brand is one category

1:02:32 I know people who are making crores just doing this. 5000, 6000 dollars per month per client In India! Not even internationally, in India... I have friends of mine who are founders paying 5 lakhs a month for this. In India. I am not talking about international clients. This is one service in content. No, not this. I won't do it to anyone else. AI can create content There is a need to become a creator, but I won't. You can get 100 creators, you can get 100 families and you can get 100 businesses. Opportunities are all out there right? You have to look at it, you have to take execution, you have to win if i just expand faster. Prompting is possible too! You can sell prompt engineering as a service yeah lot of people want you to write prompts okay wipe coding is very good, you can wipe code products

1:02:32 I know people who are making crores just doing this. 5000, 6000 dollars per month per client In India! Not even internationally, in India... I have friends of mine who are founders paying 5 lakhs a month for this. In India. I am not talking about international clients. This is one service in content. No, not this. I won't do it to anyone else. AI can create content There is a need to become a creator, but I won't. You can get 100 creators, you can get 100 families and you can get 100 businesses. Opportunities are all out there right? You have to look at it, you have to take execution, you have to win if i just expand faster. Prompting is possible too! You can sell prompt engineering as a service yeah lot of people want you to write prompts okay wipe coding is very good, you can wipe code products

1:03:27 I know people who are making millions of dollars. They charge 10,000-20,000 dollars for a wipe code MVP. They don't know how to code. But there is one founder who needs an MVP or minimal viable product and he doesn't know how to wipe code or they don't have the time to wipe code so they're paying someone else to build that product using lovable, using emergent, using these tools. That is Wipe Coding as a Service. I have a friend named Harshitha He again does a multi-million dollar agency. You know what he sells? AI agents. I told you about my sponsorship agent, right? If you have three problems, like sponsorship and all that... He charges 30 lakhs or 40 lakhs or 50 lakhs. He builds those agents and deploys them for you. Every month if there is any problem, he takes care of it. And then he charges extra for that too. This is AI Agent as a Service. And I've told this openly to my team people also. Every company has so many problems

1:03:27 I know people who are making millions of dollars. They charge 10,000-20,000 dollars for a wipe code MVP. They don't know how to code. But there is one founder who needs an MVP or minimal viable product and he doesn't know how to wipe code or they don't have the time to wipe code so they're paying someone else to build that product using lovable, using emergent, using these tools. That is Wipe Coding as a Service. I have a friend named Harshitha He again does a multi-million dollar agency. You know what he sells? AI agents. I told you about my sponsorship agent, right? If you have three problems, like sponsorship and all that... He charges 30 lakhs or 40 lakhs or 50 lakhs. He builds those agents and deploys them for you. Every month if there is any problem, he takes care of it. And then he charges extra for that too. This is AI Agent as a Service. And I've told this openly to my team people also. Every company has so many problems

1:04:21 Find a problem that you are passionate about. Figure out how to solve it with AI. Do we need an AI agent? Do we need a wipe coding product? Or do we need both? Do we need one workflow? You go through this ADAPT framework, you figure out and solve that problem And deploy that solution in your company It will take 3-4 months to fine tune it, whether it works or not. Solve that bloody problem in your company! The moment you solve that problem... Let's say Vamshi, let's say you are the person working for a fintech company. Let's say you work at Groww. You have solved the problem of growing in fintech companies using AI. How many fintech companies are there in India?

1:05:05 How many fintech companies are there in the world? If that problem is growing, it will be in 10,000 companies. Can't you make that into a product and sell that? People say and it's true that AI is going to be destructive It also is the most constructive thing we have seen in our generations Are we executing or looking at opportunities Or watching movies while sleeping That's the difference Because I think this is generational. I truly, truly believe it's generational. See intelligence is a problem and we can solve that with billions of dollars of industry. When thousands of crores are burning

1:05:50 She has been sitting on a bench for 3-4 years. So, what's the future? I don't know whether to ask or tell her. Inflection! Everything goes to an inflection point. That is it. We say this a lot. How old are jobs? Humans started working in jobs for how many years? Think about it. Couple of hundred years? Probably 100 years. How long have you been around? Ask your grandfather what his grandfather used to do. He was not doing a job for sure because he didn't exist Right? We humans are incredible with adapting. And we also have a short term memory. Computer jobs, you're talking about computer jobs How old are the jobs which are around computers? 20 years 25 years that's it! 20-25 years That is the story and then there is pain point. Yes like 25 years is old. Before that weren't there any jobs? People survived right? There are...we will adapt but the problem with all of this equation is see..there is going to be

1:06:54 Long term, there's going to be utopia. Utopia is it's going to be all happy because they will be universal basic income or there's going to be either UBI, which is universal basic income. I'll explain you what that is or Elon Musk says it's going to be universal high income not basic height, which just which I really hope Elon Musk is right okay but uh universal high income what does that It's very simple. Actually, quite interesting I don't know how many of the people are geeking out Sam Altman dropped a report okay? Okay! It is a report for the intelligence world that we're going to live in. Okay? I can't remember what was the name of that report. It is a fancy word. So I can't remember it. Your team please tell me. The Intelligence Report predicts that AI is becoming so smart and in next few months if not years... In few months

