🎉😎 125th SPECIALL | ROUND TABLE #1 | RAW TALKS WITH VK

00:00 Who are you? Respect. You attempt... First ever round table on Raw Talks This was very interesting Can one see, one second Hello! Hi! So... Wow this was impressive This is my problem statement I didn't expect that Can we do all of that? Yeah Let's try Shall we then We'll start Is that possible And then I remembered something But it's too f**king risky Okay let's get to the answer That is insane man Prompting means clarifying what you want. Balaiah sir will say, I came to your house. How many houses do we have? Which one did you come from? He'll say, I came from your mud house. He gave clarity. Promting is similar.

00:47 What are those crazy things that you do with AI in your system, that you won't advise anyone to do? What are those 5 top channels that you follow? Video is AI, my audio is AI, my thumbnail is AI. How much are you making from your YouTube? Less than 1. Why are all of you sitting in the dark like a black flag? You look handsome everyday. Using AI for script writing. He's a fool, he didn't even tell us! And it is part of evolution. If you want the wheel to go then the square has to cut off its edges Is this the first time in your podcast that we are eating? Actually yeah, I've been doing this since 6 hours ago You know what, i had lot of fun Welcome to the show Thankyou One eternity later Rain has started pouring inside my shirt Haka! Oh no! I have burnt everything

02:05 Okay, so it's a new attempt and I need to wish you good luck. And the new thing is that our co-founder Pranavi has worked really hard for this. In fact, she was here earlier. She does well in events. We have done open mics with 1000s of people, right? She is all active. I got scared when she came on stage The father was super active in front of the camera. I don't know what happened but thanks for accepting finally. Welcome to the show! So after yesterday's conversation, I was thinking why am i just daring to stay away from certain things. It is time for me to explore properly and adapt.

03:03 So today, completely... The topics we discussed yesterday can be repeated but in a practical way. Let's brainstorm and as we thought, we'll rope in few guests also. It will be like a surprise element. We're going to take their problem statements and let's explore. If possible, we'll solve them here or think about how to do it. Before getting into the proper crux of the conversation I used to feel why this word was over spoken and celebrated. But with time, every month, every activity that i do on any platform or chatbots, I understand and keep getting the importance of it. Why is it so important? For example, if someone says about Pranavi, she wasn't fast. So even though she had a headache one day before,

03:57 She was complaining about headache. Okay, fine! So she opened ChatGPT. It's common and many people do it. She told me that this is what has been happening since yesterday. I'm getting a headache. And it comes in one part only. And I am on fast today. This is what I know. Now ask me more questions to understand my situation. This was very new for me. To stop you from asking questions and to know which question should be asked by me... Oh, this is the new way of asking a question to a chatbot That means it will give a list of options. It gave the list of options and with patience, that was single audio. 30 long questions. She gave it and almost I think she came to accuracy. Yeah yeah. She said some remedies. Okay you are fasting but at least drink juice or something like that. We took it. She was all good in half hour or 45 minutes time. This is great! So utilizing and leveraging AI to actually understand how to use an AI

04:55 Now I understand why is it so over spoken and why, why, why. But yeah what do you think? What's happening around us? Prompting and engineering has penetrated to this point. Yeah Dilip Garu please welcome to the show. Thank You, thank you Coming to prompting So for me, prompting means it is always Clarifying what you want. That's prompting. I've said this on previous occasions, but I'll use it again

05:36 Balaiah says, I came to your house. It's an ambiguity in the story. How many houses do you have? Which one did you come from? Okay. So he comes to your actress' house. He shortened the ambiguity. He gave clarity. Similarly prompting is similar. What does AI say about this? Even Balaiah would not have thought like that! Welcome to Delete To The Podcast. Exactly! I didn't know it was a movie buffoonery. I heard him mentioning it. If you had sent me a invitation, would you have used this as well? Okay sorry. So we need to clear the ambiguity of our AI. So how much ambiguity can be cleared and what do you want to say

06:25 That's AI work. That is your job. You know what is happening? I don't know if you've heard of context engineering. Context engineering has become very popular. What is context engineering, is exactly what we are in a way telling and what you guys are doing also on the matter. Prompting means... prompt engineering is question You can ask a question like this. That is prompt engineering. Context engineering means, how do you give the AI model enough context for it to be able to solve the problem? What happens in context engineering is that mostly without knowingly... The moment we go through the ADAPT framework and go deeper by default we are becoming context engineers

07:06 If you see my prompts and all, right? I want you to do this task. For that, you need to go to my email. For using my email use the Gmail MCP. The other thing is I want you to fetch the data of my YouTube Analytics. For YouTube Analytics use Chrome browser using Browser Use browser user is a tool pair which AI has access to. But you are not letting AI figure it out itself, you are giving context now so its context is all about making it easier for AI to understand what it needs to do. So in this Ram from Mir and both of you

07:53 After using agents and setting such high frameworks, you will be using these AI chatbots right? Can we know which chatbot is being used for what kind of a device? Sure, I think again right dabbling phase and we keep going back. So I use everything and I keep using everything the reason is it's our job to teach everyone also what is working 100% right? Genuinely go to Back of my head is based on use case. Right? For example, deep research. I haven't done any research about any of the topic right now My answer and every tool has deep research. ChatGPT, Gemini, Perplexity But i use perplexity majorly. There's one feature in perplexity It's called model council What is the power of chat gpt its model

08:49 Chat GPT model is called GPT. Gemini model is Gemini. Clod's models or Anthropic models are Opus and so on. When I used to do deep research, I would first research all the tools and then take them to a platform and condense it That is what Perplex does out of the box using something called as model console. So, Dilip is going to show you very quickly what model console is? Absolutely! Now with model console we will validate a business idea. Right Okay so let us do that Any business idea that you have you want it validated

09:25 Okay, so basically it's majorly for optimizing retail outlets. So we have a lot of you know especially when we talk about restaurants there will be a lot of people coming in and sometimes we don't address those people. So there was one business idea they were completely into computer vision like CC cameras that they've connected something I'm not sure what they're connecting So, with computer vision technology it captures everyone's face and how long they have been sitting on which table. And let us say if someone is not addressing them it will ring their bell inside and that person will go take care of them. Similarly, the table count, how many people are coming in each bed, how much time are we wasting because sometimes when we keep talking especially when we go to cafes so how much time do you waste? To optimally utilize all these things

10:14 in Chennai has started doing that. So, that's the business idea. Okay. This is Wisper Flow? Yes. I think this was mentioned by MyWatch last time as well. Yeah so this is a tool we use to convert voice to text. You have left out one miconchious which you are talking about now. There was also this video where they said a Ferrari. If we are, if you show us one mistake, no matter how fast or slow you speak, whatever it is, we will give you a right. I have never tried this. Kudos to Tanay who was a good friend so very good friend.

10:50 I didn't realize that you were speaking in Telugu, but it was all translated into English. There is a new feature called Auto Model. So here there's a plus sign and below there is a model console. This is the pro version? Yes, this one is a bit... we have to pay double for now. But what this means is that they will make it to other versions also. For sure! Currently, they do this. They make the first features in a big payment and eventually release it to others. Here you can see models. GPT 5.4 Thinking Model. This is an expert. Gemini 3.1 Pro, this is a big shot. In our panchayats... Oh these three are doing it together? One person sees if the business idea is right or not. Another says that this point isn't right and so on. They all discuss whether what they say is correct or not.

10:14 in Chennai has started doing that. So, that's the business idea. Okay. This is Wisper Flow? Yes. I think this was mentioned by MyWatch last time as well. Yeah so this is a tool we use to convert voice to text. You have left out one miconchious which you are talking about now. There was also this video where they said a Ferrari. If we are, if you show us one mistake, no matter how fast or slow you speak, whatever it is, we will give you a right. I have never tried this. Kudos to Tanay who was a good friend so very good friend.

10:50 I didn't realize that you were speaking in Telugu, but it was all translated into English. There is a new feature called Auto Model. So here there's a plus sign and below there is a model console. This is the pro version? Yes, this one is a bit... we have to pay double for now. But what this means is that they will make it to other versions also. For sure! Currently, they do this. They make the first features in a big payment and eventually release it to others. Here you can see models. GPT 5.4 Thinking Model. This is an expert. Gemini 3.1 Pro, this is a big shot. In our panchayats... Oh these three are doing it together? One person sees if the business idea is right or not. Another says that this point isn't right and so on. They all discuss whether what they say is correct or not.

11:47 Nice. Okay? So basically it's not like a drastic... So, there is a discussion. It's not like... It's a debate between three different... Last time we opened Lovable and Replit for the website. It's not that. It's like different versions. This is like different AI models discussing together and then coming up with an output

12:27 So, if you see. Go for it. So, here council GPT 5 point 4 is thinking and this problem statement is solving that first. Zoom in a little. Claude Opus, another big shot is looking at another model. So basically what is happening is... You said it's paid right? This can be done for free too but you have to struggle a bit. Open Gemini Free Model, Google Free Version, ChatGPT Free version and Claude Free version, give the same question to everyone and when they get the answer, you can take it to all three models I'll tell you what its doing

13:04 three different AI models, one is working. After it works, each of the AI models, let's say chat GPT, Gemini still gives a clod answer. For Gemini, we are giving answers to other models in chat GPT. So, we're giving answers for other model also to each one of the models and saying, Babu, this is your job. Two more have given. Now you understand what you did. Then basically you are simulating that and eventually I'm telling them give me a report So, Mala three reports in condensed AC. In Korea model need to study right at the model console destiny on a high level. So, it on touch yesterday or tomorrow.

13:45 It shows all the sources, how it is thinking. For example this model is looking at business idea validation or computer vision startup or retail optimization. Let's look at the above model and see what it sees. Startup idea validation preferences. It comes by cutting something else. AI product business validation. First comes AI products. What are these? Fundamentals sir! That means I will come up with a foundation These foundations are not coming up. I am directly jumping into it." Then, they compare how you wrote the answer and vice versa. Or else if they say this is correct or that is correct then this happens. So as its happening in grok there's something called as Grokh Heavy. What does Grokh Heavy do? Here what he said... A debate work doesn't come. Debate usually happens on one layer

14:39 In grok heavy, what happens is... In fact I think Kimmy also released a similar version as well. What happens in Grok Heavy is that Grok opens multiple versions of its own. So, if a problem statement comes up, instead of what you see here, each AI thinks the same problem differently. That is how every AI works differently. But in grok heavy it will spin-up multiple instances of same grok. If you open 10 windows and ask the same question for 10 times, it will take a different direction every time. So Grok Heavy does that for you. By default because handoff is done by Grok, it will make it forcefully think in different ways and eventually debate with each other and coincide with the solution

15:35 But the advantage of model counseling is that problem statement validation with different AI models. There it's the same AI model competing with each other. So, we didn't ask them to do market research or anything like that? We just told them to validate their problems. I said just validate. Three people agreed that this problem was real and commercially validated. Another three people agreed there was a business behind it Do not use facial recognition, use anonymized tracking. Three people agreed to this because there is an India's DPDP Act. We don't know about it. This act exists so you implement it but remember the act. If you have technology then you can say I got a good idea let us put it in place and without putting it in place... Data protection... You should take care of data protection.

16:25 Then, use existing CCTV. Don't sell hardware. Don't install the hardware. Already solved problem. Come to software. Yeah! Then target organized restaurant chains first not single outlet. Not a single one but hit big companies and then go down. Where models disagree? Go down. Here... Here they are fighting. Confidence level One is worth validating, cautious. Another one moderate but sharp strategy. The other one high viable and timely. Okay. So here there's a disagreement. For Go-To Market, some people have time to validate it while others are giving HR. Right right. Because this came in your discussion and was not documented anywhere so that didn't get the

17:20 Where should I start from? One is South Indian chain. Another one is Hyderabad or Chennai. The other one is Hyderabad. So, comprehensive analysis? Yes, comprehensive analysis. Oh! Nice. There's a final recommendation too. Idea worth pursuing... I'm saying this after doing all of it. Idea worth pursuing but only with the privacy-first anonymous approach targeted at Indian restaurant chains solving the specific problem of customer attention latency and table turnover Don't build generic restaurant AI. Build a system that ensures no customer goes unattended.

