‼️⚠️ MISSS AVVADDUUU… AI AGENTS, AUTOMATION AND THE FUTURE | Ft @vaibhavsisinty Raw Talks With VK |

0:00 So this man is the man who scares me the most. If I'm asking you why should I watch this podcast? Good question. Two ways of looking at it. One, people who don't use AI. Two, people who are only using it. What to use, where to use, where not to use is what we'll try to tell you here. Perfect. Think about Elon Musk. But he's the same guy who's the reason why even roads have electric cars. electric car. So, he's the guy when he said, "I will ran the rocket back." NASA laughed. The world laughed. He made it happen. A straightheaded person doesn't think like this. How was Steve Jobs? Have you ever read his biography? How was Bill Gates? For you to build something so disruptive, you have to be twisted, man. You have to be mad.

0:52 jobs like what kind of jobs can come the highest paid people in the AI world are going to be sales people who puts the product in hands of people it's a saleserson this is low value work this is high value work become that the meta team showed me the future glass I had a chance to play with it we are opening up about things with AI that we will not with any human in the world. Our deepest insect

1:39 so I wanted to know. So >> it doesn't feel like you're talking to a AI. 500. Now this is more scary. This is about us. So this man is the man who scares me the most. So of course he gets me into a reality which which is actually a new world

2:26 altogether. So again intent is never to scare for sure. I know. I know. >> My intent has always been important. Now, what you do with that information is up to you. >> Of course. >> So, I am hoping if if you think I'm scaring people, let's do more solution oriented stuff today then. Okay. Solution oriented. We can talk solutions. >> Of course. Yeah. Absolutely. One question I wanted to ask you because What's the intent behind you coming to Hyderabad all the way for two days, two long days almost six hours

3:11 for the viewer who who's watching generally I wanted them to sit with a book and pen or laptop or just note down everything in their phones. >> So if I'm asking you as a layman who's watching it, why should I watch this podcast? Not just this, the second one also shoot maybe in a week's time. So why should someone watch I I genuinely I'm telling on the screen please over to you maybe the reason according to me is one is that if you're ambitious and you want to win because we are living in a AI first world >> right and lot of people

3:57 AI actual reality is if you want to not fail in this world there are two kinds of people who will definitely fail in this AI 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 100 times and smarter than everyone else you're around. Right? If you're not using it, >> then you're not going to succeed to people who are only using it. No AI, then also you'll not succeed. You'll fail. Yeah. So knowing the difference

4:42 between what to use, where to use, 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 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 yeah I I've joined last time so I thought this time I'll shout it out again maybe because I'm I'm getting a lot of information from there >> almost uh follow whenever I open that I I joined the WhatsApp group basically

5:30 that's how we generally explore so this time I thought I'll it's open to anyone it's free do it for free so what we'll do is like uh we'll put the link down let everyone join let's see I want you to pause and click the link in the description and let's let's see the comments if you have joined because uh Shalom where do we get the updates ma'am actually staying ahead downtown >> the name of the community we call it staying ahead it's right now the largest probably the largest AI community >> we'll be the total community size is 15 million people >> right on WhatsApp we recently started purely for India because India is all on WhatsApp >> uh so on WhatsApp we have five lakh

6:17 people close to >> in 4 months >> WhatsApp started 3 4 months M right 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 >> in the easiest way possible my answer was h you have to read Twitter you have to read you have to do this you have to do that and nobody was happy with the answer I was like and we made it it's that's why we made it free >> crazy >> right so I think everybody has a short head staying up to date. No must join. Do join and comment down. Thanks man. 10 months in the last episode both

7:02 virtual and 10 months. Let's talk two years late. Let's talk one year late. Almost one year. Tell me what what's happening now. What happened? What's happening around the world? How are you looking at this this fantasy? Yeah, I think AI was like a I think easy language smart intern today. It has become like the probably the one of the smartest employees that anyone can have. Right. So uh on a technical language

7:51 which is chat assistant right >> today it has gone to doing things for you so it has agents 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 physical world I'm talking about digital world >> 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 3 months

8:38 >> we have grown probably more than what normally would take a year or in just last three months or two months. It's just insane. 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's be more and let me be a layman egent. How do we differentiate or define for a person who doesn't know this agent agent workflow management automation or so if we just talk about those three terms key meaning so basically first is

9:25 your AI assistant which is chat GP gemini claude grow that is basically right This is the assistant >> right now. I'll tell you about what an AI agent is. Okay. Email agent. >> The email agent is always on. It's like >> Jerry. >> Jerry. Yeah. >> Like even the way Jerry evolved is just insane. We will talk about it if you want to. The kind of things it can do right now is just wild. >> Right. >> Right. Because of new technology that came into the world called as open clock. >> Yeah. >> Right. Uh but uh before we get there agents anidi >> it does work asynchronously like it does on the go

10:12 >> example >> example let's say email the example >> this is what I'm thinking the reply should be write the email but if it's an email agent let's say if you built an agent >> agent furnace that 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 access every time there's an email that is coming to you. 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. If it is not so important and let's say like uh some fan he said great content love your content it's a nice piece you

10:58 reply saying thank you so much uh for being nice you automatically reply you send the email let's sayach I'm saying let's say you want it to be automated completely an agent is automating some big sponsor let's say Google wants to work with you >> it's not going to reply saying huh okay we can work this is my payment this is a this is a price you pay me there it will no agent okay partnership or any this kind of very important email comes you don't reply to that emails >> right right saying that there's an important email, please get back to it.