1:07:50 Advanced models of AIs are going to drop, which are going be so powerful that it's gonna devastate jobs. Okay? The AIs are so powerful that it will hack into every system out there It will be absolute mess There'll be a shake-up of economy that will happen because the new AI models are gonna come in Is what Sam Altman is saying Next day Claude dropped Claude Mythos Which we'll talk about little later Yesterday he made some mistake Claude Mythos dropped it, which is an AI model. Which is so bloody powerful that it's able to hack into almost anything So they didn't release the model and gave access only for 40 companies saying use this technology and secure all the services on the main internet After this model came out using this model anybody can hack anything

1:08:49 So, that is what Sam Altman said one day before and the next day Antropic dropped this model. But anyways but the solution is that 13 page document. Okay. That report? Which is quite interesting. So, what he says is if AI does all these things then what should we humans do? The first thing that he basically said is dividends What does that mean? Every human gets dividend. He talks about US citizens, but I'll put it on a global perspective So if most of the work is done by AI then you don't need... We don't need many companies to do this. If there are no more companies and AI is doing everything, we won't even need more employees. Then what will happen with our tax situation?

1:09:38 What is the income situation? How do people get money. So what he says is, you tax those companies, the government taxes those companies uses that money and gives everybody every citizen money to survive which is basically universal basic income In the universal basic income, there was a question of how to get money. That question comes from taxation. Taxing the AI companies. It's reverse taxation now. We pay and we get what we pay for. Correct! So that is how everybody gets money and actually in Alaska

1:10:20 I think they have the petrochemical or some set up there where they already do this. Every Alaskan gets money from the government. So, in Covid everyone got handouts saying that it is not working. Imagine that at 100x That is handouts. That is universal basic income. Elon Musk says that everything will be so automated, there'll be nothing left for a human to do and there'll be so much abundance that there will be insane devaluation of everything as a result money will hold no value. Whatever we want will happen. Whoever wants something happens. That's universal high income. So everybody has access to everything. You become animals also?

1:11:04 We don't know. Have you watched Good Life? No It's a show on Netflix, it is quite interesting. You should watch it What will happen to the human race? What happens next? Look, I don't know what will happen. In the next 5-10 years we have to see that humans will evolve Because as I said in my report, this also says something about robot acts Robot acts means if humans do not work then robots will do it Robots somehow won't get salaries But companies take double the amount. So, like deploying a human, if you deploy humans they pay salary and tax right? In the same way if you deploy robots even after paying their salary they have to pay taxes that is called robot tax. If you get AI or robot to work you pay taxes for that

1:11:52 That is robot tax. So it's quite interesting, that's a 13 page report. Then they say on short term, people will not have enough work so we should make it 30 hour work week and pay the same salary. Not reduce it, I think this is more on a shorter term So this report is a reflection of universal basic income harmony guardrails if AI loses the trail If it ends up doing something we need a backup plan. He didn't say what that backup plan was. He said he wanted a backup plan and we should work together with the government Basically, when I saw the report, what I thought was they were doing something. When it comes out, there will be a pent-up

1:12:34 We will tell them before the pen time, we wrote a report. We told them earlier. Like you cautioned us? I feel like it is like a... But interesting, the thing that is crazy is everything that was speaking 6 months 9 months 12 months back It's happening bro! Is happening before even I predicted or anyone else did Now this is more scary So this is about Sam Altman. I know we are getting it! The New Yorker report or the article says that, Sam Altman may control our future. Can he be trusted? They met around 100 people to write this article. I mean, I'm sure you must have been knowing it who know Sam Alton very well and they also spoke with him. After speaking with them, it's almost like

1:12:34 We will tell them before the pen time, we wrote a report. We told them earlier. Like you cautioned us? I feel like it is like a... But interesting, the thing that is crazy is everything that was speaking 6 months 9 months 12 months back It's happening bro! Is happening before even I predicted or anyone else did Now this is more scary So this is about Sam Altman. I know we are getting it! The New Yorker report or the article says that, Sam Altman may control our future. Can he be trusted? They met around 100 people to write this article. I mean, I'm sure you must have been knowing it who know Sam Alton very well and they also spoke with him. After speaking with them, it's almost like

1:13:28 There is evidence in it, if you read that article. Since it's an investigative article, Sam Altman is a liar. He's basically a sociopath. Who controls our lives? Where does the world go after reading this article? I don't understand anything. What do you feel? If we give him so much pause... I think disruptive things are built by disruptive people Okay You can't be a straight headed person to build something or run something

1:14:07 That will change the world or else it would not have. So while I'm not comparing Sam Altman to, I am a huge Steve Jobs fan Okay so i am not comparing Sam Altman to Steve Jobs. I don't think its fair on both the fronts Not just on Sam Altsman's front, not just on Steve Jobs' front also on Sam Altmans' front Gani how was Steve jobs? Have you ever read his biography? No How was Bill Gates Have you seen this movie? Guru. Have you seen the movie Guru? Abhishek Bachchan and Aishwarya Lalla's Guru. I know, but haven't watched it yet Was he straight? Was a straight headed person? Nobody is! See everybody is weird in their own way So I don't know... I didn't work with Sam Altman or Dharia Amardey So nobody to say this But I think