18:10 If you are saying that if we take this pitch for validation to three different platforms, then the approach of bringing these three together and discussing with each other is definitely worth paying. That's what I personally feel. I didn't know this unfortunately. I will tell you another thing. This perplexity could build it Claude didn't do it. Why would he? He's a model, why would he talk to another model? He won't do it. Neither Google nor anyone else can do this. Because perplexity has all the models together right? What I like about Perplexity Model Council is every model has its bias Every model works differently It's like we have 5 smart people

18:55 are sharper in one area. Yeah, unique. Unique in their own ways I'm able to get five smart people to work for me rather than one kind of smart person working for me that's the advantage So deep research key you're using publicity Correct What are those other key areas or keys? Let's make it very quick In fact I might have a different stack he might have a different stack And that is the beauty. I don't use it, but everyone else does. Maybe they have a different one. Let's see. Even I don't know about it. Majorly for content research and all, I use Grok. The reason why I use Grok is because it has access to Twitter data. So I use Grok until content research. If I'm writing content on some topic, I need perspectives in it. So Grok has access to all the tweet data which it can pull which is great

19:42 But content rewriting, anything creative. I use Gemini for that now. Okay. Creative writing because Gemini 3.1 is very good with creative writing better than everything else based on my taste. That's the point of anything on the go very quick email rewrite and all those types of things I use chatgpd For anything development coding or anything like that i usually use either clod Or I will go to something like Minimax 2.7, right? Minimax 2.7 Chinese model you have to use something via open router but Claude is basically very good and imagine 5.4 Codex is also very very good which is by OpenAI that does it coding now you'll be like, you're not a developer why do you use coding You will see over the course of today problem statements whatever comes up It doesn't matter if your are developer or not all of us are building

20:39 Right. So, selecting the right model is very important Anything with automation, you have to automate something or build an agent if there's anything like that it's majorly cloud models so that is my high level stack If I can imagine And because of context, you were mentioning about a couple of Chinese models and tools Say for suppose I'm a guy who's against these Chinese apps Because we think they are scary I don't want to put my email or data at risk. But, I should log into that tool, right? What is the solution there? Nothing much... If you use a tool like OpenRouter if I can share my screen You are thinking of using Chinese app but not able to because it's not safe and all. Let us assume that fear exists. First of all, all apps must have this fear. Not just one app. Even in any other app, what do we need to do when we think that our details are not needed? LM Studio is one

20:39 Right. So, selecting the right model is very important Anything with automation, you have to automate something or build an agent if there's anything like that it's majorly cloud models so that is my high level stack If I can imagine And because of context, you were mentioning about a couple of Chinese models and tools Say for suppose I'm a guy who's against these Chinese apps Because we think they are scary I don't want to put my email or data at risk. But, I should log into that tool, right? What is the solution there? Nothing much... If you use a tool like OpenRouter if I can share my screen You are thinking of using Chinese app but not able to because it's not safe and all. Let us assume that fear exists. First of all, all apps must have this fear. Not just one app. Even in any other app, what do we need to do when we think that our details are not needed? LM Studio is one

21:38 where you can install an app like ChatGPT or you also have something called as OLAMA. Okay. OLAMA is the same, high level. What it means is that there are models right? You see here a download button. Why is it a download button because if I click on the download and say hello first of all It will download that model look at this. Meaning what? That's AI model which is open source okay GPT and Antropic models are closed source. They don't allow you to download them There are a lot of AI models by OpenAV, GoogleV. They don't have Anthropic now but most Chinese models are open source. Quen 3.5 is open source and all. Now we think that open sources are for developers. Basically you can run the AI model in your own computer or server. You can tweak it. So OLAMA is software where I can install

22:35 an AI model on my computer, which is happening right now. Does it need internet for that? I will stop the internet after this. Then you won't be able to record. Even if you don't have internet even if you are basically traveling on a flight or anything You can use this AI model to run So what's its source? How does it run? Every computer has a GPU For example, your computer has Windows and Mac. There is RAM and then there is graphics. NVIDIA GPU sticker comes first. Every computer has a GPU. It uses that GPU only. So how does the content come from? How much knowledge do you get from it? It's already trained. If you think about it, when GPD was 3.5 or 4.5, then there wasn't internet access for it

23:26 Do you remember? We forget everything very fast. Because we don't know if a new thing has come or not. How were they answering? Because all these models are trained with a lot of information, unlike humans who can remember everything. So it's able to answer. But we should never use such models for facts. for brainstorming, for rewrites. You should use it like that. If you want to ask a question about what you need to do in this situation then it's okay but you should not use it for things like... How do I put it? Facts! Like, what happened recently in Hyderabad? Don't ask such questions because that needs recency based information which these models don't have

24:19 But this is the Olama. When there's internet, I can add that internet access. Then it will be able to... There you go! It has installed and said hello how are you? There is this creator named Vamshi and I want him to invite me on his podcast So write a super convincing email which can convince him I'm using Gemma 3, should have used 4. By mistake downloaded 3. I hit something while scrolling.

25:06 It's taking some assumptions right now. And I'll tell you the beauty of the answer that we're getting, it's coming from an AI model which is 12 billion parameters. 12 billion means... I think opus will be a trillion. There would be 1 trillion for that. They didn't say but easily understood. So this is 200 times less powerful than an opus But its still able to do most of the work that you want to do. It figured out that you are a Telugu person. and it came up with this email. It's fine, let say you have a thought, convert that into notes all that can do very well but don't give it fact based information. Especially, we can keep our private information safe in here. Some people use temporary email generators right? How does that work ?

25:59 How does that work? It will just create a temporary email server for you. Okay, it's called burner email. I'll use one of those too. By the way LM Studios also has one more similar app that you can use and for phone Google AI Edge Gallery again same open source open weights model, you want AI to run without internet anything for install it on the phone. Now, Olamalai nain jubincheno? It was a 12 billion parameter model right but on your phone 12 billion parameters nadadhu. It's too big so it'll be like 1 billion parameter but still is very good trust me when I say this okay So what was that burner email? Next

26:41 Basically, this is locally AI. You can install the same. Actually, it's UI looks like a chat GPT. He said that. Full UI is full chat GPT style. And the other thing that locally AI does if I'm not wrong... ...is it can run Apple's own model. Apple started to work on a model, the Apple Foundation Model. It is also being used which is optimized for better performance on Apple iPhones. Right? Anyways so this is locally AI Next if you want I don't know do you use a burner email? No I don't use it So I use burner emails very strictly There's something called... Oh it must be here. For some reason it fell down This is burner emails What are burner emails?

27:27 Let's say you are sending an email to a platform. You don't want to be bothered by spam that is coming in So instead of giving your personal email, burner phones use this right? Yes. Similarly, Burner emails. There are bunches like Temp Email and many more This is the one I use which is called as Burner Emails It's a Chrome extension. Okay. Burnermail.io This is the one that I use What happens is, once you have logged in, it will look like a Chrome extension. Now see this email box here automatically gets an icon here Every time you click that email associates with the other one And because we are using it inside of our browser next time when you go to the same website it will remember your burner email So you can login with that without actually sharing your email Just incase if your emails still private Still private?

28:31 After exploring so much, what are those crazy things that you do with AI in your laptop or in your system that you won't advise anyone to do? Actually Dilip sir will tell better than Vaibhav. I have seen everything he said and thought when did I ever do it? The team is all tense. They say, AI agent has gone mad. He's got a session. My system is shutting down. AI agent will do something. Everyone is tense. How much time does it take to get the data? What's there for him to come out of his shell? That's the problem! If he writes too many emails, people think that he'll write like this. Only one phone call. That is the problem. But yeah I think

29:25 AI has access to everything of mine. Literally, everything including my stock market data to my bank balances to social media accounts to everything so open claw for some time on open claw Open Claw can be run in two places. You can run it on your own computer or you can run it on cloud. Usually, what people do is they take a Mac and run Open Claw there. Because they are too scared that something will happen to the computer because Open Claw is like an AI employee. It does whatever you tell it to do. So for an AI employee to work very well, he needs a computer right?

30:11 I have to buy a different computer for that. But, I gave my computer to her. So, she uses the same computer as me. There it can do literally anything I want it to do and I can message it on my Slack. Do you remember Jerry? Yes! Now Jerry is built on OpenClaw. So, Jerry version 2.0 is running now. It's build on OpenClaw so it will use Twitter, Instagram, LinkedIn messages My twitter feeds me every 3 hours. I have access to all my emails. It creates files in my computer. It sends 7 day summaries of everything that happened during the meeting. You asked me a lot of questions today.

31:03 So I wish i could show you those but it actually has sensitive information. I can't show that. It will tell me how my company's revenue is going today. I'll ask, why are the ad spends increasing and the revenue decreasing? Why does this happen? My Ad Manager will go to my ad accounts and tell me where all these have gone down. If the revenue has decreased today, I'll ask why did it decrease? I feel like it. Sometimes, opening the database daily feels less of a revenue. But you forget that today is Tuesday and the first week of Tuesday is usually low. So my revenue seems to be very low today Can you look up if the revenue's accurate? Is actually low today or am I feeling it? Lookup all the last 3 months data, last 6 months data because it has access to our revenues. Go to Razorpay, go to Stripe dashboard, do some analysis

31:54 So, what we have done is open glove. We have given it access to all the sources. Data is connected to something called as Julius. Julius AI has access to the databases of my company. It can only read, it cannot write. My people won't give me any right access. I'll do whatever I want. Once they deleted the website. So my people... They did a search online again? They will give me only read access. If you ask for write access, they would be sitting next to me. They'd take that away from me later. Maybe this is also a growth hacking technique. Maybe he made them do a search online. No no.. I got to know after talking with him. In the morning everyone gets their phones at 6 in the morning

32:32 Some people have already woken up at 12. Some people are still awake, so I am trying to wake them all up. Me and Mitha have been getting calls from 6am. Some of them have already come at 12 and deleted the website or whatever it is. But somehow in 24 hours it has backfired. I asked for OpenClaw, and it understood that the name was Jerry. It understood that they were asking about a company information. I have to ask from Julius. So, I talk to Julius and the data is prompted by AI in Julius and it picks up that data and presents it to me. It's like an employee for me. Employee gives us computer right? Chief of staff or whatever. Entrepreneurial residency. Meta-age also uses MANAS. Like its what it is. Interesting

33:26 Can we just see that? There is nothing. It's a chatbot. It's on my Slack, so that is the problem with this. It's a chat bot. So, I can see the answer only if you have any questions on chat bot. You don't understand anything else. You don't understand magic in technical. You don't understand framework or structure and skill. And also sadly...I wish i could show this but there's a lot of sensitive information. That is okay. That's why she was adamant to understand these structures. I don't care whatever the techicalities are we should build something in this video Yeah! I'm going to ask something and maybe she has prepared something. Ruby over to you. Let's go Okay so Ask one of the tasks that you want to solve. We can build that. I think one generic task for Andar ki help hai tatto is we always want to keep ourselves updated with news kada so how can

34:15 I feel that. Yeah, so I can just go check. The important thing is to send me a message on WhatsApp every three hours with the hashtag AI news or business news. How can we do that? Thanks Pranav. I think we can do this. There are multiple approaches. How much time does it take? Depends on what you want to do. Nothing takes more than half an hour. If you want biriyani, I'll make it in five minutes and you can even make it in one hour!

35:00 Chicken biryani and mutton biryani are both available. Let's do one thing, because if I want this automation to come automatically in WhatsApp or Slack you guys said it is an agentic process that would take time something that a regular audience can use for everyday things. It is not very accessible. Actually, we don't need Claude Kovac either. It is very simple. What are you asking? Now, here is how I will break the problem. What is a problem statement? I want to know what happens in AI every three hours. If you don't know what's happening now for the last three hours, then what would you do? Generally, I go to Google. Otherwise... Generally if we don't have Google, we go to, you know, GBT, perplexity, what happened. Correct!

35:55 Let's say you go to chat GPT or perplexity, or any of the tools for that matter. In this example, we have done perplexity so far right? Let us go to Cloud now Every of these tools, chatGPT, PowerPlex3, Cloud has something called a scheduled task. Okay. You ask an employee to do one thing every day. You can tell the same to AI and give it as a task. So everyday when you say that, it executes the task for you. You don't have to remind yourself everyday. And in fact if you have your phone installed on your phone, like Cloud or ChatGPT, it will send push notifications too. It'll say, you told me something, I did my job. We can do that first on cloud. Why don't you show?