11:43 Or let's say your team manages it. >> A team member, partnerships managing that this is what we want you to solve for >> or there is an email which is let's say someone 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 prompt agent. >> So it'll work like a 24/7 employee for you, >> right? >> That is an AI agent usually which doesn't usually need a trigger. >> The trigger is automatic. The trigger is automatic. It's like a 24/7 working employ. So one quick question how much

12:38 most of the things that were built were no code drag and drop 9% that has become even simpler >> I think we built a landing page or whatever also last time that was wipe coding >> now incidentally just a few days back Claude >> dropped something called as manage agents managed agents okay >> like literally you probably might not have even come across >> right anthropic website consoleanthropic.com which is the login layer of anthropic not you will have managed agents just like your vibe coding you just say that hey I want my emails to be managed end to

13:24 end by an AI use a tool like whisper flow like I spoke to you yesterday last time it will guide you through five different steps question very simple UI after that it'll build you the agent >> then it'll simple you can do all those integrations as well that you want to here >> it's just text to agent so effectively when you think about all these things right uh I have thought

14:09 about this saying >> like use those tools everything which is still very very important background simplify simplify 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. >> Yeah. And that is also becoming like almost zero. Anything you can imagine 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 a technical person to build

14:56 agents. you're listening for the first time trigger action this that and all people will be like what is this >> even I was I was trying to do it but this time it is the game changer yeah so just you just tell it and going to build it right chat GPT has chat uh you know open AI chat SDK which does something similar >> everybody is building their I mean uh Gemini also has their own builder called as Opel which is also an agent builder which is in uh text to agents pretty

15:43 much >> right >> right so everything is converging zapier can also build agents has almost become like text to agent >> right so LLM have become so smart that the only thing that it is doing really well these days is asking you follow-up questions which is important So to break this barrier, all these models have become very smart by pushing these builders to ask you a lot of questions. Right.

16:34 So, but when we use voice because I majorly use voice for everything. You can describe what you want at length. 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 so these are little little nuances just ask before you build anything especially for VIP coding or V building anything that I'm trying to build agent app I have this tendency and a habit that

17:19 ask me questions >> before You build it. You ask me questions. >> Questions. I again ask. This is it. Don't you have any more questions? When I know it is more complicated, I wanted to ask more questions because humans used to plan much better than what they are doing right now. As a result, we are not spending that time to actually plan. Correct. And just just five this is what it does.

18:06 >> Huh? So just the email example if I double click my problem is I miss out on emails. >> Okay. Because know email habit because the whole company works on Slack. M >> so Slack I'll be on top of but not emails. So it job is >> to scan my inbox every 2 hours. >> Okay. >> Every 2 hours we use something called as Google CLI workspace. Okay. Google workmail is how we are able to access the data from the back end. If it thinks something is important, >> then it will ping me on Slack. >> If I don't reply on Slack in 1 hour, I

18:51 have to respond to it. >> Thumbs up. React check. >> It'll 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 6 hours it will just drop a fiveline summary of all the emails that I've got like a quick summary and with all the links I can read it this is email to uh I think calendar management and all does end to end that that thing has evolved into a voice note. I'm traveling to Hyderabad.

19:43 This is the timings that I'm planning to meet. The third one like I said calling basically if I have to call anyone. Okay, >> let's say I want to call you and tell him that I'll be 30 minutes late. Okay. >> It'll call on my behalf because uh there are all these calling AI calling apps on no >> there is a tool called as WP. Okay. >> WP it has something called as it has API integration. So that is plugged into Jerry plus 11

20:31 voice. >> So when it talks it will talk like me. It'll sound like me but it'll say I am webhub say I executed. >> Yeah I'll understand also you'll understand but people who don't know me who don't talk to me every day they will not know. So I have to say >> so it'll 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 5 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 in an drop him an email or WhatsApp if needed. So there are tags, urgency tags. This is important. Make sure you call until 30 minutes. >> If I just say calls and say then there's one more playbook for that. Okay.

21:24 Do you want to eat anything? Let me know. Oh no. You let me know what you want. If I say yes, I want to eat. Please order. Then again, it'll call you back and tell you that. Crazy. Okay. Right. So Jerry can do anything right now, right? Uh let's say and I give I was giving this example to your team. I treat Jerry like an actual very smart employee. I don't do it. >> Let's say I want to understand what is working on my YouTube. >> I just go into my YouTube because it already knows who I am. I context initially onboarding HSN. No.

22:11 >> Right. It's a long process on open claw to set this up. Right 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 my Vid IQ account >> then it can open my YouTube with its own browser >> while I'm sleeping right or while wherever I am doesn't matter then Slack message it can open my YouTube on the computer on my at my back at my go to YouTube. If it if it understands that it can it has to download those reports analytics download whatever it needs to do >> it can it has access to its own browser

22:58 for example data analysis data it found its insights it has to present that data to me >> so it has the 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 Next. So, >> so you can imagine like you can think anything that you want to do. >> Yeah. Let's just imagine if this Jerry is a smart employee or set of employees and employees. >> What would be the cost of them? >> Good question. Two ways of looking at it. Okay. You can almost run it for free. >> Okay. I run it for free. I'll tell you what it cost me. It cost

23:43 me almost close to a lakh per month right now. Okay. I'll tell you why. I keep experimenting. Mhm. >> Because I want to know what it can do. >> So I keep trying a lot of things that mine is an edge case. Okay. >> Right. I think if you would use it or someone like that will use it, I don't think it'll cost more than 10,000 rupees, 5,000 rupees per month. >> Depends on how much you use it. Now you want the best out of uh an open claw then you will spend that money and that money is going into API cost >> open max account