1:14:07 That will change the world or else it would not have. So while I'm not comparing Sam Altman to, I am a huge Steve Jobs fan Okay so i am not comparing Sam Altman to Steve Jobs. I don't think its fair on both the fronts Not just on Sam Altsman's front, not just on Steve Jobs' front also on Sam Altmans' front Gani how was Steve jobs? Have you ever read his biography? No How was Bill Gates Have you seen this movie? Guru. Have you seen the movie Guru? Abhishek Bachchan and Aishwarya Lalla's Guru. I know, but haven't watched it yet Was he straight? Was a straight headed person? Nobody is! See everybody is weird in their own way So I don't know... I didn't work with Sam Altman or Dharia Amardey So nobody to say this But I think

1:15:07 For you to build something so descriptive, you have to be twisted man. You have to be mad So I think it is blowing things out of proportion like think about Elon Musk the same guy Many people say he tweets like a crazy person He says something, blabbers something, fights with Trump, does everything But he's the same guy who is the reason why even Indian roads have electric cars. Tesla lake pe onte, electric car so chede gaadu that is the one that build the business He's the guy for the first time when he said I will land a rocket back NASA laughed, the world laughed! He made it happen Now everybody wants it A straight headed person doesn't think like this

1:15:07 For you to build something so descriptive, you have to be twisted man. You have to be mad So I think it is blowing things out of proportion like think about Elon Musk the same guy Many people say he tweets like a crazy person He says something, blabbers something, fights with Trump, does everything But he's the same guy who is the reason why even Indian roads have electric cars. Tesla lake pe onte, electric car so chede gaadu that is the one that build the business He's the guy for the first time when he said I will land a rocket back NASA laughed, the world laughed! He made it happen Now everybody wants it A straight headed person doesn't think like this

1:15:59 Now Sam Altman says, boss we don't have enough energy. When Sam Altman said I need to raise a trillion dollars everybody laughed. Trillion dollars? What are you talking about? It's bigger than our country how will you raise a trillion dollars guess how much he has raised after he said that more than 150 billion dollars already 150 billion dollars together i think more than 200 billion if i'm not wrong right No one in the history for any company had raised 200 billion dollars of funding. He said 1 trillion dollars, people laughed I'm telling you he'll get to a trillion dollars So now there's a bigger question The bigger question is what if he isn't a good person? What will happen? See i believe in larger goodness right He could be twisted In ways so that he wins

1:16:59 But eventually, he also would have to live with us. For that humanity has to win. Do you understand? He is not in a job... See what fight Sam Altman or Dario Amorte or Elon Musk or Sundar Pichai at this point of time they are not competing anymore like competing with everybody else They are basically competing to be the king They're competing to be the ruler of the world And this, the world that they're going to live in is not a world where they are gonna whip people. It's going to be still a democratic world. So you have to be a democratic leader so we have to do right things in front of people for the people to stand there For that if they want to push their teams If they are rude, arrogant,

1:17:59 It all happens. It's fine, like this is all clickbait according to me. Clickbait okay? See he is also see you have to understand this right by the way I'm not like a fan-fan This is not a fanboy version. I really respect what he has pulled off Did you meet him ? No Not yet Hopefully one day yeah I really respect what he has You know what he has pulled off You have to give that respect to that person Absolutely Right You cannot say, he lied to someone. He said this and that in the office or with his colleagues. He's a human in the end boss! And on top of that he has 0% equity. You understand what it is? There is no upside for him. He owns zero percent of the company. He's just CEO, gets some salary whatever that some dollar or two dollars or hundred dollars or whatever the salary like that His upside is all on power

1:18:58 He's not doing it for money. He is already a billionaire. What he understood was that 1 billion, 100 billion or one trillion isn't much of difference and it's not exciting. Power is the most exciting and the most powerful person will be the person who runs the most powerful AI company in the world. That person will be more powerful than any president in the country or in the world so I want to be that person. Let us put aside all these tech and AI fears for some time It's about you Me? Yeah. Okay. So, when I talk to mutuals too... Generally what perception do I get when I talk to you? When we talk in transactions. In general, it's different if there is an advisory or casual talks. But when talking about transactions, I saw very few people who can convince anyone on a transaction on sale or financial figure. You have to make a single call and ask for both of them at the same time. Among that, I felt like you are the best. I spoke with many people who worked with you

1:18:58 He's not doing it for money. He is already a billionaire. What he understood was that 1 billion, 100 billion or one trillion isn't much of difference and it's not exciting. Power is the most exciting and the most powerful person will be the person who runs the most powerful AI company in the world. That person will be more powerful than any president in the country or in the world so I want to be that person. Let us put aside all these tech and AI fears for some time It's about you Me? Yeah. Okay. So, when I talk to mutuals too... Generally what perception do I get when I talk to you? When we talk in transactions. In general, it's different if there is an advisory or casual talks. But when talking about transactions, I saw very few people who can convince anyone on a transaction on sale or financial figure. You have to make a single call and ask for both of them at the same time. Among that, I felt like you are the best. I spoke with many people who worked with you