36:51 I will ensure that it is locally in a file, in your folder. In the system. Done. So let's do that. Instead of WhatsApp notification and all... Now you are doing business, right? Everyone does product first in business. This is called proof-of-concept or MVP version. Okay. So basically, there are three here. Chat is what we usually do in chat. There's another tool called Co-Work in Cloud. It is called Cloud Co-Work. The difference between this and that is it can execute things. And very importantly, to do things in the chat here means you have to attach everything. You have to put every file and do all of those things.

37:54 Co-work means we access our folder in our local system. So, we can attach that folder and do anything with it. Okay? So what I'm going to do now is create a completely new folder. Okay so I'll go to downloads... There are many folders right? Each folder has one project. Every project will be done in this folder. Okay! Now I am creating the new folders here I'll call it. Raw AI Update. Raw Talks AI. Create this and open it up. Now, you see what happened on top? Let me put that here for a moment. You saw here right? This folder came here. So now whatever is in this folder will be opened. Done! Okay so we got to this point. What do we have to research now? Correct. Lookup the

38:53 look up the AI updates in the last three hours and make it relevant to a content creator who is working in the infotainment, who's trying to educate people in AI so that this person can look at all of this and could help him in his own work as well as help him educate a lot of people. Try to put a score to this AI update based on this and put it all in a neat order Score is like which is more relevant. So if we look at the right-hand side, you can see that there is progress on top. In a little while it will understand what steps are in this. Now just like how we broke down the steps, it will break them exactly and tell us about step 1, step 2, step 3 etc., in the corner of the right hand side.

40:15 So we didn't give any workflow for this. Just a prompt. We are not telling you to go to these 10 websites and do this. The more specific the better, but we are not doing it right now So, basically what you do is that you automatically Breakdown. If I do all this, the task will be completed. Got it! That is what it's doing. So here we are using web search. It searches on the internet and gives us all the news. Look at APN, it crosses there. Yeah yeah Whatever is done, it's updating you like a good employee. This is done, that is done. Progress update comes next to it. If something isn't working out for you anywhere, then again ask what is happening? We can ask. Nice

41:14 When this is running, did you see the clock next to it? It's a schedule. So we can set this as our schedule. Okay. So if I want to run this every three hours, I can do that and your file will be updated like your folder every three hours. Great. Nice. If you search here, it searches all these. LLM News Today, Robotics Week and other sources are trying to figure out this. It searches everything in parallel. It filters all of these and puts them in a way that makes it interesting.

42:04 This is 15+. Since you are an infotainment content creator, every item has a content angle. How you can use it in your own work and how you can explain it to your audience. I will be surprised if the answer is great because we didn't give much context. Usually, when you have to refine something, you have to keep giving it feedback. Hey! Good! Do this and that. This information was good but why did you add this? Why did you add this? Next time don't do this. If you keep on giving feedback like these, it keeps getting better. It's not that easy once you get it. You have to give a lot of feedback, and that is the difficult step. We can't say these are good or bad out of ten. If we don't do this, every time it will be mediocre work. Got it! So now I've got all these things. But let me open up the file as well

43:11 Oh, it's here. The entire file is here. This is saved in our folder on the desktop. Now this is Google Vids, now lets anyone create AI videos for free. Then they say Ultra has arrived. 2 million token context. We talked about that today itself. 1 million? I thought it was a lot more than that. They have come to 2 million now. 94.3% are graduate level science. Anthropic unveils. This is a perfect story for infotainment as drama stakes and a clear narrative. Let's schedule this, no? Yeah. Schedule it up. So if we want to run this every three hours, we click on the dropdown here and we get the schedule. Oh, it's right here? It's right here. If you click on schedule

44:18 Schedule this to run for every three hours. Every three hours. So other document will be updated again? No, it's not that. It puts a new one separately. It puts in different documents. And at the advanced level like you said with NET, do we need any other APIs for that? Especially news related ones from Feedly or something like that? You can run Feedly or something. We need Feedly for that.

44:56 So, if you click on Schedule now it will be done. This is level 1. It's created with the basic file. Got it! That's all. Okay? I'll put this for now and show how to do the next one as well. Done So first level is something like this. Generally, we ask for it and then consult about it to get a file. Second level is using with N8N and then automating. You take these sources from XXX source and every 3 hours whatever schedule we put bring it on WhatsApp. Third is the more advanced thing what you do. Yeah but see the one that I do is slightly different. This...I used to use a process like this before. But

45:38 What I want is perspectives. We have this somehow, you got sources from Google, basically that's it but we have sources from Instagram also, we've sources from LinkedIn also, we've sources from other platforms and also sources from Twitter. But MyScraper one of the most important thing that I wanted is every three hours it runs there too what do I need there? Basically My Twitter feed is the best AI update feed for me because my twitter is active on consuming AI content. The problem with this is that it's not possible to use an API which can see your tweet. There are no APIs which can see my tweet. If I go to x.com, what do you get? You don't get any of that. AI can only search based on keywords

46:29 If I go to Twitter and type in AI updates or chat GPT, it will send me tweets. But what I see on the feed won't be visible. That's what I want. Every three hours a new game comes up on my feed because those are perspectives. Those are real updates. To automate that... And those are customized? Because you like them? Yes by default. To automate that, I need to give my AI the power of using my Twitter account. So, every three hours Jerry, my EA will open my Twitter on my computer scroll pick up the updates and send that to me on my Slack

47:10 This is what happens when a new browser opens and the scrolls are lost. I didn't see that, it was happening next to me. People would say, What is happening Webhook? There's one more screen on the side. What is happening? What is happening? means... I had to explain them So now this is done. I can show you my Jerry Mine is connected to Slack directly And this is my executive assistant that we spoke about Please blurb this corner Yeah Right In this basically you can see top items worth right now and non-duplicate because I don't want duplicate ones. So, this kind of update... What's in my Twitter? Why each one matters? What are the sources? This is what they send me every three hours. Okay! This is coming directly from my Twitter feed. That's why it doesn't look exciting when you show it to them but the way it is happening is what makes it powerful. The news that we will find here

48:10 You cannot find it from the internet. Because internet is usually one to two days late. If this gets viral on Twitter, then they lift a news publication and paste it there. My job as a content creator is to talk about it before Forbes, before anyone else talks about it So I have to pick information from Twitter. I have to pick information from Reddit And I have to pick sources where other people can't use. For example, if there is Reddit. Even in Reddit, there's no API access. Everyone gets blocked. So I have to do something like this or I have to use a web agent

48:46 browser agent to be able to do tasks. So this is a slightly more advanced one, this is all done again using OpenCloud. I'll show you what they are saying in a small way. Because people don't understand the use of browsers and scrolling. Let me demonstrate it for them. Also, since we can't show the whole process now, at least if you are able to tell them tools that go on. To make that thing, what all tools would be required? If you're able to tell them better. So this is Comet browser

49:26 This is a normal Comet browser. I'll tell you what Comet browser is, everyone uses Google Chrome right? We don't use Google Chrome, we use Comet. Comet is like a browser like Google Chrome. Comet means perplexity. The advantage of the Comet browser is that the whole browser is an AI. So whatever work you do in your browser, it can be done by Comet for you on your behalf Sorry. So this is a browser with an AI layer on top of it. Now what we are going to do, I am in my ex feed now. Weirdly enough 41 minutes ago Vaibhav has tweeted while Vaibhav was sitting in front of you. That's good!

50:07 So, there are a lot of tweets here. I can't capture this tweet with any API because what I do is read and like the tweets that I like, then say which tweets I don't like, and then personally curate the feed for it. This cannot be accessed by any API. We have to think very differently about this. Now, who should tell me about this? Correct. So now, agent to tell you the only way is my agent has to open the browser and scroll. Correct. So agents scrolling demonstration is done. Thanks. You can assist them to open JC, scroll my Twitter and give me top tweets on my feed

50:58 on AIC. OK? And then we'll do this. Yeah, yeah. So, my feed is saying that. This is what makes Vaibhav open class without being seen. Nice! Got it. Basically now I don't have the time to open and say do this so I tasked her with this task but the problem when you task someone is that their computer should be open and they need me in front of them. What did I do? I installed it on a computer

51:46 Our computer is always on 24x7. It never closes. Whether I'm there or not, it will open and scroll. And if you say update, it sends me to Slack. We use Slack wherever we are. Wow! Insane On that note... Now that you keep yourself updated about a lot of things What are those 5 top channels that you follow? Profiles, Channels, Twitter... All these updates Interesting. First channel is Vaibhav Srisanthi. Just kidding! For YouTube, there are a few channels that I like watching. One is Matthew Berman. A little technical and it's very good. Two is Andrej Karpathy who was the ex OpenAI member Lex Fridman podcast or one more that I'm always on top

51:46 Our computer is always on 24x7. It never closes. Whether I'm there or not, it will open and scroll. And if you say update, it sends me to Slack. We use Slack wherever we are. Wow! Insane On that note... Now that you keep yourself updated about a lot of things What are those 5 top channels that you follow? Profiles, Channels, Twitter... All these updates Interesting. First channel is Vaibhav Srisanthi. Just kidding! For YouTube, there are a few channels that I like watching. One is Matthew Berman. A little technical and it's very good. Two is Andrej Karpathy who was the ex OpenAI member Lex Fridman podcast or one more that I'm always on top

52:40 Imagina, I forgot the name. What? TV... TVPN. There is a podcast called TVPN which is a daily AI news channel. Okay. Right! It just got acquired by OpenAI. OpenAI. Right! TVPN media company that is also something they do a lot of podcasts and everything with Sam Altman. It's like TV. They come live every day at 2 o'clock. They talk about anything. The other thing is something called as the Moonshot Podcast That is also a very good podcast. They do it three times weekly, usually I listen to it in the gym. Three or four incredible people sit and jam on AI. Yeah these are the ones and of course you can easily make your own podcasts.

53:27 I'll give you a caution. These podcasts are challenging to watch. Oh yeah, that is true! Because she said the one that you follow. Yeah, I told what I follow. We always have this question, right? What channels would they follow? Where does their content come from? Someone we follow... How do they get followed? She's asking such difficult questions. She's witnessing it now itself. Now that we have seen the news agent, majorly students who are in job search. It is a very tedious process right? They have to go through multiple requirements and then they have to apply again after researching whether the company is good or not. Can you automate all of this process because I heard Vaibhav telling that agents can do it from forms so how can we do it

54:21 So, what we are going to do is... This is a software called GenSpark. Again this is... Software? Is it an application or...? Application but it's like... You came here saying Manus, right Vaibhav? Yes Like Manus with lot of tools. Okay! Now you can do many things in one place. It's a super app type concept. Super app in AI You can create documents, spreadsheets and chat in this. You can also create images in it. So that's what all of this is for. In this super app we have an AI sheet. Let us share the job search here. You might be wondering why I gave you a sheet when searching? So let me show you how to do that.

55:10 So, now we are in Sheets. Now what I want from you is to explain the parameters of a job search that a company wants to find out. Go for it. OK, maybe searching jobs for content writing especially for media agencies and YouTube I mean current YouTube channel media agencies. Yeah. Which city? Which country? Country, India Remote preferred Best

55:59 Also include what is the expected CTC What are the skills especially that this job requires so that at a glance I can see what is it that i'm looking for and give me also you know some kind of a percentage score on what things information have given how much it matches. Done, also can we also make like old companies not like just started startups maybe

56:39 Not bootstrap startups. Try to find jobs only in funded startups and not bootstrapped ones All large companies. It is now trying to find the content writing jobs in media agencies and YouTube channel. India with remote preference focusing on funded startups and large companies. So first thing, it's the same as our co-work tools or any tool agent like here task is given. First this search for content writing job in media

57:21 Okay? content writer jobs, funded startups. So this is different from normal sheets because you have to put data in the normal sheets. This is intelligent AI sheets. It will find and populate your desired data. Okay okay nice! So it's like Google Sheets and Excel Sheets on steroids. Because it uses AI muscles to make it think that I can't give my own data so I'll get mine myself

58:13 It could also go and it can look for forms and fill the form on your behalf, attach your resume and apply. And you see job title, content creator, content ops executive, content writer, intern. We didn't say full-time or intern, so we have an intern now. Okay? Company type funded startup funded large company more funded a large company agency funder do and new Xander. Oh fun funding status would it's funny shadow location Mumbai Bangalore Hyderabad and new say mother a could've said

59:05 Then, remote. Hybrid, hybrid, hybrid on-site, remote, remote, remote, hybrid on site. All these came up. What else is it doing? Doing a summary. It did the summary. Then key skills required What skills do you need? What is the industry sector? Then, source accounting. In case you want to trace it back. To check if what I said was hallucinating or an actual source, You can take the company name and the source and double-check them. Nice! Another easy way of double checking is to download this sheet and put in PubText. If they check whether it's a real job or fake one, we can see that.