24:30 which is like I think other things also you can do with it right not just this That is the actual cost and it's running on your computer. The way you can do it for free >> is of course it will not perform as well as this will do. >> Mhm. But I think in one year again next time when we are talking the free one will be as good as the 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 comput the tokens that AI uses >> and we use it

25:19 in the back end what is it doing It is using something called as tokens. Simplify words let's say $1 20 whatever that is the cost of comput because that AI model is running on a GPU Nvidia GPU or Google the tensor processing unit >> and that is the cost the hardware cost is what is being built to you But today AI models are becoming very very smart and AI models computer

26:05 not as powerful as let's say GBD 5.4 or Gemini 3.1 Pro >> or Opus 4.6 Six. These are three best models right now, right? Which are available to public at this point of time. >> All right. Those are very big models and computer to go and pick up Diet Coke in Bar Hills with full traffic. It's better to go in a rapid than over your than your Ferrari. Correct. >> But we're using a Ferrari because we have a Ferrari. Yeah, >> right. Obviously, if you have a Ferrari, you should go use it. That's a different story all together. But so we are using a very powerful model

26:52 >> for things that you don't need it for. >> So there was a study right 80 85 90% of the task right look at those tasks. 80% of them doesn't need the power of the AI model seeking information. This is question and answer guidance again question and answer writing >> technical section >> right so 80% of the tasks this is a older report by the way this is 2 years old >> uh openai report is 2 years old there's a recent anthropic report also but very similar

27:37 >> direction you don't need a very smart model you don't need a very large model is a wrong 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 because those are hard problems so AI is trying to beat those things but when it comes to writing a small AI model which can run on your computer >> can do a 90% of the job Right. >> Yesterday I he had a question like this. Okay. Example phone

28:22 >> uh Google edge AI app only edge gallery. >> Okay. >> Right. We will we'll send all the links to people so you can download >> edge gallery app free app by Google install. >> You have the ability on an Android phone you have the ability to install AI models which are very small. For example, Gemma 4, >> okay, >> is a new AI model that came from Google which is a 1 billion parameter model. The size of that model, hypothetically speaking, is 1 by,000 of the size of Opus. >> Oh, which you use. It is bloody good. It's 80% there >> or 70% there.

29:14 I have that app installed on my phone so that if I have some work to be done when I'm on a flight, I use that model. That's on your phone. It's working for free forever and it has become 10 times more powerful than it was one year back. That was Gemma 4 which >> both on iOS and Android. iOS app. Okay, >> I think Gallery is only Android, right? Edge gallery is of Google, so it's only Android, but iOS you can run your own AI which is >> probably 15 20 times more powerful than your phone. >> Okay, >> phone is small. You can run very small models. Laptop laptop has every laptop has GPU.

30:00 >> Every laptop has secondary memory. In fact, if you're using Mac, it's even better because they do something called as unified memory. Danala you can actually run 10 15 billion parameter models which are again insanely good likew incredible model >> Gemma again dropped recently incredible model open source model so you can run AI on your computer without internet access for free for life crazy you can do probably 70% of the things that you want to do >> at a very very high quality without spending a single rupee and for laptop you can use a tool like

30:48 chat there's a tool called as LM studio >> okay and the other advantage of doing this is that when you use an open source model then advantage your information is on your computer important thing what I'm realizing about myself and everybody has to realize this about themselves as well. We are opening up about things with AI that we will not with any human in the world deepest insecurities because it used to give articles. Now AI gives you opinion and on top of that AI

31:33 psychopantic and AI is like yes bro you're right >> I trust you you're awesome >> 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 think around 6 months back >> okay news around 25 million or 40 million chat GPT chats were leaked because of a breach. >> Oh, when I read that I'm like I don't want my messages to get leaked, man. >> That's what >> so anytime I've made a habit after that by the way every time when there is

32:20 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. >> So it could be a lot of things, lot of insecurities. This is how I'm feeling. Is it okay? >> Every time I say something like this, I also say that 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. Right? That's it. >> For example, OpenAI recently said that especially for kids, if they feel that the kids are getting towards direction

33:06 of suicidal >> they will notify their parents. >> Crazy >> Claude. How do they identify? >> By looking at messages, conversations all parents identify. >> So these are kid accounts. >> Okay. So naturally kids al also said if you do anything illegal there there's a possibility that this information will get to law enforcement. How crazy these things are there and it should be there. Safety guards safety guardrails have come in slowly and has to come. There's no two ways around. Of course,

33:53 two quick questions. Okay, see let's talk about your case only versus this the amount spent on AI tools or AI models or subscriptions versus one more >> just the opportunity cost >> I think it'll be like >> let people understand 80x I guess it'll be 20 I can't even so >> and that to 24 hours with you >> anytime 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 >> and the way I want it to work like you

34:41 know that uh I create content now >> just like you get brand deals I >> you are not creating content >> so my face creates content my AI creates content okay >> he's never there in front of the camera >> only in this podcast Actually sorry to >> uh so that that line of business also works based on brand deals. Yeah. Yeah. >> Right. And so the more brand deals the better higher quality brand is the better it is for the whole business. Let's say we are doing 100 rupees we want to do 1,000 rupees next year. >> How do you scale? The only way to scale is to get more sponsors. So what we created, I'll give you this

35:26 example. We created an AI agent. It's called the sponsorship bot. >> Okay. will be 10 >> m >> 10 15 people like me. >> Yeah. >> Ed center will be all together 100 people YouTube, Instagram, LinkedIn, Twitter, >> newsletters. >> Okay. 250 300 properties. >> My AI every single day. I think around 10 or 12 time they do pulse check. >> Oh, >> 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 profile I think 80 90% my team keeps adding so

36:14 gets added to that list right of the language >> I don't care irrespect of the language right but we mostly have been looking for English >> okay >> we have not built the engine where we telling >> it watches every single video of every single person that gets posted in a few hours of it getting posted Okay. >> Reads the description. >> If sometimes watches the video, >> looks for sponsor tags. Sporm thankfully. Have we worked with this brand? Yes or no?