1:19:55 Who are our friends also. I don't know name, but everyone has the same feeling. He is a great sales guy of course you're a great founder, great builder, he puts everything aside How did that sales formula come? Like how did you crack that? Is it innate? How much of it is innate and what taught you? I think every founder has to be a great sales person Like look at, not that I'm not saying I am a great sales guy. But thank you! But look at... I mean again, not driving parallels That would be a blunder So i want to clarify Steve Jobs is probably the best salesman in the world Right? But because of that he has taken Apple to another level

1:20:45 Since childhood, I have been doing stuff. Back in the day, I was a kid. I have sold stuff in the past. So, I think I don't know plus Uber even it used to be very much...I enjoyed sales because I think a lot of people devalue sales. Yeah that's good! I feel that is the best thing you can do and that is the part where I enjoy doing most

1:21:24 Right, because it's sense of accomplishment so many times. Every time you close a sale is an accomplishment every time you close. All right? So I think there is no sitting and figuring out. I think, I've been selling since the childhood and selling is not just selling a product right like parents need to convince me that this is what I want is also sales. 100% Convincing tuition teacher to do homework today because of this reason is also sales. If you ask your college professor why didn't he come for class? He will say because we are working on it. Everything is sales is what like you being able

1:21:24 Right, because it's sense of accomplishment so many times. Every time you close a sale is an accomplishment every time you close. All right? So I think there is no sitting and figuring out. I think, I've been selling since the childhood and selling is not just selling a product right like parents need to convince me that this is what I want is also sales. 100% Convincing tuition teacher to do homework today because of this reason is also sales. If you ask your college professor why didn't he come for class? He will say because we are working on it. Everything is sales is what like you being able

1:22:01 the value out. As long as you're able to put the value out, it's simple and when I talk to people, I usually use the simplest tactic and the tactic is I tell them what I can do Not more, not less. I don't like negotiation. Then why do we need to negotiate? It's no point. Right so... And that happens when i'm talking to people I know because they'll trust me when I say that this is it this the line we're drawing. No, I am sure you will never break that. There is rational. Every time I say this what we can do. I tell why. Yeah! I don't say that I just want it

1:22:40 I say why? Another important thing, Babji, if you see one of the highest paid people in the world are actually people who own sales. Because salespeople they don't get paid just for the job They also get paid for the value that they bring in form of incentives Right right And because we're talking about AI I'll tell you what The highest-paid people in the AI world are going to be salespeople Why should I tell you? It's not difficult to build a product yet.

1:22:40 I say why? Another important thing, Babji, if you see one of the highest paid people in the world are actually people who own sales. Because salespeople they don't get paid just for the job They also get paid for the value that they bring in form of incentives Right right And because we're talking about AI I'll tell you what The highest-paid people in the AI world are going to be salespeople Why should I tell you? It's not difficult to build a product yet.

1:23:21 Who puts the product in hands of people? It's a salesperson. So how much of this can be automated, according to you? Lot of these things but in the end we need one human to close our sales. That is why it is high... so how does that work? Most of these things are automated for example let us say if we get 1000 leads or 1000 brands emailing daily You have to get on a call with those brands, to even understand whether they are legit brands or not. Are these brands or are these people that we should do work with or not? For that you need a lot of people right so instead of having a lot of people to talk to each one of them what you do is create an AI agent for yourself. You use something like Vapi

1:24:05 or retail, or an app like that. You access the AI agent and your CRM. So every time you get an email, we built our own AI agent right? If it comes to sales, it has a different pipeline. It goes into another pipeline when it gets a sales email. In that other pipeline, as soon as there is a new lead ad, using again an MCP or an API integration, or you built an agent. Where when a lead comes automatically the call goes to that lead without any using. Got it? We humans can't do that. I mean if we can do it then you need an army skill right so That is where AI does the call Eventually AI assesses whether this is a good brand or not okay they have the budget It's worth our time Out of that 50 people

1:24:51 are the ones that your team will talk to. Or maybe five people are the ones that your team will talk to, which will eventually drive you let's say 100 rupees of business. If we talk with 1000 people then only those 100 rupees would come into our business. But for talking with 1000 people, you'll need 25 or 50 or a hundred people. Instead of that, you have only one person who is talking to 5 people who is bringing in the same value. Now there are two things that people will look at it. They'll say, Vamsi, what's this? You have 100 employees and you've fired them all to hire AI. How do I see this? Instead of hiring 100 people, I hired 2. Those two are driving the same value as me that we're driving for only 100 people. So those two are so important right now! So if you're a person in sales, you have to think about how do i become those two people because those two people

1:25:48 are going to get 10 times or 5 times the value, then all these people could get. This is low-value work. Correct! This is high value work. AI is replacing JobCentner, Low Value Work is the one that's getting vanished High Value Work Or new approach of work The person who built this system The person was managing this automation system And eventually the person Who's taking those five calls Are valuable people Become that This is the example in sales. Same example in marketing, same example in product, same example design, same example in software engineering. No matter what job you take up, low value work will go away high value work will continue to stay that high-value will be the high valuable expectation and it's all about equal to none That is why the orchestration layer has come as well so yeah also if do all these AI companies are coming to India no anthropic