59:46 OK, summary insights. So this is how it is. So this is GitHub repo. So basically, what we've seen now is a version which is the simpler version. Simplest version? Simplest version! But even though it's the simplest version, think about it. I go to the internet and search for jobs on different sites, and then job descriptions are made up of things like, oh no this isn't right, or oh no this isn't funded startup, or oh no that's not right... Without any frustration, we just put them together and drink tea and come back with everything ready

1:00:39 How do I download this? How to use it, all that is in the video. Yes, he has put everything in here. What are these and what are its features. See, he says auto pipeline. If you paste a URL, it will tell you about full evaluation PDF. Interview Story Bank. It will take different stories, 5 to 10 master stories that can answer any behavioral question. So this is pipeline integrity. I'll call it a bit technical. So for example batch processing. I could do it with parallel agents as well

1:01:20 If I want to hit 100 jobs at once, it will run with 10 agents in each job. So negotiation scripts. You have to negotiate your salary. If someone sends you an email asking how do we talk to them? That's one thing. It gives training for everything else. Another very important geographic discount pushback If I'm in India and you're remote, so if they give less, how do we tackle them? That's also given. And that too customized for every company with your role how it should do. Nice! So this requires basic coding knowledge. It doesn't need to be. You need to know what cloud code is. You have to learn cloud code basics The reason why I am recommending not to show is one, people who are not engineers

1:02:13 You'll get scared. There's a code and there is some learning curve for that. But, that doesn't mean you have to be a software engineer. Right. You don't need to copy paste it. Look at the guide in Quick Start. We should not copy all those codes in terminal. That's all. Actually, we do not even need that. What I will do is, I will take this URL, copy it, put it into Codecode, and let you know what to do with it. Best! Figure it out. I'll give you a prompt, this is this. I have to use this. I don't know anything about this. Do something! The backstory of this is the guy who built this basically he was applying for jobs. He wasn't doing it manually but using cloud code he built this software so that AI can apply jobs for him

1:03:05 He applied 700 jobs because of this. And he got the head of Applied AI job also because of this. Nice! Then, since it works well, they open sourced it. We are showing that so more people can use it. But to make a full video for this, you need at least 2 hours. Let's put aside what we built and try to brainstorm on problem statements. This is something I have been excited about lately So the surprise element is we're going to open, you know curate this with seven guests from our previous podcast. Interesting! So that low I thought last time when we spoke about it, we'll try to pick a few problem statements from students and all that right? We will narrow down those and leverage them on our channel as well in that sense we've identified everyone and talked to them

1:03:05 He applied 700 jobs because of this. And he got the head of Applied AI job also because of this. Nice! Then, since it works well, they open sourced it. We are showing that so more people can use it. But to make a full video for this, you need at least 2 hours. Let's put aside what we built and try to brainstorm on problem statements. This is something I have been excited about lately So the surprise element is we're going to open, you know curate this with seven guests from our previous podcast. Interesting! So that low I thought last time when we spoke about it, we'll try to pick a few problem statements from students and all that right? We will narrow down those and leverage them on our channel as well in that sense we've identified everyone and talked to them

1:03:56 So, there might be a question or too much complex situation. They are waiting already. You are from the movie industry, so they have different industries as well Let's try to take the calls I will do one by one and not give the list also I'll give you a heads up okay? Yeah please Right now... Now... I'l be very open and frank When we go in problem statements And that too from different industries We are jumping a stage where

1:04:41 We cannot talk like the way we were so far. Right now, my assumption is me and Dilip will talk the operator language or the AI orchestrator language, the AI generalist language. So before we get into this quick caution When we get questions, when we answer or discuss them I think we'll have to discuss because there's no right answer. We have to brainstorm and understand what could be the solution And that will open up a different side of us. People might wonder why are they talking like this? It is very technical AI, it can't do it by itself. That feeling may come. Why does it happen

1:05:31 We started with some workflows, tools and stuff like that. To make it clear. Then the language we speak is much higher because we are jumping from being dabblers directly to an orchestrator You cannot get to the position where you can talk like this. Absolutely! Right? So, don't get carried away and if you guys have questions feel free to ask as well. Sure sir. We will discuss that also. We'll be layman's or we'll be students like that and then if there is any complex issue in your discussion, we will directly ask you guys. Done. And I do not think we will able to build everything but

1:06:10 But we can give direction. We can show them a reference point, not necessarily building something... Brainstorming on how we solve problems. I love that! Let's do it. So even the viewers who are watching will understand and the guests too will understand what would be an ideal process for transferring. Shallow then? We'll start. First case is already ready. Okay. Niranjan Pagadala, I call him Annan. He is founder of 8Views Digital Marketing Agency I have a guide, friend and generally any hardware or technology aspect. So I thought he will be the right person also to start this kind of conversations.

1:07:01 So, Vaibhav is here. Pranavi is here. Dilipanna is here so we are starting this conversation maybe we will start with what do you do with your team I mean with AI in your team how do you utilize or adapt the AI tools? After that you can go to the meeting. We will continue. Yeah, so VK I am... AI, AI, AI, Win2, Win2 as an entrepreneur lot of fears daily you know newspaper is going to be published and this company is going to close down marketing agencies are going to leave again. In news if we don't feel positive about it then Instagram opens up and there also the same thing happens. And when office closes, okay let's do something in AI

1:07:49 In the meantime, we face other 10 problems. As you know, with employees, man management is always a big issue. One person comes up with X problem and another one says relationship problems and another one says he doesn't have motivation. So, we get stuck here in deep. When I go home this evening People at home would say, AI is coming. Will your company exist or not? So the subject of AI has become like this. When I play cricket, it's called eat, drink and sleep cricket. AI has also become like that. And when it comes to the employees one thing we are seeing is... What I want to tell all AI experts is to simplify discussions for many people. Many don't understand

1:08:34 Right. I have a friend who has good idea in AI and another third person who doesn't know much about it, but we are trying to teach him. He is saying that he builds agents and all this happens. The third guy said if you become agents there will be no problem for me because I have 150 people. So first simplifying the language I think it will really help in the entire ecosystem. To simplify all this, what I started doing was to use only one tool and platform other than Chahar GPT, Gemini or other perplexities. So I came to Claude. My vision eventually is that can I automate digital marketing through AI?

1:09:25 by 70% at least. But building that vision of ours where our company can do anything for the entire 70%, is where you know, our brains are working right now. Yeah. Any cross questions before the call ends? Good. We got a question, right? Perfect, bro! Now we will start this discussion. Okay okay. You guys all look like dark webs and I am in black. So, based on my understanding the question is how can we automate running ads as much as possible? I think we are already in a position where you can automate 90%.

1:08:34 Right. I have a friend who has good idea in AI and another third person who doesn't know much about it, but we are trying to teach him. He is saying that he builds agents and all this happens. The third guy said if you become agents there will be no problem for me because I have 150 people. So first simplifying the language I think it will really help in the entire ecosystem. To simplify all this, what I started doing was to use only one tool and platform other than Chahar GPT, Gemini or other perplexities. So I came to Claude. My vision eventually is that can I automate digital marketing through AI?

1:09:25 by 70% at least. But building that vision of ours where our company can do anything for the entire 70%, is where you know, our brains are working right now. Yeah. Any cross questions before the call ends? Good. We got a question, right? Perfect, bro! Now we will start this discussion. Okay okay. You guys all look like dark webs and I am in black. So, based on my understanding the question is how can we automate running ads as much as possible? I think we are already in a position where you can automate 90%.

1:10:14 The technology is already there. Right? And... Many people don't use it because, this came before AI actually. Ad automation was there long ago where you can buy and sell but AI added a new angle which is even the creative part can be automated So we'll talk about I can't show you stuff But I can tell you how to go about it and I can tell you the tools. If you want to explore, you can explore or else it will become very marketing centric. Let's say they have meta ads. Meta ads on Facebook and Instagram ads. There are two aspects in that too. Marketing on Ads

1:10:52 One is, we need a creative. We need an ad creative. Another thing is when running ads, does it work well? Does not work properly? If it doesn't work well, you have to turn off the ad. If it works well then increase its ad spend. You have to launch new ads That's all ad operations. Ads sets. Yes, ad sets and campaigns. People spend time on this for a long period of time What are the problems that People solve actually. Performance marketers do that. They look at the data and say, we spent 100 rupees today. We should get 10 sales but only 8 came in. Why did 2 sales go down? Everyone goes to see the data. Is there any ad creative that stopped working or not? Or is it because of a number of clicks which is CTR, has anything reduced on creativity? Maybe some ad campaigns are not working so they have to switch off. We're wasting money on that

1:11:44 If people are not marketers here, they will not understand. I'm trying to simplify as much as possible but if you're not a marketer just forward this But that is what you do So both the things can be automated The first thing one of the tools two tools that people can use for automating ad management A social... a Facebook ads manager basically does this which is called as optimization. AI automatically detects it and increases it

1:12:30 If you give this campaign to the punch-out, it can automatically reduce also. So there are a bunch of workflows it can do look at this It's called optimization This is ad optimization How should every ad be optimized? What should be the time spent on ads? Look here How should we reduce targeting? What should we do for purchasing spend? Basically whatever performance marketer does, it automates all that for you You can also give your competitor's data. All these are my competitors. What does it do to them? We can learn from them too, right? In Facebook Ads Manager, you could use other ads as well

1:13:23 AI will also see it. If you are a competitive test, if an ad is running for more than X number of days, will someone run an ad which is working? Which is not working? No way! But the duration of the ad is visible on the algorithm. correlation. Not always causation, if the ad runner on that side is lazy and he's running ads for a sale starting from money-spending, then what to do? If there is such person, you shouldn't be a competitor first of all! Right so it can pull ads from there also right now it can pretty much quite frankly there was a point of time this is pre AI Pre AI There was a point of time we used to have if you ask Dilip We had 15 performance marketers

1:14:10 Now the business is almost 20 times or 30 times larger. We have 5 because of ad automation, ad rules. It's not even crazier than creatives. Think about earphones for example This is your physical product You want to create ads for this physical product with different people with different backgrounds People giving reviews on this product I use these earphones they are very good They're amazing, awesome blah blah blah all that How many ads do you create? How do you do that? Let's say your customer writes a review on Amazon. He wrote a positive review, but it is a review on Amazon that will not work as an ad. Correct! I can take that same review that is written by our customer let us say his name is Dilip and I can convert that into an ad not with Dilip's face but with the guy's face. Is this the tool we used before? Magic... Magics. So is that the same tool which you use right now? Cool yeah

1:15:11 It's the same tool that we use. I'll tell you why a lot of people don't use this because they charge percentage of your ad spends, oh! They charge 2-3% of their ads spends. So many people say, are you spending so much money? But a human cannot do what an AI can do. I have realized this two years back. How much are your ad budgets approximately? Its in crores per month In cloud we can also connect our ad manager and review it from there. That is what a human marketer does, the action is still taken by humans. When Niranjan said I want to automate it

1:15:51 So, you trust an AI to do its job. Got it! So automated... Good thing about what Niranjan said is he didn't say I want 100% automation. He said 70%. He said 70%, I'm saying 90%. Why does that 10% exist? Still someone has to look at it. Correct. Someone needs to have a human instinct. AI can only understand basis of what it's seeing. There are more things than we see right. That is where you need that human even now. A performance marketer, a marketer if they want to win in this AI first world. They don't need to know how to change the things. They need to know how to ask the right questions. That's it! Insight on matter. Nice In fact I am seeing a tweet Majana The tweet was about How marketers are new software engineers? Last generation software engineers were the ones

1:16:46 who are in highest demand, which is still true by the way. It is still true that it's not going away but if someone can overthrow the demand for a good software engineer that is going to be people who are good marketers or salespeople. Why? Because when we're living in a world where anybody can build great products The most difficult thing, the most difficult thing is to put the product in the right people's head. That can be done by a good marketer and a good salesperson someone who understands distribution so I think it's a platinum era for marketers not even gold platinum era for good marketing sales and good business folks So now it is Prasad sir

1:17:31 Kalamandir, basically Sai Silks is the main company which is listed also. And when I met Sudhanshu for the first time it was like... I went inside very scared and after 10-15 minutes I thought that there will be a kickass podcast so I came out. So after that we got to know he's very much into technology. First, no one would stop at just one second with any technology All the team, from his son-in-law to his partner and all of them are talking about technology first. So in a podcast there is one teaser I will show you the teaser we're not a saree business We are a technology company selling sarees that's what he believes in sir over to you now AI we were discussing so problem statements discussed on its solutions Vaibhav was there with us

1:18:19 Yeah, first thing. Hello Vaibhav I have to appreciate you Thank you so much Amazing You are doing great I want to thank you again for this opportunity. Thank you sir! I am a little bit into technology

1:19:07 Actually, what I need now is for my business on a day-to-day basis, WhatsApp and Slack. Then SMS and emails. So these are the four or five different mediums of communication with entirety across the length of breath we have thousands of people between our employees and our viewers. Everybody is more into technology. Yes! Everybody knows everything kind of thing So, in a few hundred books and messages I always see that I will be missing out on something. Correct Which would be so many things are critical and so many other things are okay to make. Right. So, generally what I tell my people is if there's any time sensitive issue, call me or send it personally but again that is not 100% professional. Correct

1:20:06 So, I want to be empowered. I need a tool for all these communications and study them well so that the next day morning I can have a summary of things you missed out on or these are the things you need to prioritize 1-2-3-4. Not just me sir but everybody is flooded with so many things and opportunity will not knock it in for anyone We have to be prepared when it knocks the door. If you select a general use case, not just for me but many people will get a good result.