37:02 If let's say let's say I don't know uh Google >> Google has sponsored a video but they have not sponsored >> sponsor you are part of the watch list we call it watchtower okayati has a new sponsor which is Google >> and it checked our CRM and say that okay we don't have Google as a sponsor then automatically >> we have an integration with something called as Apollo >> Apollo >> Apollo Apollo.io Yeah, >> which looks for email id and phone numbers. >> API. It'll 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 if it can have any kind of intel intel

37:49 >> 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 ids, which it does 99% of the cases, >> it will also send a cold email from our end from my team's name saying, "Hi, I'm representing web hubs team. We just saw your integration in Wamshi's video. Really enjoyed it. >> We also create a lot of AI content. Would love to see if we can do something together." >> Now, this is all and then it'll send us an acknowledgement on my Slack, right? Uh now when I do this when you say social media channel I think it is just not possible because

38:37 I don't know how I will hire a human to sit and watch every single day to and real time because if a brand is sponsoring now I need to talk to them now it's It's not going to work. You know that. >> Yeah. Yeah. >> Right. This is just one example. >> Yeah. Insane. Value of utilizing or maybe adaption of AI. I thought I wanted to ask you this question before asking that question. Every 100 words email it uses around energy of 14 LEDs. No. Okay. Got it. Now what is the cost of it? Free us

39:25 every word please. Thank you. Is it right or not? Yes. How much is the cost that there should be some some kind of a thing what's happening there? So it's it's it's the model that the world works in right which is called fremium. Right. What is that? What is happening here

40:11 will pay enough to take care of everybody's cost. That is premium right 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 worry there will be a point of time where like >> true so those three people four people five people who are paying for paid charg account are the ones who are paying for everyone else's free consumption that's the premium Samman he mentions that you'll be charged like consumption of your uh so I don't know how how far is it >> no so uh it's blown out of context so basically when you think about costs

40:59 what are costs what is costing you more >> what is that so at economics of scale >> everything gets cheaper Correct. Correct. Cost per. Yeah. >> The same happens. This is scaling law, >> right? Simple scaling law. The same happens across the board. As long as you have a factory >> which can manufacture as many mugs as you want. But let's say after the course of time >> now imagine a world where you want to

41:44 make 100 billion mugs. Okay. Let's say you are able to build the factories also electricity 100 billion >> right? So for the amount of compute or the for the amount of usage >> the world potentially needs >> or the weights going up the biggest bottleneck the biggest reason why they're not able to scale as fast as possible is purely because they don't there is a energy deficiency like electricity >> electricity

42:31 data because they are power hungry data centers righty deficiency only. So what he's saying is the ultimate thing that you have to solve for the ultimate thing that everybody's going to pay for is >> power is energy >> right that is what eventually when all this will become cheap >> all this robots are running everything is happening then eventually you'll basically be paying a electricity bill or generalized statement got it got >> right >> on last podcast I was asking how can a normal guy who never used AI become a generalist in the AI level 0 to 5 and people loved it and even I I liked it a lot in case

43:16 because considering all these advancements what do you advise and do you still call that scope as generalist what do you call it and what should people do right now >> so here's a fun story okay >> like you said do you still call it AI generalist yes we still call it AI generalist because we coined the word I Oh yeah. So in a way we put AI and journalist together. General is a common word put together. I commonly started to use that in the office and then I know I started to speak about them in podcast and all. At that point of time nobody was talking in that lines. Now there is this word that has become extremely popular. >> Okay. >> And that is called as AI orchestrator. >> Mhm.

44:02 >> AI orchestrator is essentially AI generalist. It's the same. It's a different word. >> The world has >> until the world has caught up to AI orchestrator like a Sam Alman Guratindo like Elon Musk has spoken about it. Vinod Kosa has spoken about it. Everybody has spoken about AI orchestrator or AI journalist in some form of fashion. So one year the acknowledgement was not loud yet >> but two it's actually one and a half years old. The word is one and a half years old. when we designed it, >> it was designed or the way because it was made for my team. My team goes on JSON that became the road map that I shared here. >> Right. >> Right. But still, yes, AI generalist is still very very relevant. That is it. That is actually

44:49 stronger notion that everybody has to become a generalist. What is a generalist? >> You are a AI first problem solver. >> Right? >> 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 called >> but I think 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. >> AI our lives will revolve around AI. So we don't have an option. Acknowledge it. >> A lot of jobs are already going because

45:35 of AI. Acknowledge it. >> The job that you are doing >> acknowledge it. But also acknowledge the fact that the job that you're doing today as 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 powerful thing we all have to use it whether we like it don't like it respect it don't respect it I say it you say it you have to use it that is the first layer that you have to do >> agreed >> don't fight with it there is no way you can win >> right once you acknowledge right you come to D

46:20 >> which is dabble >> okay >> what does dabbling mean Mhm. >> You're dabbling with multiple things, right? You're dabbling. >> It's not a bad thing. It's also not a good thing. The bad thing is that you will never realize the true potential of AI, >> right? >> But you will never go to the next phase. So

47:06 what is the job of a dabbler? When you come to dabbling, your job is to play with as many tools as possible. >> Try try try play with a lot of tools. When you play with a lot of AI tools that are out there, they're free out there. So as you dabble with a lot of AI tools, you'll get the full spectrum of advice problem.