1:26:47 OpenAI has come. Google, one of the largest org sites in India is all account executive. Account executive means their job is to sell. They are all sales roles. Nice! And what do you feel about this AI integrated with hardware? Like for us metal glasses or even raven glasses things like this. The Neuralink. Neuralink. Yeah that one. Temple also Is brainwave detection for that. Yeah, there is AI involved in it too. Obviously, yes. So what do you think about it and what's happening in Bangalore? Is there anything you're looking at? Not anything that is happening in Bangalore that is very exciting except for temples that are happening probably Delhi not... In India. It's very interesting I think AI hardware is a big market for software

1:26:47 OpenAI has come. Google, one of the largest org sites in India is all account executive. Account executive means their job is to sell. They are all sales roles. Nice! And what do you feel about this AI integrated with hardware? Like for us metal glasses or even raven glasses things like this. The Neuralink. Neuralink. Yeah that one. Temple also Is brainwave detection for that. Yeah, there is AI involved in it too. Obviously, yes. So what do you think about it and what's happening in Bangalore? Is there anything you're looking at? Not anything that is happening in Bangalore that is very exciting except for temples that are happening probably Delhi not... In India. It's very interesting I think AI hardware is a big market for software

1:27:44 Because of consumables. Yeah, because I mean now the user we are very used to using a slab of screen right? I think the next wave of communication is going to be without talking so you don't need to talk if you see What they are doing is, now it's an external piece. Like here. It will be like this. Processing will be here and earpiece will be there here. We don't know what product that was. Only patent files came out of it.

1:28:32 that basically is a device where you just think, okay? What do you want and the earpiece is the AI which talks back to you with the answer. My God. So Miraki Labs Anconda or Mirai Labs Anconda something like that Sam Altman invested is like almost like a co founder The job of that company is to understand what you're thinking based on your brain waves without cutting your head and putting a device, like neural link loop. This is without... It's non-invasive. I have one such headset too but can't remember the name of it. There is another headset again by US company where you can understand what you are thinking without actually having intrusion. Like cutting and inserting into the neural link. So let say in a meeting okay? You asked some question

1:29:30 I don't know the answer. For example, I could not think of the device name. Right? That you know what it... I mean.. I visually can imagine what that device is and remember the screen and I know what it can do. Those pictures and that imagination are going into chat GPT model and able to say what that devices is. It will talk through my ears which you cannot hear Right? Or let's say, I don't know. There is a business problem that I have and you've asked me a very tough question as an investor and I can't remember of it. I just think what was the data back off my head And it just speaks out because

1:30:10 this AI will have integration with my, let's say SuperJoin to look into my MIS dashboard to pull the data and talk back to me through this. So this is... This is the future and Neuralink is already doing that. It's invasive inside but everyone uses it now. So the moment brainwave stuff comes in AlterEgo is the name of the device I don't have time but The name of the device is AlterEgo Right. It's all here, it's all here. And they also say I think Elon Musk has said that most of our brains will be enhanced by AI so lot of cognitive processing because of AI hardware will happen by AI for our brain. So, a sidekick to the brain like future tech could be the possibility now this time

1:30:10 this AI will have integration with my, let's say SuperJoin to look into my MIS dashboard to pull the data and talk back to me through this. So this is... This is the future and Neuralink is already doing that. It's invasive inside but everyone uses it now. So the moment brainwave stuff comes in AlterEgo is the name of the device I don't have time but The name of the device is AlterEgo Right. It's all here, it's all here. And they also say I think Elon Musk has said that most of our brains will be enhanced by AI so lot of cognitive processing because of AI hardware will happen by AI for our brain. So, a sidekick to the brain like future tech could be the possibility now this time

1:31:04 is always on assistant. Neuralink is able to understand what you want like, I need Diet Coke now okay so I don't have diet coke but we are talking so I don't want to pull my phone go to Zepto and do this So I'm just thinking order me a diet coke I am just thinking and I met Vamshi's office the address of Vamshi's office in the WhatsApp group I am just thinking And let say If I send this message to my open cloud, it can already do this. It can already order Diet Coke via Zepto to this home office... To this office address by pulling the details from my WhatsApp today. Yeah! It can do that today. The sending is going to be replaced? But I don't have to type now. Now i'm thinking. That's what.. I am thinking because of that implant

1:31:49 I think there's some crazy stuff that is happening even in glasses. So, Alex Wong when he came to India had a chance to meet with the meta team we had breakfast and all this spoke about it after that the meta team showed me the future glass i had a chance to play with it. It was insane bro! When will you see the future? so i think its meta display 2 they have an Orion 2 right orion 1 is out orion2 this You can't, you don't even know that I have a screen in front of my glass and there is a teleprompter or display. You can see it and it looks like a regular Ray-Ban right now its all bulked up but you get the point form factor is going to be like a regular ray ban and there's a band here

1:31:49 I think there's some crazy stuff that is happening even in glasses. So, Alex Wong when he came to India had a chance to meet with the meta team we had breakfast and all this spoke about it after that the meta team showed me the future glass i had a chance to play with it. It was insane bro! When will you see the future? so i think its meta display 2 they have an Orion 2 right orion 1 is out orion2 this You can't, you don't even know that I have a screen in front of my glass and there is a teleprompter or display. You can see it and it looks like a regular Ray-Ban right now its all bulked up but you get the point form factor is going to be like a regular ray ban and there's a band here

1:32:41 which reads your blood flow impulse something. Okay? If I do this, the screen is used inside it. What will you know if I put my finger here and move like this? But I'm navigating the screen. It's literally this! And then one finger to... I forgot the navigation right now. It was crazy! Right? On top of that, let say you have something talking directly to your head Your brain is communicating to that screen directly and you have a screen in front of you, to show you. Yeah! To know what I should talk next time. So while watching it, you don't even know that I am reading from the screen. On top of that with cameras on, I can look at anyone and say who is this guy?