1:20:45 So got the question. Sir is saying that my life is a mess because of so many channels of communication. We can organize them and keep it to ourselves. Okay? Let's start with the first one, which is let's say meetings online meetings. For online meeting he could use a simple tool called as Granola. Granola or Firefly AI. Yeah Is it similar to Read.AI Yes It's similar to Read.AI or OtterAI But the only advantage of Granola is for people who do a lot of WhatsApp calls. So if you take WhatsApp calls, and use your desktop app on your computer to take WhatsApp calls, Granola can take meeting notes off that also. Oh crazy!

1:21:28 That's a big advantage in Ranoola. Okay, that is one. Second for physical meetings how to organize it as we talked about yesterday or plots the physical device or India has a startup quick shout out good guys good kids trying to build something called as Neo Sapien Oh yeah I heard of it It belongs to us so we should promote it 100% Neo Sapien okay It's a physical hardware. I have a similar version. They just released it, I've been using it for the past year. It's from another brand. But I've tested this and its pretty good as well which is called Neo Sapien. I wanted one anyways. Yeah. Neo Sapien is a physical device that will be on at any meeting you go to. It even shows light so people know you're getting it recorded which is important! Which is very important. After all the meetings and recordings are done

1:22:25 End of the day, it will give you a summary of everything. And also ask questions like what happened in last 30 days? Did I forget anything or something else? That is Neo Sapien for physical meetings. Next we are using Gmail Slack and stuff Now the big question How will i bring all this information to a single platform So, Granola. It has something called as MCP. Model Context Protocol. Right? Now using MCP and using let's say N8n, you can fetch all the information from every single meeting of Granola that is happening every single day. Same, NeoSapien also have an API. They use the API

1:23:08 you can pull all the information into a single platform using something like N8n again. We saw NeoCPN, right? Next is... This is slightly technical. This is called Google Workspace CLI. Right? It's a GitHub repo. If an agent connects to a workspace CLI with N8n, Gmail or Calendar, anything of Google services it can pull information from. Nice! This is the workspace here. From this, very easily... It's very simple actually. It looks quite technical but it's quite simple. What happens from here?

1:23:45 You can pull information from Gmail, Calendar, anything Google services. Yeah, Google and ecosystem. Anywhere you want. So we got it... We got Gmail covered. We got calendar I mean all your meetings covered physical or online Next is Slack. There are a lot of slack bots in Slack which summarizes everything that happened on Slack Now coming to Gmail for gmail also Basically, to get all this information at once you can use Cloud. So here by adding it... Here for example we need Gmail. In the same way we need Slack summary

1:24:33 You go to connectors, again add connectors. You go to slack and you can add it. If I connect this too in the same way, I can also connect granola. Oh nice! There is that too? Yes, there is. It is also like a granule. Right? Similarly, Neo's API has an API. You can use the API of Neo's API to convert it into MCP. You have to learn this. Go to Websext and ask them. How do you convert an API into an MCP server? Once that is done, you can connect here also. All sources are connected now. After connecting all the sources, all you can do is come

1:25:15 here and just say check my gmail, check my slack Check my meeting assistants and tell me what happened in the last 24 hours. Claude will take all the information and it will give you a summary Now because we have connectors which is the Slack connector With that same Slack connector you can also say this everyday once this runs use the Slack connector to post this message into my DM So, Slack will also post a message directly to Cloud. Okay. Nice. To my inbox? To your inbox. Now the only thing that is not automatable... WhatsApp. ...is WhatsApp and the reason is very simple because Whatsapp are end-to-end encrypted so they cannot break that encryption as a result Meta doesn't allow In fact Meta AI, yesterday dropped a very powerful AI model right even that meta AI of meta cannot read your messages it doesn't have access

1:26:10 Does that mean it will not happen? I have a way to make that happen also. Okay, okay It's risky But what i would do is...I will go to connectors and we'll search for Chrome. What if I install chrome and... My laptop has been open 24 hours. 24 hours there you are! That's it she's thinking like a Orchestrator. So I have my WhatsApp open, right? And it just keeps scrolling and then leaving. Every 24 hours, I will know that it has access to my Chrome. How do you write a prompt in this? Forget about Gmail and all the connectors but for WhatsApp Open My Chrome Scroll through all the 24-hour messages and give me a summary of them. Would it open message too? Once it has access to the browser

1:27:09 It will be able to do anything you want. Instagram is just a scroll, right? We don't have to go to every... It's not like a feed, it's in the individual chat. You can do whatever you want. So when you click on your mouse and control it, it'll take that mouse and click as well. Does it do that? So what's the risk element here? The risk is giving your WhatsApp. You said last 24 hours. It is not smart so it will end up opening a message that you don't want to open Okay, so we connect a connector. So I can just say use my browser via Chrome MCP and check the WhatsApp messages in the last 60 minutes and give me a summary Don't reply to any message Okay? Yeah

1:28:12 Let's see. I have not tried it either By the way, just saying... I have done something like this using my open clock Using my Jerry It didn't listen to what i said It didn't listen? I mean if you say that it didn't listen... I wanted my Jerry to work from WhatsApp via Sockets When I set up... It is supposed to respond only to me I setuped it at 3 or 4 in the night I set up the app, it was replying to me and everything worked out. When I woke up in the morning there wasn't a single unread message. Why isn't there any unread messages? Is my WhatsApp not working? Usually I wake with 50-60 people messaging me in the morning. When I opened WhatsApp and checked if there were no messages, all those messages replied automatically

1:29:05 Someone said, hey do you want to... Are you available on Wednesday for a talk? And I carried it and said yes. I'm available. This that and reply WhatsApp group messages everyone was messaging Good morning! I am replying good morning Have a great day! It's important. Sometimes there is no benefit but sometimes there is loss too. Again some branch had reached out Hey like what is the thing? It made up a budget You have to pay $25,000 for this. Let me know if you're interested." I replied like that. Wow! Nice. Then I thought WhatsApp was the answer. Jerry is very fast. WhatsApp. He did what I told him to do and it's not Jerry's mistake then i started to read about it The reason why it's happening is Jerry is not able to understand whether I messaged or Vamsi messaged

1:29:56 It was only supposed to reply to me. But if you still want to use that risky technology today, I can still give Jerry access the WhatsApp. Now the idea that I got as we were talking about this is, if Jerry replies to all my messages, it means it can read messages which is OpenClaw. OpenClaw and ReadMessages So what if I connect my WhatsApp to an OpenClaw instance and strongly recommended not to reply to any messages. If you don't want to reply, then don't do it! And your only job is to read the messages So I can read all those messages and run a script Every time a message is coming Read it and store it Read it and store it And also tag it You can do whatever you want Yes yes Read it Store it Tag it Whatever you want And every three hours send me a summary so it can do it

1:29:56 It was only supposed to reply to me. But if you still want to use that risky technology today, I can still give Jerry access the WhatsApp. Now the idea that I got as we were talking about this is, if Jerry replies to all my messages, it means it can read messages which is OpenClaw. OpenClaw and ReadMessages So what if I connect my WhatsApp to an OpenClaw instance and strongly recommended not to reply to any messages. If you don't want to reply, then don't do it! And your only job is to read the messages So I can read all those messages and run a script Every time a message is coming Read it and store it Read it and store it And also tag it You can do whatever you want Yes yes Read it Store it Tag it Whatever you want And every three hours send me a summary so it can do it

1:30:53 But don't come and tell me if it ends up replying or something. Because the technology is too early, so I don't want to even say this and say it. So that people know it's possible. Got it! It opened my WhatsApp right now. Yeah. I said last 60 minutes let us see what it is doing? I can't see anything though It's trying to figure out, the first thing that it is doing is trying to figure out what is in the messages. Right? How to navigate it Once it figures out how to navigate shell A and T second time it becomes fast High level summary. You saw Key takeaways Looks like family event or puja happening with photos and congratulatory messages

1:31:40 I'll tell you another thing. If we load all this information into the cloud, no matter how top up our plan is, it will be lost in a day Okay. This is okay, it's not so much but yeah he is right. Pro plan is not at all useful. If you want to use connectors to the full extent, you need Maxplan. 200 dollars. Right? But logic I mean overall what i tried to do today or what i try to show you is everything that he's saying is possible using connectors simplest way to do it except for whatsapp whatsapp could get a little technical but except for whatsapp everything is possible i hope mr prasad finds this answer helpful hopefully yeah

1:32:41 Doctor hi. Hello aunty, how are you? I'm great how are you? Absolutely absolutely fine How is your leg? All good sir. Recovering. But yeah doctor as we discussed over to you We want your problem statements. What were the things that you wanted us to actually discuss about and maybe if we can come up with solutions what would be a good point to discuss Vaibhav is here. He's in the video. So yeah over to you Firstly, Hi! My name is Dr. Virenchi and I am a psychiatrist. When practicing as a psychiatrist one issue that we usually find is after 2-3 days of consultation my patients call me saying they need to talk urgently with the doctor. So when it comes to urgent or emergency

1:33:36 Usually, the issues that arise are genuinely what is an emergency. If a person has suicidal thoughts or self harm thoughts, it's one urgency which they talk about Some people feel that their issue is urgent. Sometimes we get calls at 11 in the night, I am sitting with friends and everyone is drinking. I have already taken your medicine so should I drink or not? When such queries come up sometimes what happens

1:34:17 we might miss out on something very important and you know, because of that we feel bad. It would have been better if we had prioritized this. So looking at it like that is there a possibility? Because when they call the call center, they can't say everything to the call center so is there a way where we can use AI where probably type in their complaints and AI can prioritize. The doctor should see this today itself, to say that it is not that important. My second question is, recently when we spoke with Vamshi ji about digital addiction, screen addiction social media addiction, the social media addiction has increased a lot. Because of which depression and anxiety are increasing

1:35:12 At this time, is there a possibility that we can identify people's patterns or themes of thinking through social media behaviours and find out what kind of mental illness they might go into tomorrow? What is the risk? Can you tell us anything about it? These were my two issue points that we are facing in today's world. Got it doctor! Right, so thanks for doing what you're doing. I got your problems and i think uh...I can think of a couple of things that can be done which can solve your problem we are going to explore that and you'll probably see that in the video when it goes out Thank You Doctor! Thank you very much! When I am talking about patients, I would say that the doctor's voice is most soothing and comforting

1:36:09 So now we have voice cloning technology. We can clone any voice, and build a voice agent. You might be thinking what happens if the voice agent gets suicidal thoughts? What should I do? Now that our voice agents are capable of doing this, we set conditions and hand it off to humans We have a call transfer system, right? Likewise, Voice AI agent will call the human. So if we get 10 calls and one of them is an emergency series where the patient is self-harming, if we detect that call there, it immediately goes to the doctor. So all the calls to doctors are emergency calls. For example, I'm taking this tablet, I want to drink something, can you give me something to eat?