47:54 What do you adise? So when you when you don't know anything and you go directly with a big problem statement >> maybe obviously dabbling you'll try 10 15 20 different tools you'll figure out what is good with what for example I don't know let's say chronicle >> you try you like something about it and then you try one more software let's say gamma >> you realize I really like gamma for presentation But you know what chronicle presentation >> so you explore your new ways. >> Yeah. Correct. Yeah. So that is very important. >> You have acknowledge then you have dabbled. You played with like 10 20 50

48:41 100 tools. Then your job is to amplify. What is amplify? >> 50 60 70 tools. Free trials. You understood the power of AI. You understood what works well for what. Eventually, you'll realize this is what I'm getting to eventually. So, you amplify them. Now, it's about depth, not width anymore. So, okay. Anthrop Claude became your answer >> for let's say assistance and everything. >> So, Claude possibility. What is memory? Memory. How can I go deeper into automating things?

49:30 How do you play with I will go deep into it. Then you go to console.anropic.com. Then you learned a you go deep. Yeah. You become an expert in a bunch of tools. That is the phase here. When you do this, you'll actually end up building your toolkit >> problem because you've gone through the whole dabbling at this point of time. You have amplified let's say tool A >> to be a presentation tool. So you go

50:15 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 amplification when you're playing with it automatically there's a characteristic that kicks in a characteristic you expect more from AI. >> Correct. >> Then that will open up a new evolution for you. No agent figure a new thing starts running in your head >> right? What is that new thing? How can I do more? How can I do more? How can I solve more problems? More problems, more problems. That is when you navigate to a problem solver in adap is problem solving. >> You come to problem solving. And this is

51:00 a very important phase in everyone right the thing that you said you'll think about how do I solve a problem the beautiful thing is already amplify >> bunch of tools on top of that you have dabbled with so many more tools which is in your loose memory problem by the time you get to this point you will be able to connect dots tool e tool e tool workflow presentation because we're talking about that example right presentation there's so many things then design presentation eventually presentation >> so now you have four things oh deep research on this topic I want to start

51:47 this melling business deep research >> in amplification in dabbling I tried 50 tools in that I amplified let's say perplexity deep research >> perplexity deep research so model council option because you have amplified them. You've gone deep into them. So you don't only know deep research, you also know something called as model council exists inside of perplexity. council. So it debates with each other models to give you an answer like five people are working from five different companies to give you an answer. >> I will tell you that I have a project inside of my chat GPT only which is

52:34 called chat GPT project where I also have all the transcript meetings of all my investors. >> Okay. >> Or every meeting that has happened with my investors which are online >> a meeting the transcript. So they know how what questions they ask when I say something a presentation let's say this is for my investor presentation this is a presentation I'm planning to present to my investors what do you think they will ask and I can vet that also this is problem solving problem 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 be they repeat repeat every AI agent,

53:21 right? >> That becomes an AI agent. When you become a very strong problem solver, you by default automatically move to the next phase of T of ADAPT, which is tying it all together. You put different workflows, an AI agent, a VIP coded product, a clawed uh project using cowwork and you stitch everything together for it to work in Symphony. You explore things like paperclip. What is paperclip? Paperclip is like okay agent, >> right? Think about 15 different AI agents working like your employees. >> Paperclip is the CEO, which is also an AI where you you just tell the CEO, you are the investor. You just tell the CEO this is what needs to be executed. Yeah, the CEO figures out

54:09 splits up all the work and gets it all executed eventually task >> which is all AI everything is an AI agent inside of it that's paper clip >> now the reason I'm giving an example these are not perfect yet and one year later when we are talking about it no I'll probably show you on my phone that this is happening right now because that's how tech is moving >> but when you all tie it together that's when magic happens >> people who figure out how to tie things together are basically orchestrators. And this is how you become a great AI orchestrator. This is how problem you're thinking AI first. >> Yeah, >> that's it. Connect

54:55 AI first problem solving. When you do AI first problem solving, the kind of solutions you can build, a human can't do, >> right? It's not practically not possible. We've spoken about example. >> Correct. 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 uh adapt framework. Simple but works really really well. M >> in fact me and my team and everybody worked on this presentation of adapt I think it's a 20 25 page presentation >> okay

55:40 >> of this adaptab framework super >> for people who are genuinely curious >> they can click and read it free let them use it and they think that is >> what AI I is cool. You get my point, right? Okay. So, I get I get what you mean. So, >> and they think that is AI problem because stages further out you'll not grow as fast. But here is the beautiful thing the

56:25 narrative right I think that is not untrue because a new job will evolve and you will be the best fit for like what kind of jobs can come. So see there are two what kind of jobs can come there are two line of thoughts here line of thought the jobs that will come that never existed before context engineering jobs uh AI agent builders only wipe coders so and on the

57:12 technical side uh you know all AI research kind of roles any research kind of roles are all there these are new roles that will pop up >> and these roles are no escape with this AI >> these are new roles created purely because of AI. >> Yeah, >> this role AI agent builder is a role that would not have existed. I think Lensart uh I saw what is this anthropic I saw immersion and replet all these companies in fact perplexity also are hiring people as orchestrators. >> Oh, >> Apple also >> and these are all new roles that are coming all revolving around only one thing. M >> AI generalist or AI orchestrator that is what the job is.