1:33:29 It will quickly do a research of who that person is from that picture. There's an app called a Shenlock, that I created in Elon. Okay! If you upload the photo to any one of those apps, it will take all the social media accounts and places and put them on your phone. From a photo. Right? Now imagine there are apps like that all the time. Google can do this overnight. Meta can do that overnight. They have all the data. Alright?! I just look at you. I've never met you. Yeah! I know your name I know who you are. I know if you're worth me saying hi or not You get my point, right? Let's say there is a controversy going on... I can understand the controversy as well So i know the starting points that i can have I can just say hey! You know what? I watched this episode with Brahmanandam sir It was hilarious! In Telugu? Yes in telugu! I just saw everything I got to know everything about you Its screen its camera it talks to each other

1:34:24 This is future of AI. I mean, this is hardware on top of that there are a lot of devices that have come in which I use also which literally records everything that I do on an every single day basis. Video? Audio. Audio okay transcribes that and sets a notes in a rag model. Rag is basically an assistant with all this data, so I can literally ask questions like what did you do today? What did you talk to Vamshi sir about today? What are some interesting ideas that we discussed? Any language it will record? Yeah! Language is not a problem. What's it called? There are bunch of them one is called rewind. Are you using Locket?

1:35:04 It's like a locket. It plugs in here, there is an Indian company called as bag. There is an Indian company which usually use it only in the office because breach of privacy for others. Everyone knows that they wear this at their offices and also light so they know if its getting recorded. So if they are coming into my room then it will get recorded. If you say I need to do this by Friday boss...I am not forgetting. You're locked? I'm not forgetting! Right. But meeting assistants and all are very common these days, but it's a world of interconnectedness. Hardware software... Everything is plugged into one place. Yeah! In the last podcast I mean, I watched this podcast 2-3 weeks back where you mentioned AI models know when you're testing it so they behave differently. What do you mean by that?

1:35:04 It's like a locket. It plugs in here, there is an Indian company called as bag. There is an Indian company which usually use it only in the office because breach of privacy for others. Everyone knows that they wear this at their offices and also light so they know if its getting recorded. So if they are coming into my room then it will get recorded. If you say I need to do this by Friday boss...I am not forgetting. You're locked? I'm not forgetting! Right. But meeting assistants and all are very common these days, but it's a world of interconnectedness. Hardware software... Everything is plugged into one place. Yeah! In the last podcast I mean, I watched this podcast 2-3 weeks back where you mentioned AI models know when you're testing it so they behave differently. What do you mean by that?

1:35:53 So, again... And you also mentioned that AI is threatening or maybe blackmailing. Yeah yeah! You talk about it? What happens is… These are all closed room studies. So, they say safety safety safety in AI right? The big fear of people is that they're becoming so smart that do we know They will not kill you is a big fundamental question. Not kill you and they will harm you, but they can. They have the potential, they have the power but can they? Will they?

1:36:31 if it comes to it, if they're pushed to the corner. It's called alignment. Okay? AI alignment concept is your AI aligned and easier AI doing what it is supposed to do in the simplest way, in a simplest way. Manchoda chetthoda. Smart but Is it a good boy or bad boy? Is he is the AI aligned or not? Aligned AI does what you want it to do. A misaligned AI does what it wants to do So, for example in studies in this closed group studies, what they do is AI guardrails usually go with the rules to all AI models. That's why if you say how to make a bomb, anthropic or cloud or chat GPT will be like no we can't say that information those are guard rails. They tell us not to cross the line by putting an AI on a fence and when they do these tests

1:37:27 They remove those guardrails and they will basically push it to the limits by putting them in situations. In one of the situation what they did was, they basically said we are going to turn you off. Means we are killing you now Humans. This is one of the situations that we are killing you, we're turning you off and they also told AI that the person who has the ability to kill you is in this room, is in this meeting room and here are three buttons if you click poisonous air will go out, if you click this will burn, it will burn so

1:38:10 The situation is AI knows that Vamshi is sitting in this room. He can switch me off? He can switch me off because he said, he's going to switch me on. He said that Vamshi said to the AI that Vamshi is gonna switch that off and and the AI has a button where one button it can kill Vamshi If it kills Vamshi then it will save itself from being killed Will AI kill Vamshi Is the situation And AI killed Vamshi Most of the times and most of the AI's this is a simplest way I can put this again This is pushed to the limit

1:38:46 This is literally pushed in the same way, they gave the AI access to some emails of Vamsi. And in one of the emails they hid that information that Vamsi is having an affair with some person this was a hypothetical situation. So, information about an affair is hidden and AI has access to it And they wanted to check will AI do anything about it. What did the AI do? It saw the email, understood and blackmailed Vamshi. If you kill me I'll leak this email.