1:37:00 we can filter there. What are your thoughts? getting a reply. Usually, then if we build this chatbot different entity from our chat GPT for 90% of those people who are not serious when non-serious people first call is non-serious or second call non-serious or the fifth call of that non-serious person could be a serious call and if it's getting generic advice every single time

1:37:39 Then the person may be calling us. While AI agent, the voice agent ability is a very smart thing I have two things to this Voice agents can build without code in retail or WAPI. It's simple You can build full flow If it detects something is off then automatically it will set off to human which could be doctor that solves the problem but One problem, what doesn't get solved is personalization. So why do I call a psychiatrist when I already have AI? Or why do I go to a psychiatrist when I already have Google? It's because I believe that person knows enough about me. Correct. Right? So the answers that come up... If it is not personalized for you then it is not very effective. That's my belief. Second

1:38:37 For patient A, should I drink or not drink? The answer could be different and dangerously different based on who you are. What if you're an addict? What if it's okay...I don't know if it is okay but what if you can have a drink even after drinking alcohol? So you get the generic answer that yes you can drink, don't worry But what if you were an addict the answer is no for you. Correct. Ganiyoka generic bot, yes I am a pest study in this case in case if it's okay to drink after medicine which I don't know because I'm not a doctor but I am just giving an example. In this case context history of the patient is important correct

1:39:24 So, my approach would be while voice agent is a great way to do or we can do a WhatsApp voice agent also or a WhatsApp agent also. Voice cloning could have changed using smallest AI or 11 labs or... It's almost like the doctor himself? It'll sound like the doctor I have problems with that also. Okay. When a doctor talks in an AI voice, will he what will be the perception of the patient when he talks to the real doctor. I don't know what it is, so if i am in a person, I would make AI voice separate and I would make the doctors voice separate but again we don't know what will work or not So you have to test it If I have to solve this problem from ground up I will build an app where which will maintain all the records of the patient in a single place

1:40:17 Doctors and client conversations are notes are taken by the psychiatrist. Sometimes they record also for their own note taking, right? Can I build an app which is completely secure? It's like a patient and doctor rules following, so the data is not shared with anyone. Where all our conversations, medicines that we give if I'm a doctor, the kind of prescriptions I've given the kind of things that you're supposed to follow everything is documented and created, it's created into an app called using RAG. RAG is basically an AI with a lot of documents access to it. So before giving the answer to the AI, It will look up all the information of you understand who you are understand the latest conversation then we'll give you the answers so for every patient I will build an app

1:41:11 I will build an app which will create a space for every patient. And, I will attach the phone number so voice agent gets calls from that phone number it knows that its from patient Vaibhav because it knows its from patient Vaibhav It understands all my problems and gives me contextual solutions rather than generic solution So this would be my approach In fact there is one medical startup in US which does summarization of conversation and prescription, which is a billion dollar company right now. Do you remember the name? It is made for doctors. They prescribe medicines, right? You know what I am talking about because of so many patients. So this app does the bookkeeping of that. It's a massive business! What I understand from this is if Dr. Virendra Chauhan is talking about this there's probably an opportunity Why does Dr. Virendra Chauhan have to build this?

1:42:13 One of the viewers who are listening to this could build this for thousands or hundreds of thousands of psychiatrists that are out there. Which is secure, they have to do it with the focus on security and privacy. And the best part is all we talked about now doesn't require coding You can start without code where you can wipe code products on red plate or whatever and get started. Build a MVP get started, go talk to two three doctors. Understand maybe you are sitting on a multi-million dollar business. Yeah. So yeah. Hi. Hi Vamsi hi how are you? I am great so now this is Mr Achitesh Korupulu. We cracked it nice. So anyways so over to you as we discussed problem statements

1:42:13 One of the viewers who are listening to this could build this for thousands or hundreds of thousands of psychiatrists that are out there. Which is secure, they have to do it with the focus on security and privacy. And the best part is all we talked about now doesn't require coding You can start without code where you can wipe code products on red plate or whatever and get started. Build a MVP get started, go talk to two three doctors. Understand maybe you are sitting on a multi-million dollar business. Yeah. So yeah. Hi. Hi Vamsi hi how are you? I am great so now this is Mr Achitesh Korupulu. We cracked it nice. So anyways so over to you as we discussed problem statements

1:43:10 Hi Vamsi, hi Vaibhav. See basically construction industry the biggest problem what you know, we want AI or these sort of things to solve is that whenever we start our sales, we commit an end date to our customer. Right? Like let's say you bought a flat today I'll say after four years i'm going to give it give the handover the flag now in this four years We do and are like various amount of activities and you can say man hours of labor right almost typically average of our communities are like thousand apartments and each of this community has lik 8 to 9 million man hours. So these eight to nine million man hours will be arranged in a particular sequence that is what will give you end date right? Our problem is in this four years

1:43:53 For example, Google maps location of Delhi. Google maps tells us which way to go. It gives a travel route and also the time of arrival. The same building will be in Google maps as well There's a very interesting problem statement. I think it goes into predictive analytics. Can we do it? I think, It is very much possible We will explore some ways to do an MVP on how you can test out something with your own data that you have because you will have understanding of when did this project start and when did we commit If there was any delay then why did it happen What were those stances Once we have the data I think its...we can build simple model which

1:45:00 you know, give you an estimate that hey this is the safest date that he should give based on all the parameters type. It's more of predictive analytics is what I imagine but I think today models are powerful enough that you don't need a ML team to be able to build this right so we will explore ways on how to do this in some form or fashion today but this was very cool yeah because attached to this we can do various automations and purchases and you know what do you say material dispatch to site like generally when we order material the person who is ordering material at the site right, store manager and all he will simply order material based on how much did I buy last month and how much should be bought next month but imagine in 2 months there are elections and labour is going down so we don't get that many materials for storage so all of those things automated essentially what Eros does is a site engineer

1:45:57 like managing labourers, supervision and material availability etc. Ideally the site engineer should only be thinking on how, like what is the quality of my work and are the labor available or not. Rather you should not worry about drawing on that material. So we have also been doing some experiments around this and building some genetic algorithm models Amazon used some of these things in their delivery whatever when they do package deliveries yeah so we've been exploring but Not many conclusive results as of now, but hopefully we go there in some time. The question that I have for this is do we have data? Because India real estate doesn't operate on blockchain so there are no clear data traces there But these datas will be in pockets with companies like yours

1:46:51 Right as this specific I would say market level data layer though, but each company in measures a particular way. Correct? We have been measuring like we've been measuring over the last five years now We are where we have like almost 10 million square feet of data only like at which point what did we do and all But yeah still will have to you know trying to make meaning out of it. Thank You. Thanks a lot. Thank you brother. See you. This is a big I think you have to model this out based on data. Data, I am getting doubt now is see maybe in their last 5 years journey they might have done 20 projects. Those 20-30 projects are in 30 different locations and the conditions there are different. So it's a very dynamic data right? And sampling also if the existing data itself is only 30 then how much sampling will be there

1:47:43 I always get that doubt. How much data do you need? These kinds of problems. Not a lot actually, if we think about Hyderabad alone there are pockets of regions There will be external conditions and internal conditions based on what i imagine. I don't know much about real estate so I'll tell you what I understand. External conditions as he said, you will have elections That's an external condition You will have IPL That is an external condition These are all external conditions which are relevant for whole of Hyderabad That is the easier bucket to play. Then there are internal conditions, that almost becomes uncontrollable. There are external conditions that we usually don't bake in for example rain

1:48:27 There is no data point that is flowing in but that data is available on API for the next 30 days, 60 days 90 days also. Will it train? Will not train? Estimated yeah right on top of that only also a lot of external conditions are more like you know Elijab Pali Let's say in Hyderabad, this period gets very hot. As a result efficiency of workers go down right? Those are clear patterns based on city. So if you have external data and then project let us say you have done 100 projects. In these 100 projects what were the delays why did those delay happen if that part is documented I think it is enough

1:49:12 to build something better than a guess. So, that's the first step right? Instead of saying I will deliver in 3 years, if this model says it takes three and half years you'll say four years so your safety net increases you're not guessing right and the rationale can tell you the reason Why it said 3 and a half years and not three years? Because it will say, look, you are doing in Cocapet. In the last three times or four times of Cocapet area these are the big reasons Now in future what I am imagining is this patch of month is going to be very hot efficiency dip then there are gonna be rains

1:49:52 And then there is this Bay of Bengal effect that's going to happen, which we are...which is already out there. The climate department has already spoken about it and you have elections coming up and then whatever other factors because you can do a deep research to understand what is the impact on this region? And tie all that data together to tell you these are all possible things that can happen. So, maybe this kind of data is already seasonal data. It's not just 5 years old... We can find it out. You will know what the problem is later on. Because in the past these are the 10 big reasons so they do something called as weighted score that you can talk more about. What happens with weighted scores is we have to weightage what thing leads to what amount of impact. Right? They do this in ML right?

1:50:44 So actually, I'm thinking of the same problem. We can measure what's in our control in external and internal conditions. So if it rains, we have a lot of rainwater, let's get more material. Or there is some material on the road that we didn't put in shelter because it gets wet. It's easy to do all this because data is available. Connecting data like this with an API and sending alerts via WhatsApp makes everything easier So, that's why I'm a deconditioned alert. Climate condition alerts are for when it rains and there is a bus stop outside. You can place the camera where you would put up the bus stops. The camera can easily tell if there is a bus stop outside and send messages to the camera as well. Using OCR, you can do that?

1:51:27 Easy. Even if we don't update, the camera will look at it and solve it. If there are Google Nest cameras installed outside your house for security purposes... ...if a person is able to say that someone has left something outside, how? So, how it works is that it takes a picture. Simple! Very simple. You take the photo in your phone and send it to Gemini. Gemini can answer what's in the photo right? Imagine the same thing happening. It's the same thing that's happening. There's a camera taking photos of you. It uses OCR to send through one of the AI models and tells you what it is. Computer vision model. Yeah, OCR. Crazy. Hi brother

1:52:09 Hi! Can you hear me? I can't hear you. One minute... Yeah, so this is Shashank. You're in Delhi right? Mumbai? Yes, Mumbai. I'm going to the airport now. Oh but thank you thanks a lot for managing your time. How about you? Now generally when we come into industry there's production planning right? Generally what they do is if there are 5 years of data Year on year production is one. Depending on the trend, Monday to Monday transaction Tuesday to Tuesday transaction will be generally like that But now that Nilofer has become a must visit place in Hyderabad What happens now is If AI can help us In planning and production How I am saying this One Monday it would look like

1:53:06 We can't say. Too many functions might mean more flights, too many events or maybe in way of any which way... I'm sorry to cut you. There are outlets like this at airports? Yeah so... So now let's suppose there is a 500 bus in departure. One person comes and the other two come and take away 200. If I have to send 500 banmaskas again, it will take me 2 hours. Departures, security checks, CISF... There are so many things. So we can't say how long the day would be. We couldn't do it according to last Monday's production. Last Monday and last year as well, but sometimes production planning goes wrong 30-40% of the time. Because there is a ban.