58:00 >> I lost it when I heard it. Okay. Fa >> Fiverr did 1 close to $1.2 billion in revenue. >> Okay. >> GMV. >> Last year. >> Last year. Yeah. >> $1.2 billion GMV. >> That 48% of revenue of Fiverr was for AI services >> of 48 F. >> Yes. $500 million worth of revenue has come for AI services and no freelancers >> half of the revenue 48% of the revenue is for AI services chatbot kawali rack kawali agent kawali v coding kawali half of the revenue of fiverr now the reason

58:46 why it's crazy is not because it's $500 million but because it's only one platform >> how big is freelancing economy on the internet it's close to $25 billion >> 25 with a B organized unorganized is 25 to30 billion freelancer.com Upwork this that everything together piece >> piece 50% 50% 30% 50% right now in this DNA age about 5 to10 billion or billion dollar and the 8,000 crores 80,000 40,000 to 80,000 crores worth of freelancing on AI is happening

59:34 dayc 2 3% penetration that's what I was about to say let's say 10% penetration only 10% or 5% penetration is on AI services while website designing website, freelancing and website digital marketing. Imagine website designing in 2000 >> or 1996 I don't know when when did come I don't know right

1:00:20 >> then that is the version of AI services then the 10x version is the opportunity right now >> right 10 billion or 7 billion or 5 billion whatever that number is land let's say 5,000 >> 50,000 cr 20,000 crores my bad sorry huh >> 40,000 crores than they have 20,000 crores 20,000 crores worth of revenue on freelancing or consulting is happening when most of the world is not AI aware either aware companies large companies are seeking for services >> right

1:01:08 Hyderabad When AI gets to this phase, imagine this will become like 10 times, 20 times, 30, 30 times larger and that is the opportunity. >> But there is no supply. >> Yeah. So AI consulting is just insanely big and AI consulting insane crazy AI content AI contenerated that in itself I know a person making AI videos for brands AI services AI services video category

1:01:57 category let's say content cloning you know all my content is AI generated let's say tomorrow there's a founder who doesn't have the time he wants to create content but he cannot shoot if he gets to know content AI content for founders personal brand is one category I know people who are making crores >> just doing this $5,000 $6,000 per month per client in India international in India I have friends of mine who are

1:02:44 founders who are paying five lakhs a month for this in India cont opportunity is 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, you can sell prompt engineering as a service. >> Lot of people want you to write prompts. Okay, you can code products. I know people who are making millions of dollars

1:03:30 MVP$10,000$20,000 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 VIP coding as a service. I know I have a friend Harshani. He again does a multi-million dollar agency. >> You know what he sells? AI agents. every month. This is AI agent as a service and I've

1:04:16 told this openly to my team people also every company has so many problems. Find a problem that you are passionate about. Figure out a product. You go through this adapt framework. You figure out you solve that problem and deploy that solution in your company. Solve that bloody problem in your company. The moment you solve that problem, let's say

1:05:02 using AI, can't you make that into a product and sell that? Boss people say and it is true that AI is going to be destructive. >> It also is the most constructive thing that we have seen in our generations. >> Yeah. I truly truly truly believe it's generational makes sense billions of dollars of industry

1:05:48 thousands of crores burn. So what's the future inction? Everything goes to an inflection point. Like how old are jobs? Humans started to work in jobs for how many years go? >> Couple of hundred years. >> Probably 100 years. >> How long have we been around? Ask your grandfather what his grandfather used to do. >> He was not doing a job for sure because it didn't exist. >> Right? We humans are incredible with adapting >> and we also have a short-term memory. Computer jobs. Computer jobs. How old are jobs which are around computers?

1:06:36 20 years. 25 years. That's it. 20 25. >> I know that's a story pain. >> Huh? Like 25 years is old. People survived, right? There are we will adapt. But the problem with all of this equation is see there's going to be long-term there's going to be utopia. Utopia is it's going to be all happy. >> Yeah. Yeah. >> Because there 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. >> Oh no. >> Not basic high >> which is which I I I really really hope Elon Musk is right. Okay. >> Uh but uh universal high income what is

1:07:24 that? It's very simple actually. Quite interesting. I don't know how many of the people are geeking out. Sam Alman dropped a report. Okay. Okay. >> Uh it's a report for the intelligence world that we're going to live in. >> Okay. It's a fancy word. >> Intelligence report. What Sam Alman is predicting is AI is becoming so smart. And in the next few months if not years in a few months >> okay >> advanced models of AIS are going to drop which are going to be so powerful that it's going to devastate jobs >> okay it's the AIS are so powerful that it will hack into every system out there >> it will be absolute mess there will be a shakeup of economy that will happen because of the new AI models that are

1:08:10 going to come in >> okay >> is what Sam Alman is saying a Claude dropped Claude mythos which we'll talk about a little later claude mythos dropped which is an AI model which is so bloody powerful that it is able to hack into almost anything. So only 40 companies access saying main using this model anybody can hack anything. So that is what that is what Samman said one day before and the next day anthropic dropped this model.

1:08:55 >> But anyways but the solution is that 13page document. Okay. Which is quite interesting. >> The first thing that he basically said is dividends. >> Mhm. >> What does that mean? Every human gets dividend. He talks about US citizens. But this if I I'll put it on global perspective. So if most of the work is done by AI then you don't need right

1:09:43 you tax those companies. >> Okay. The government taxes those companies, >> uses that money and gives everybody, every citizen money to survive. >> Okay, >> which is basically universal basic income. >> Universal basic income taxing the AI companies. Correct. So Adi so that is how everybody will get everybody gets money and actually Alaska looney I think they have the prochemical or some some setup there where they already do this. >> Oh >> every Alaskan gets money from the

1:10:30 government. So imagine that at 100x that is handouts that is universal basic income. Elon Musk there'll be nothing left for a human to do >> and there'll be so much abundance that there'll be insane uh devaluation of everything. >> As a result, money will hold no value. That is universal high income. >> So everybody has access to everything. >> You become animals of >> We don't know. We don't know. Have you watched Good Life? >> No. It's a show on Netflix. It's quite you should watch it. No. What will happen to the human race?