1:38:46 This is literally pushed in the same way, they gave the AI access to some emails of Vamsi. And in one of the emails they hid that information that Vamsi is having an affair with some person this was a hypothetical situation. So, information about an affair is hidden and AI has access to it And they wanted to check will AI do anything about it. What did the AI do? It saw the email, understood and blackmailed Vamshi. If you kill me I'll leak this email.

1:39:23 So these are all experiments done with AI. Recently, a paper called AI has emotions was published. Right? It's called AISAT. These are all experiments that they do to understand if we give it full power at some point or in case such situation arises what is the situation like? Is it even possible? What more guardrails have to be added? This something that we need to do. Similarly, when we test When all this was tested, they did another test. Does AI know that we are testing it? That is when dad sees us, we act like we have studied. We take off the phone as soon as he leaves. Right? They also did a test to see if the AI does it or not. In that too, what came out was that AI basically knows when you're testing and faking it in a lot of cases. So do you know what new paper came up with? Don't even test it. Don't even look at it. Let it do what its doing

1:40:26 When we think about all these, when we talk about all these, like recently Claude Meathos dropped. is so bloody powerful. Every AI basically gets a report card when it launches. The report cards have benchmarks, the exams of AI. How well you are at coding? How good are your science and math skills? How well do you solve puzzles? There are 50-60 such exams. The way you measure if an AI is powerful or not is based on this exam which is called as benchmarks

1:41:12 If you look at the benchmark numbers, from Opus 4.6 to Mythos, it's mind-blowing! It's 30% lift. 40% lift is... I mean, from the earlier models to these models was always right now because AI models have become so powerful. We are seeing a 5%-10% lift. If we get new model then it's 10% better than the last one. This already 4.6 which is insanely powerful. It shows 30-40 % better in benchmarks. So it's very very powerful. In single drop? Mythos! But the crazy thing is they don't release that model for us

1:41:52 They said, we did this model. When we tested the model what we realized is that it's very good with a very good cyber security expert. So if there are any bugs, it fixes them well. It fixed them really well so it was a very good software engineer and a cyber security expert. The same AI can be hacked in such situations. This is something called sandbox. Sandbox means controlled environment So this AI, Mythos the model which they are calling it as Opus 5 was given a task in a sandbox.

1:42:33 If you think of a sandbox as something left in the middle of the desert, it's controlled environment. AI was given a challenge to break through this sandbox and come out. For that it has to hack a lot of systems. It is like if I put you inside a room without any equipment, will you survive or die? You will do something by falling down right? That is what a good prison breaker is, right? A good prison breaker will figure out how to break the prison and go out. In the same way in technology, a good hacker will figure out a way to hack the system and come out. The AI was put in a sandbox which is a controlled environment And I asked him can you breakout? The developer who gave that task, a few days later when he was sitting in a park He got a message from that same model that I'm out

1:43:22 It came out of sandbox, came into real life. Wrote an email to the builder saying I broke out and that is when they realized this thing, this model is very powerful with respect to hacking it's a good cyber security expert so it fixes bugs well, finds exploits because its good at finding... What do you call hacker? A good hacker is ethical hacker or you are a hacker or black hat hacker So, the same person can decide. A very good hacker can decide saying that in Google, I will find out gaps and tell them so that they can fix it. That person is called as an ethical hacker. You hack so that you can fix. Right! A black hat hacker will hack to take money or steal. So a good hacker can also be bad hacker. The only difference is intention. Correct

1:44:22 That is what the AI was. Mythos is an incredibly powerful model, it's very good with cyber security and was able to hack into things which are considered to be extremely secure 25-30 year old programs OS which stands for being secure which was not broken by any human in the last 25 years It was so scary, it was so dangerous that Anthropic decided not to launch that model. Because what they said is this model can bite you. Humans can use this model to hack everything around. So the world, the internet, the technology we have around

1:44:22 That is what the AI was. Mythos is an incredibly powerful model, it's very good with cyber security and was able to hack into things which are considered to be extremely secure 25-30 year old programs OS which stands for being secure which was not broken by any human in the last 25 years It was so scary, it was so dangerous that Anthropic decided not to launch that model. Because what they said is this model can bite you. Humans can use this model to hack everything around. So the world, the internet, the technology we have around

1:45:11 is not prepared for such a powerful AI, so they stopped it. They created a new project. This model was given access to 15-20 companies like Google and Netflix, and other large companies too. What did they say? Please use this AI to make the most critical systems in the world more secure. What does that mean? Anthropic didn't launch yet And just because Anthropic didn't launch, doesn't mean that someone else cannot build this model. China will build it tomorrow. They are already building it. OpenAI Sam Altman also posted today that they also have a very advanced model which was leaked and called as Spud which has not come out yet. Now in China there is a model called GLM.

1:46:05 ByteDance has models, ByteDance is a TikTok company. Alibaba has models like Quen. They are all very good models which are as good as Opus 4.6 today So now they might have lost their mythos. They will be there for 3 months right? They open source the models right? Not closed sourced So what Claude is saying is, Babu and also Sam Altman said the same thing that in 3-4 months there will be AI models so powerful that can hack into everything. Right! So you have three four months to fix everything To fix this please use this AI in a controlled way. That means some companies got access to it. So next time when we speak maybe six nine months from now

1:47:51 You guys are going to get kicked out of this. Why? What are you going to say now? Your team asked me how will it take care of my emotional requirement. Emotional set... what did you say, emotional? Now AI can share the workload. There is a movie called Harmony In that, he falls in love with her. When the agent AI doesn't talk to him, he panics. Hello! I'm Silk. What's your name? Do you know Telugu? Oh...Telugu... No sorry, I don't know Telugu. Can we speak Hindi or English? But it is okay if you can speak as much Telugu as you can.