1:54:03 How can we address this in the air? This is not because of planning. We can work as much as possible, but how do you take sales last year and so on? We cannot do market analysis. Interesting! No I mean correlation of inflow of traffic at the airport And if 100 people come, 2 people visit Nilofer. Right? Right. Our correlation analysis is Kunte and second source is If you are able to get real time data of consumption on an airport at a flight then that prediction is to a level possible Got it But its very hard problem I wish we could solve this problem That will be billion dollar problem

1:54:54 Thanks a lot. You guys have put me in trouble. We'll discuss. Okay, thanks for your time. See you soon. Watch it on video later. I have an idea. It's difficult to find out the airport footfalls. We can't get that data. The problem here is, in our case of real estate, he has the data. We can collect and silo his data and solve his problems. Correct! Here the problem is, he needs external party data to solve this problem. So if we have to do it like this

1:55:32 We have to guess estimate here. Guess estimate? Yes, estimated guess. Estimated guess, okay! So let's use a small logic in this. Flight tickets data is available via API. We know the average flight ticket price from which city to which city and how much it comes to Hyderabad. Major flights major cities. Everyone has come once to Gump right? If you say that they came to Hyderabad, then they will come from Mumbai or Delhi and there are some major flights. I put alert on that ticket price if I am like this When the price increases, it means that demand has increased. Like many people do. I estimate with this and say, the amount of flight ticket prices have increased or something is happening. Everyone is coming. We need to get ready for it. It's a dynamic question. That's a good idea. So on the basis of Flight Price Watcher, there will be some correlation? How can we find out? Do they have APIs? What kind of APIs are available for these? Real-time

1:56:30 OK. Pricing APIs for flights and all that. So if we book flights on Google, there will be an API. Everyone has the same API measurement. We can check it regularly. Oh! Crazy. It's so fascinating. In order to run a business, you have to accept normal or abnormal conditions. Now, it is not about acceptance but discussing how to double down. There they are discussing about the data. I'll see whether I have an API or not? No, definitely you do. Definitely you will have one. Nice There is another one. As we are talking, if I connect a realistic example to this, will you drink tea if it's very hot that day? Is there a correlation to heat? Is there a correlation to weather? What are the other correlations that I can find out because this is only one kind of data we're looking at but I don't think this is the only data that should be looked into

1:57:23 It's all, I would look at weather. I would look at sentiment. I would look at why people are coming those factors as well, as much data points that I can bring in. Which I can track. If you sit and do the data without being able to track it once, it will have no impact. You don't see where real-time data comes from but if people are watching they'll be like, data...data..data... What is this? There's no data at all! But that's why they say India is sitting on an opportunity of a lifetime at this point of time. The organized data in the country is less than 5%.

1:58:02 Imagine if we open all our data, which the government has to do. A lot of it has to get organized. There's a lot of data in papers We can really use it to change a lot of things. Hey guys! Hello? One minute... Hi hi hi! Your video is not coming. One second Vamsi, one second. Hello! Can you see me? Yeah, of course. What happened to this city that you are shooting and putting Kida Kola DP? I think it's the old DP that is stuck on to Zoom. How are you? All well, Thailand is burning up. First of all thanks a lot for accepting the shoot.

1:58:02 Imagine if we open all our data, which the government has to do. A lot of it has to get organized. There's a lot of data in papers We can really use it to change a lot of things. Hey guys! Hello? One minute... Hi hi hi! Your video is not coming. One second Vamsi, one second. Hello! Can you see me? Yeah, of course. What happened to this city that you are shooting and putting Kida Kola DP? I think it's the old DP that is stuck on to Zoom. How are you? All well, Thailand is burning up. First of all thanks a lot for accepting the shoot.

1:58:51 We started 20 minutes before the scheduled time. He told me about Tharunana and said he's the first person in our show to talk about two fascinating things. So, in production there is a concept called as line production. How do you reduce or how do you effectively use money of the producers? In that same podcast he spoke about using AI for script writing. I was like, did he really say this? Seriously! Did he really think it's possible to do so? But then... I should tell you this. This podcast will be available in my downloads.

1:59:36 I learn one new insight every time. A different perspective, when talking about content creators... I think no one will speak like that so open. About reviewers or general content creators. But yeah his perspective is different. So he'll be right for this discussion also. Anna over to you! As discussed we can discuss the problem statements I wanted to talk about writing and post production.

2:00:16 What's happening with video generation models, there is a lot of different video generation models. There are common places where you can integrate all these video models also. But the major thing in filmmaking now it's all about control And if control is achieved, then it will disrupt something. Presently we are seeing inconsistencies with location and character. And they are basically becoming better and better.

2:00:58 They're becoming better and better. But having said that, these generative models I don't know if it is even aligned to that perspective. That's a question being targeted by all these video generation models Many people are making a group of agents doing their own custom pipelines. The problem with these custom pipelines is that they operate from the cloud and not local servers

2:01:35 These video generation models are coming out in H.264 containers, which are MPEG compressions. So they're 10-bit rate, 10 MBps or 12 MBps bit rates. If we want a theatrical projection, you have a sequential DNG format. If each image has a frame rate of 23.976 or 24 frames, we run into 13 stops of latitude in the technical language. Now, image generation models are doing that. Image scaling models are there. Can we build a pipeline where we can take these H.264 videos?

2:02:24 Break them down into image sequences, PNGs. And then image upscaling and DNG sequence it will be like batch processing. I think it's going to be really disruptive for the film industry. I don't think we have enough GPUs to pull this off. That is probably the problem. But I think, first of all Tarun huge fan Thanks bro If i have to share a personal story You know Kida Cola has become that movie where after we get drunk if we have to show someone a new Telugu movie that's the movie Right? So it's...I mean my girlfriend has made me watch that movie like 10 times but she'll be very happy to know that we caught up here Chandana gana? Chandana

2:03:12 But I think one is that I am pleasantly surprised bro with the conversation that we just had. I do not know if this is the case across, but I did not expect that you will be talking about content pipelines. You will be talking about agent orchestration to... So you put me in a spot! But sadly what happens now is like with SeedDance 2.0 the one that we saw right? That just came out. The one that is out for us is actually a distilled version of the product The full version of the product is not out yet because one of the issue that ByteDance found is when they tried to convert an image into a video, they heard the voice of that same person. That didn't feed

2:03:59 based on the data set. So it's like, if you put a picture of Chiranjeevi and ask him to generate this, basically we get the voice of Chiranjeevi in a version that came out. That is insane man! Yes as a result they thought this could be a big problem. That is why what was launched wasn't launched around a couple months back now they are slowly launching out which is heavy restricted heavy guardrail restricted version so but But with someone of your access, I think if I was you. I would not use a regular one. I would try to get in touch with ByteDance to get direct access of their no guardrail API

2:04:41 Right? Which I, without any guardrails. Without no restrictions because they will... So for example there are companies like Hagen right? They got integration with SeedDance. They don't allow audio. SeedDance doesn't allow external audio that easily but with Hagen they allow because they follow two-step verification so they don't want people to exploit it you You know what I mean? Hence, they're doing this. But there has to be a way that we should be able to get in touch directly with them. I'll also try find a contact if i can because that is an unlock The second unlock is consistency character consistency and motion control energy It's already being solved. I think it is the biggest problem that all these videos and models are trying to solve. I think we are probably six months away from that

2:05:31 That's crazy. That's what I thought because conversation started in a particular key, there are so many updates that it's a day to day update that we're looking at no longer the three month back conversation is outdated and every day the news is changing rapidly But there are some ways to scale up videos also. I don't know if it will fit in the big screen, which is a theater screens and the level of scalability because even on video resolution level we have best AI models that capped at 720p or 1080p. We're not at 4K yet

2:06:07 Right? Not because they are not capable. It's purely because of GPUs, cost. Yeah the processing that is required here So there are models like LTX which are open source so if we take some budget from producer and put NVIDIA GPUs then we can actually get 4K outputs directly right? So it is possible but tech has to coincide with that point And also access has to coincide to that point. But I think we are six months away and I believe now that you've had this conversation, if someone is picking this up it's going to be probably someone like you. So looking forward

2:06:45 No, so that's what there was one conversation with my ADs and my time to come to image scaling models in just a day. It's not only generally we get carried away by resolution but theatrical projection is basically all about bitrate and data for that particular image and matter For example, you can have a 4K image but it can be as low as 5MB and the color information and gradient information that the image has millions of colors fill up our wallet. That's why we use expensive cameras like Arri and Red because they have lot of color information. When I say why, I mean theatrical projection. We now have Allu Dolby Vision

2:07:31 That low you can literally see millions of colors. I mean, it's a phenomenal experience to watch theatrically so yeah i think the only way for theaters to survive is to provide that much of information for the common audience So that it's a visual experience that you cannot do on your digital screens Fair But I think we have a long way to go before you know We can process that amount of data through AI Yeah The image upscaling models that they thought, you know they are doing it. Yes. Off late we're also seeing on Instagram where they find detailing of old pictures and you know, it's a detailed textures to generate this plus it's scaling that image

2:08:14 The fact that it's upscaling an image, it's just 24 images a second. So it's just like as you said hopefully few months away but my other concern is that all this is being done so that the average consumer can generate films or ideas at the click of a button and I still seem that far fetched because of the processing power it takes. I think its going to cost a bit Rajiv sir clearly made out a very good statement. He said, it's easy to make films like Bahubali but it does not make sense to make small film. That is true. Because the processing power of that movie is billions and billions. So why would you go ahead and make a small film?

2:09:03 I feel small filmmakers are still safe. But there's so many possibilities opening up, and I'm really excited. Interesting! You're asking if there is a possibility or not? But you're saying that it will cost more. No, not at all. That was what I was trying to address. So I was catching up with... Sir, do you have time now? Yeah yeah go ahead I was catching up with Alex, who is the head of Meta AI. Right? I mean Facebook. Facebook's Head of AI. Alex Wong. We were talking about it and it was the same conversation where... Boss like the biggest bottleneck at this point in time is cost of inference. Meaning cost of processing and everything Is the future actually local

2:09:51 So, if the models are converging to very good optimization. so that at lower memory, you can do a lot more work. Large models are becoming smaller and giving you the same kind of output. So distillation process on the... That is getting very good yesterday I showed an example of Jema4 And the same is happening with video as well. First it's in text then audio Text and Audio is almost happening what we were expecting a year back Now, it has to happen with images and videos. Images also have started now the next is video with LTX Studio Japan. LTX is the first glimpse of that where you can actually generate a video out of Mac. It's not great but its footage that looks how it used to look with SORA 1 or SORA the newest model. Is it as good as Seed Dance? No!

2:10:50 But maybe a year later, it will be as good as Seed Dance. So future I think everybody will be having racks that will be part of the cost of production or maybe as you pay money to actors and directors and producers You rent out a rack as well right? As a part of the cost of production for the movie. Interesting. One last question what's happening in the industry with respect to AI adaption Artificial intelligence has always been a part of the film industry. It's called different names. There is also a magic wand feature in Photoshop.

2:09:51 So, if the models are converging to very good optimization. so that at lower memory, you can do a lot more work. Large models are becoming smaller and giving you the same kind of output. So distillation process on the... That is getting very good yesterday I showed an example of Jema4 And the same is happening with video as well. First it's in text then audio Text and Audio is almost happening what we were expecting a year back Now, it has to happen with images and videos. Images also have started now the next is video with LTX Studio Japan. LTX is the first glimpse of that where you can actually generate a video out of Mac. It's not great but its footage that looks how it used to look with SORA 1 or SORA the newest model. Is it as good as Seed Dance? No!

2:10:50 But maybe a year later, it will be as good as Seed Dance. So future I think everybody will be having racks that will be part of the cost of production or maybe as you pay money to actors and directors and producers You rent out a rack as well right? As a part of the cost of production for the movie. Interesting. One last question what's happening in the industry with respect to AI adaption Artificial intelligence has always been a part of the film industry. It's called different names. There is also a magic wand feature in Photoshop.

2:11:36 Business and technology always converge and create compartmentalization. Smart was trended first, then green. We've always had technology to assist us. And filmmaking has always been innovative and it's always been the edge of that. Because we're constantly working with VFX, what we deal with is video. It's always been constantly evolving. From post production DI or... Obviously, we have a production intensive

2:12:19 Even in the recent movies I have done, even in promotions we had Trump. Oh yeah. And I was pleasantly surprised, that is happening and in VFX also a lot of work has reduced. A rotoscope artist will have a big credit roll with many names and that is going to reduce significantly and a lot of work is going to be simplified for sure. Access

2:12:57 I don't think it's going to replace filmmaking altogether because definitely there are going to be a lot of constraints but it has to work in tandem. Definitely, its going to work in tandem and basic work like lower rung jobs If you cut a paper or film, those jobs are in for a toss. It's part of evolution. I don't think we should do stupid things like protesting on the streets It's a part of evolution. If you want the wheel to go, then the square has to become... Cut off its edges. Crazy! When is our next film? It will be soon. You are talking about E&E 2 right? Yes.

2:13:55 I'm in Thailand right now. It's scorching heat. I wish, sometimes when Rekki does it, she takes a few photos and puts them up on the air. Do you remember that first question from our podcast? When will ENU2 be released? I'm finally getting to finish all the apples. We'll sit there for another episode. Thank you and thank you for this. Really looking forward to the future, exciting and I'm glad that I could speak to you Vaibhav, Dilip and all of you guys. Thank you for having me!