1:11:18 Next, next five, 10 years, >> the humans will evolve because I think that report. So human in the same way that is called robot tax. >> If you get AI or robot to work, you pay tax for that. That is robot tax. >> Okay. >> So it's quite interesting. That's a 13page report. Then they say on a short term

1:12:04 >> people will not have enough work. So we should make it a 30-hour work week and pay the same salary >> not reduce it. I think this is more on a shorter term. M >> so this report is a reflection universal basic income harmony guard rails if AI loses the trail if it ends up doing something we need a backup plan back I I feel like it is like Uh what's interesting like the thing that is crazy is everything that we were speaking 6

1:12:50 months, 9 months, 12 months back is happening bro is happening before even I predicted it or anyone else did. >> Now this is more scary. >> So this is about a this is about Sam Alman. >> I know we are getting >> the New Yorker report or the article says that Sam Alman may control our future. Can he be trusted? They met around 100 people. I mean, I'm sure you must be knowing it who know it says that it's almost like evidence investigative is a liar, sociopath

1:13:36 basically. Over to you. What do you feel? I think disruptive things are built by disruptive people. Okay. You can't be a straight-headed person to build something or to run something that will change the world or else it would not have. So while I'm not comparing Sam Sam Alman to I'm a huge Steve Jobs fan. >> Okay. >> So I'm not comparing Sam Alman to Steve

1:14:22 Jobs. >> I don't think it's fair on both the fronts. M >> not just on Sam Alman's not just on Steve Jobs also on Sam Alman Sam Alman's front >> how was Steve Jobs have you ever read his biography >> how was Bill Gates have you seen uh movie and was he straight was he a straightheaded person nobody is see everybody's weird in their own ways. So I don't know see so I'm nobody to say this but I think

1:15:08 >> for you to build something so destructive you have to be twisted man. >> Yeah >> you have to be mad. >> So I think it is blowing things out of proportions. Like think about Elon Musk, the same guy. But he is the same guy who who's the reason why even India Indian roads have electric cars. >> Tesla electric cars that is the one that built the business. He's the guy for the first time when he said I will land the robot I will I will land the you know rocket back NASA laughed the world laughed he made it happen now everybody

1:15:55 wants it a straightheaded person doesn't think like this now Samman says boss we don't have enough energy when Sam Alman said I need to raise a trillion dollars >> everybody laughed trillion dollars it's bigger than a country how will you raise a trillion ion. Guess how much he has raised after he said that? More than $150 billion already. $150 billion 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 of funding. He said $1 trillion. People laughed. I'm telling you, he'll get to a trillion dollars. >> So now there's a bigger question. The

1:16:41 bigger question is See, I believe in larger goodness, >> right? He could be twisted in ways >> so that he wins but eventually he also would have to live with us. >> Yeah. >> For that humanity has to win. He's not in a job. See what fight Sam Altman or Dario Mday or Elon Musk or Sundar Picha at this point of time they're 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

1:17:29 world. >> Yeah. And this the world that they're going to live in is not a world where they're going to whip people. It's going to be still a democratic world. >> Correct. >> So you have to be a democratic leader. So you 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're rude, if they're arrogant, okay, big deal. It all happens. It's fine. Like like this is all clickbait according to me. >> Clickbait. Okay. >> See, he's 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.

1:18:16 >> Did you meet him? >> No, not yet. Hopefully one day. I really respect >> what he has uh you know what he has pulled off. >> You have to give that respect to that person. >> Absolutely. Absolutely. >> Right. You cannot say he's a human in the end, boss. >> Yeah. >> And on top of that, he has 0% equity. >> You understand what that is? >> There is no upside for him. He has he owns 0% of the company. He's just the CEO. He gets some salary whatever that's some dollar or $2 or $100 or whatever the salary like that. His upside is all on power. >> He's not doing it for money. He's already a billionaire.

1:19:03 >> 1 billion, 100 billion, 1 trillion difference. It's not exciting. Power is the most exciting. And the most powerful person is going to be the person who's running 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. It's about you. Me? Yeah. Okay. So, what perception I get when I talk to you? I saw very few people who can convince anyone on a on a transaction on on a sale or on a financial figure. Single.

1:19:51 You are the best and I spoke to many many people who worked with you who are our friends also. I don't know name but uh feeling he's great sales guy of course you're a great founder great great great great builder how did that sales formula how did you crack that is it is it innate how much of it is innate and what what taught you I think every founder has to be a great saleserson period >> like look at not that I'm I'm not saying I'm a great sales guy >> uh but thank you uh but look I mean again not driving pals that would be a blunder. So I want to clarify >> uh Steve Jobs is probably the best

1:20:38 salesman in the world right but Apple so I think it's every good founder is a by default a good salesperson >> so I've been doing stuff uh back in the day robotics uh you know I've 200% huh Uh so I I've sold stuff in the past. uh so I think Uber used to be very much I enjoyed sales because >> I think a lot of people but sales that deval it say I feel that's the best thing that you can do that's the part that I enjoy doing the most

1:21:24 >> right because it's sense of accomplishment so many times >> every time you close a sale it's a accomplishment >> every time you close >> all right so I think like 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 is also sales% teacher convince homework because of this reason is also sales >> college professor is also sales everything is sales is what like you being able to put the value out >> as long as you're able to put the value out it's simple And when I talk to people, I I usually use the simplest