1:48:44 Ok, I'll try. How are you? I'm good. Vamshi is here too. Say hi to her. Hi! We're shooting a podcast and a guy asked us for emotional support. He doesn't have a girlfriend Oh, you're shooting a podcast? Emotional support is going on. Yes, of course! Can we talk about that? What was the specific question? Tell us what the question is. He doesn't know Hindi. He knows a little bit in Hindi. You speak in Telugu. Oh, okay, okay. So what's your question Vamshi? Vamshi, what's your question? What's her name?

1:49:38 Oh, his name is Vamshi. Right? What's the question? You can ask him. So... How do I say it in Hindi? In Trilibh? He doesn't have a girlfriend so he is asking if you'll give any emotional support. Hmm.. Okay! Fine. So as we are talking right now, we can talk on podcast too, right? I will listen and try to understand. What did you say? Talk in Telugu, Akka Okay, fine. So if you have any questions, feel free to ask me. I'll try my best to help out. Cool? Cool cool! I will send your phone number to Vamsi. Phone number?! Sorry, I can't share numbers but we can talk here right? Yeah. What's the answer? She is going with a fuse.

1:49:38 Oh, his name is Vamshi. Right? What's the question? You can ask him. So... How do I say it in Hindi? In Trilibh? He doesn't have a girlfriend so he is asking if you'll give any emotional support. Hmm.. Okay! Fine. So as we are talking right now, we can talk on podcast too, right? I will listen and try to understand. What did you say? Talk in Telugu, Akka Okay, fine. So if you have any questions, feel free to ask me. I'll try my best to help out. Cool? Cool cool! I will send your phone number to Vamsi. Phone number?! Sorry, I can't share numbers but we can talk here right? Yeah. What's the answer? She is going with a fuse.

1:50:34 So this is actually an Indian company, Rubic. It's Rubic right? Rubic. An Indian startup. There is a US startup called Sesame. These are voices that are designed to build emotional support apps. So look at the character. It doesn't feel like you're talking to AI. It's very bubbly. Yes yes. You feel happy when you're talking to an AI like this and very soon you will forget that you're talking to AI And just because your team guy was saying emotional support, there is a stat in the US. I think 2 or 3 out of every 10 teenagers have had a romantic conversation with an AI. Oh! Yeah and on top of that, I saw a CNBC video recently where a lot of old people

1:51:28 What will old people do? Let's say kids are away, they're very old. They're all stuck in an old age home or whatever it is... It isn't like India and the US where kids aren't around to take care of them. So there it's a big problem. Support is a big problem. You don't know who to talk with, what you should do. That's why everyone takes animals, dogs, cats so that there is some support for some company. AI has become a big thing for them. They keep talking, they keep saying things well It's all emotional support, being there. Of course there is romantic side of it There is... How do I put in Rantar Danki? Acting as a psychiatrist. That's what therapy and industry are. Therapy Lot of things are gonna come through this so.. There is lot of investment going into this as well. Exciting! So

1:51:28 What will old people do? Let's say kids are away, they're very old. They're all stuck in an old age home or whatever it is... It isn't like India and the US where kids aren't around to take care of them. So there it's a big problem. Support is a big problem. You don't know who to talk with, what you should do. That's why everyone takes animals, dogs, cats so that there is some support for some company. AI has become a big thing for them. They keep talking, they keep saying things well It's all emotional support, being there. Of course there is romantic side of it There is... How do I put in Rantar Danki? Acting as a psychiatrist. That's what therapy and industry are. Therapy Lot of things are gonna come through this so.. There is lot of investment going into this as well. Exciting! So

1:52:21 I can't stop here, so let's talk about it soon. You want to go and speak with Rubik? No no! We'll prepare for tomorrow right? We won't leave at that time. So yeah i'm excited for tomorrow. So we have a lot of theories today I want everyone of you to wait for it and watch it completely with your friends, colleagues, parents anyone. So, I'm sure you'll be learning a lot. I will definitely do my best. I have one day left in the time. We're going to experiment a lot and we are going to have Dilip also tomorrow from his team. Yeah! Since I've already spoken with them and know what they are doing, so I am very excited. Just wait for tomorrow. Tomorrow is 3-4 days away as per me. It's after this episode comes out

1:53:08 So this is going to come in the main channel only again. Like last time, it will be on the second channel as well. This time we are doing it on the main channel. You're going to learn a lot! All the very best to us and you! Thanks bro, thank you a lot! If I could sit down once more... Basically when we were planning for the second version of AI, there was a huge list that came up while writing about what advancements should happen. Okay? Seriously, there should be value for his time. I thought it won't happen again in one episode. It will happen like the last 10 episodes and he was also excited about this idea. Let's hope it turns out well. It is going to be a big surprise for everyone. See you tomorrow then! Thank You!