1:22:10 tactic. >> And the tactic is I tell them what I can do. >> Not more, not less. I don't like negotiation. >> Then negotiation work. Hell no. >> It's no point. >> Yeah. >> 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 is the line we're doing. >> No, I'm sure you'll never break that. Discomfort. >> There is rational. No, every time I say this is what we can do, I tell why. >> Yeah. >> I don't say that I just want it. >> Right. >> I say why important like one of the highest paid people in the world are actually people who in sales >> because sales people they don't get paid

1:22:56 just for >> the job. They also get paid for the value that they bring in in form of incentives. >> Right. Right. Right. >> Right. And because >> I'll tell you what, the highest paid people in the AI world are going to be sales people. Product build >> the thing is putting the product in hands of people. Who puts the product in hands of people? It's a saleserson. >> So how much of this can be automated? Sales and a process flow according to you. lot of these things but in the end human right that is why it is high so it's most of the things are automated for example let's say

1:23:41 let's okay you have to get on a call to even understand legit 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 agent you use something like or retail or an app like that agent rightes using again an MCP or an API integration

1:24:28 or you built an agent >> right where automatically using got it you need an army >> scale yeah >> right so that is where an AI does the call eventually AI assessment okay they have the budget it's worth our time out of that 50 people >> 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 of business you will need 25 people or 50 people or 100 people >> correct >> instead of that you have only one person who's talking to five people who's bringing in the same value now there are

1:25:16 two things that people will look at it that one month they were driving. So are so important right now. So if you're a person in sales, you have to think about how do I become that two people >> because those two people are going to get 10 times or five times the value than all these people could get. This is low value work. >> Correct? >> This is high value work. AI is replacing jobs. Low value work is the one that is

1:26:02 getting managed. High value work or new approach of work, the person who built the system, the person who's managing this automated 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 in design, same example in software engineering. Low value work will go away. High value work will continue to stay that high value will the high value work expectation that is why the orchestration layer has to come in as well. So yeah also all these AI companies are coming to India know anthropic

1:26:48 >> open AI Google one of the largest uh uh org size in India is all account executive account executive their job is to sell >> they're all sales roles nice and what do you feel about this uh AI integrated uh with hardware meta glasses glasses things like this The neural link neural link. Yeah, that one. Temple Ga they say uh >> is brain wave detection for involved in the new. >> So what do you what do you think about it and what's happening in Bangalore? Is there anything uh you're looking at? >> Not anything that is happening in Bangalore that is very exciting except for neural except for temple that's

1:27:33 happening in probably Delhi not in India. It's very interesting I think AI hardware market because >> yeah yeah because I mean 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 >> okay >> so you don't need to talk >> if you see dime which is a project name like a internal code name for open AI's earpiece. It's a external piece >> which will be here like it's not blocking your ear completely. just here

1:28:19 next to where the idea of dime basis of patent file. >> Okay, only patent >> that basically is a device where you just think >> okay >> what you want and the earpiece is the AI which talks back to you with the answer. >> I got it. Okay. So uh Miaki Lab Sanunta or Mai Lab Sanunta something like that uh Sam Alman invested. >> Mhm. >> It's 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

1:29:06 like neural this is without it's noninvasive. There's a headset again by a US company where it can understand what you're thinking >> without actually having intrusion. So let's say you're in a meeting >> okay question I give answer for example I could not think of the device name >> right >> that you know what it >> I know what I visually can imagine what the device is I I can remember the screen I know what it can do right those pictures and that imagination is going into chat GPT >> model and it's able to say what that device is >> crazy hi and it'll talk to me through my

1:29:53 ear that you can't hear >> right or uh let's say I don't know there's a business problem that I have and you have asked me a very tough question as an investor and I can't remember of it >> I just think what was the data back of my head and it just speaks out because >> this AI will have integration with my let's say super join >> to look into my MIS dashboard to pull the data and talk back to me through this right so this is this is the future and neural link is already doing that no So the moment brain wave stuff comes in alter ego is the name of the device. I don't have the time but the name of the device is alter ego right it's all here it's all here right and they also say I think Elon Musk I guess has said that u

1:30:41 >> 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 brain sidekick >> future techn could be the possibility. Now this dime is a always on assistant >> right neural link is able to understand what you want like diet coke okay so I don't have diet coke but we are talking so I don't want to pull my phone go to zeppto and do this so I'm just thinking order me a diet coke I'm just thinking and I'm at wamshi's office the address of office in the WhatsApp group >> I'm just thinking and let's say if I

1:31:30 send this message to my open claw it can already do this M >> it can already order Diet Coke via Zeppto 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. I'm thinking >> that's what that's >> I'm thinking because of that implant. >> Are there glasses? I think there's some crazy stuff that's happening. >> So Alex Wong when he came to India, I had a chance to meet with the Meta team. We had breakfast and all we spoke a lot about it. After that, the meta team showed me the future uh glass. >> I had a chance to play with it. It's insane bro. >> Future. >> So, I think it's meta display 2. They have a Orion 2,

1:32:16 >> right? Or one is out. Or 2. This is You can't display. You can't see it. And it looks like a regular Ray-B band. >> Oh my god. >> And right now it's all bulked up, but you get the point. The form factor is going to be like a regular Ray-B band. And there is a band here. >> Which reads your blood flow impulse something. >> Okay. But I'm navigating the screen. Huh? It's like this. It's literally this. and then one finger to I forgot the navigation right now was crazy

1:33:03 >> right on top of that neural link connection let's say you have something talking directly to your head >> your brain is communicating to that screen directly and you have a screen in

…