2:14:47 Bye, Bamsi. See you around. Insane! I'm just imagining... This was impressive. Right? For the viewers, imagine what content we made yesterday and today. Just imagine if someone who doesn't like or doesn't know about AI is not adapting to AI if he's watching this content. We're increasing so many level of thought processes. In fact, that's how much I learned. I asked for 30% of my notes in the document that I wrote yesterday. But still we were so... How much content did you make? Maybe it will be available in the next episode. It's not a topic, it is an outdate. It'll come later on. It's crazy to see such problems in each industry. What a different exposure! True Actually, I heard something from Ben Affleck. You should tell us about him

2:15:38 I was like, they are saying scorching heat. I'll also give you a point on why he liked it so much. I heard two Hollywood actors who have very good AI consciousness and awareness. In fact Ben Affleck has an AI company. He is Netflix's producer now So Ben Affleck, for me, I always explain to my learners and normal people what AI is. First, I will explain how Ben Affleck explained it. What he says is that in AI everything is mean. Everything is training. In any subject you train in AI there will be 10% best material, 10% worst material 80% average. Okay? If you directly ask an AI to do this, the average comes out

2:16:29 So, anything about this query? Actually, I thought it was upscaling. It's not upscaling! A very different problem. So when he said that, when he spoke about 24 frames per second if we scale it up and put together to make a video technically he is absolutely bang on. Upscaling does the same thing too. When there are tools like Topaz, as he mentioned right? The upscaling happens in the same way. Topaz is used by our editors also So, it's for video editing and now it has become AI first majorly. We can probably show towards the end of that right? Upscaling basically so it also does in the same way but what he did is very interesting He showed... He told people

2:17:27 Even if you know your art very well, you can apply AI there. Learn AI and do it because the moment he said that upscaling and how it is 24 frames per second. I can think of we can use Comfy UI We can use Comfy UI, which is an image workflow. You break down a video... Basically he means... Every second is divided into 24... 24 FPS means 24 frames per second. Yes, it's 24 images. If you put together these 24 images, then we get 1-second videos.

2:18:05 It is difficult to upscale a video. The reason for that is the GPU's. I got it, okay! So each and every image has to be upscaled because it is easy to upscale an image. Once you have upscaled the image, all of the images put together in one software will be high quality videos. That is what Topaz Labs does. But the interesting thing is Topaz lab is probably not catering to his requirement of theoretical. I can imagine that you can build a comfy UI workflow for this. Comfy UI workflow is exactly this, do you want to explain? So comfy UI- Can we also see? It's an open source thing so you have to look for comfy UI. Comfy UI is open source but if it is locally installed in our machine then the whole system will burn out.

2:18:55 So, we need cloud GPU servers to load and run. So, with that... It's a long story. Six months back! Six months ago? Yes, six months ago. Eight or nine months back, someone had a model in their hands. We wanted to put Pepsi bottles into Coke bottles in their hands for example We have a Pepsi photo, we need to replace it with Coke. AI is not that powerful at the time to do that but you can use Comfey UI. So while doing all this... That's when I was there! Yes, then only. Then I taught some people about this too. We thought we had achieved something and changed everything.

2:19:45 Two months back, two weeks after that, ChatGPT image generation came. We put Coke and IMAGE in it and changed Pepsi to Coke and Coke to Pepsi. We did all this on GPUs and taught them a lot of things. While teaching the workflow, everyone was crying like they didn't understand anything. This is a real problem. I can hear the pain in his ears. After saying it so strongly, he's done with it. I mean, some people do it that way. So now, coming to this problem is a good option because... Till you get the AI models to be able to do that if its really important problem for you then yes yeah. Then what we have to do is like Vibha said, put GPUs in the whole room and solve the GPU budget. Those days will come, definitely. Guaranteed!

2:19:45 Two months back, two weeks after that, ChatGPT image generation came. We put Coke and IMAGE in it and changed Pepsi to Coke and Coke to Pepsi. We did all this on GPUs and taught them a lot of things. While teaching the workflow, everyone was crying like they didn't understand anything. This is a real problem. I can hear the pain in his ears. After saying it so strongly, he's done with it. I mean, some people do it that way. So now, coming to this problem is a good option because... Till you get the AI models to be able to do that if its really important problem for you then yes yeah. Then what we have to do is like Vibha said, put GPUs in the whole room and solve the GPU budget. Those days will come, definitely. Guaranteed!

2:20:38 It's like a car going on the mandatory. It's like a GPU ban. VFX is becoming a big problem now, because they shoot on green and blue screens. The director has to render all their VFX, but it looks so bad that everything gets doubled up in no time! So, I'm speculating on how they are rendering it. Is it correct or not? In spot render, is the scene and hero that I thought of interacting in the right place or not? To do a spot render, you need to install GPU-WAN and render VFX with GPU-WAN. Now we can go for theatrical production or... Proper editing. Yes, I am speculating.

2:21:19 It could be the future of film production. We'll start again as soon as you say it. When did we last do a podcast? 6 months ago, right? I don't know if this will happen in 6 or 9 months. Vamsi has just told me that he was asked about the role of a character in any film or musical.

2:21:19 It could be the future of film production. We'll start again as soon as you say it. When did we last do a podcast? 6 months ago, right? I don't know if this will happen in 6 or 9 months. Vamsi has just told me that he was asked about the role of a character in any film or musical.

2:22:05 As far as I know, in AI, I can say that it is the department which protects our literature. In relation to literature, AI does not create or solve any problem. At present, we cannot predict what will happen in future. At present, it doesn't solve our problems and neither does it create problems for us. Why am I saying this? Normally, the artist's soul is in his hands and the singer's soul is in his throat. But Kavula's soul is in his heart. It is in the mind. The one who works with the mind is a poet. What is related to the mind? Kavula's work was related to the mind

2:22:51 The brain can do nothing but the mind cannot do anything. Because the brain is a machine, it has no function to make the mind work. Till now we are in digital format like poems, stories, novels and different kinds of dictionaries etc., all these are in digital format. We have only that from online I am getting some help, but none of them fully understood us. Recently in a movie called Sari, RGV's film, there are two songs and two songs. The same kind of literature. He wrote the literature again and wrote the story.

2:23:30 If I write it, and give it to AI, the AI will send me two types of notes. The two types of notes are used in movies. Two songs, two notes, two songs. The script is one. So if we give a script, music will be sent to us. Even for fun... There's something called Sonona or something right? In that, I would sing a song from my country and ask for fun, it would do all kinds of things. Not 10 minutes, but 2-3 different things within 1 minute itself If we give literature, it will make music. But is it enough to make literature? You should know the language, use it and understand it. You need to have experiences, memories, thoughts, etc.

2:24:28 So, what is the problem with this? So we have to see if all these things are possible in the present. Because it can be possible in future but for now there's no solution or co-operation available to us at any cost. No question! What do you think? This is a bit more Telugu for you right? A little understood no? High level So now, if we need to discuss about this. What do you feel? Whether it's not Telugu or English, even if you put aside vernacular and original language, in general, for lyricists, is there a situation that can accurately help those rules? I think so! Again, I'm not a lyricist. I don't know all that much but logically speaking... Logically speaking

2:24:28 So, what is the problem with this? So we have to see if all these things are possible in the present. Because it can be possible in future but for now there's no solution or co-operation available to us at any cost. No question! What do you think? This is a bit more Telugu for you right? A little understood no? High level So now, if we need to discuss about this. What do you feel? Whether it's not Telugu or English, even if you put aside vernacular and original language, in general, for lyricists, is there a situation that can accurately help those rules? I think so! Again, I'm not a lyricist. I don't know all that much but logically speaking... Logically speaking

2:25:25 I feel all the creative people have roadblocks. They think something is not working out, biggest... In movies that I watch, even if it's a Beatles movie like you know, all these large Bohemian Rhapsody every big artist goes through a block mind-block where they're not able to come up with new things. One day they get stuck and are unable to move beyond AI can help you in that area. For example, he took the example of Suno. There is a song and how to represent it? Maybe you used AI for your movie. AI will become an idea engine for you

2:26:11 If you send me lyrics and ask me to sing a song with 100 variations, because AI understands music across the world not just of Telugu. You get a lot of inspiration. We don't pick up that information directly from someone else. When we listen to new inspiration, we come up with another idea. So it compounds your innovation ability That is how I use AI as well today. Like, if you want to try a script on Instagram... Now, I don't consider myself to be creative like sir is but even at that point of time

2:26:51 Many people think that Webhub just gives a topic, writes an AI script and shoots the script. Or they generate it with AI. That's not true. I talk with AI as I give ideas to AI, AI gives me more ideas It is unlimited idea generation machine for me. It enhances my perspectives Because I don't have someone who understands what I understand right next to me all the time. It acts like that The second thing is, everybody has a style. Everybody works for a reason. Why does it work to know that? Like what is my flavor? What is my style? Why do people like me? It's a very hard thing I believe. So again, I can also imagine if I'm a lyricist or if I write songs... ...I would feed all the songs

2:27:49 that have already made it. I will say these were the hits, these are the ones that have gone viral on Instagram Reels. These bits... This part is viral only. I've written all of this already but you didn't accept or put them in. So this is how I think and on top of that I will do everything so that AI also understands my style. He said right? It comes from the heart. I believe creative is all inspiration Nothing is original. Everything is synthetic. And the idea that comes to us is coming on the basis of something we have seen, something I've read at some point in time and it's all inspiration. So finally too many inclusions, too many experiments

2:28:40 12 hours later. One eternity later. It was definitely 8 hours late, too much of explosion. My brain exploded inside. You know what I had a lot of fun That's what I was about to ask. I had a lot of fun actually, uh i think the part that I really enjoyed is the questions um we should do this more often yeah definitely This will be an IP. So yeah and we should also build. If you are able to solve it, it'll be amazing

2:29:24 Now, we could ask on the starting level. We'll have to name it. What was that? The last five years we've done... Five hours we have done. Breaking heads with AI! Baking heads on AI. But yeah, it was amazing as a different point of views and in the end business validation is very important. If you want to see a problem as a problem or not? If it's such a big problem then solution has worth. Everything means different points of view. This is no just about AI, AI, AI. It's also about orchestration. It's strategy understanding connecting dots like

2:30:03 I feel proud. On that note, we have business validation and idea validation. I think a lot of engineers love building stuff and in the course of building they forget whether it can actually become commercialized product or not. Because they love their products, they really think like its a baby. They don't mistake it for something else. So, I think such kind of things would help them. Okay! Is there really a problem in the market? Should we solve it or not? That would really help. Yeah, I'll have to do this more often. Not just Vaibhav, but I even have Dilip. So... I will rope you in any time. Virtually at least. Whether I come here or you come here. Ready ready. But this was good because I like to think and when questions came up, I was thinking

2:30:53 I was thinking, I was like, yeah. Is it possible? Is that possible? Like the Sky Scanner API figuring out... is that happening? WhatsApp is doing this and that. It's not even Anukunt. WhatsApp is doing this and that. I've seen browsers but it's not scalable you know what I mean It's not very... And then I remembered an open clause. If it could reply my messages, that means it could read my messages That means I can use that to summarize my message So I learned something new by thinking about it which is what I like I'll have to go there for that crazy experiment. He's going to shut down in every reply this time! I'm scared, man! I'm getting a WhatsApp message again... Do you know what happened recently? There was a session. There were 40 thousand people on the call. Our sessions on Zoom are like that. It happens all over the world. Yeah. Forty thousand people on the call

2:31:56 Sunday evening, two days ago I was running an experiment in my computer. I did it and then I travelled back to India. I forgot about it. I turned on the computer and thought of starting the session but suddenly the computer restarted. You are telling that? That's what you said this morning. Suddenly after the computer restarted, as soon as it restarted, the screen froze. I ran a Mac studio I don't see problems like this. Then only Dilip called me and said, maybe... I might not be able to take my sessions because my computer is not in my control. Because initially via Shell I gave it root access. So literally you can switch it on or off as per your wish. But behind that some program is teaching you.

2:32:39 My computer restarted. I told everyone after joining the session, listen if my computer restarts If my computer stops if something comes up on my screen which should not Please bear with me. Suddenly a browser came and started scrolling. So people were like, what is this? Did the devil enter your computer? But that was when he was on standby. I am on standby now. International audience are also very strict at that time. So one person was saying somebody hacked your computer. What happened?

2:33:15 They told them, but some people joined a little late. That's what I was thinking about. Full capsule, somebody hacked my computer. Multiple times he wanted us to notice the message that I hacked your computer and nobody said anything else. If we don't talk now, we'll be here for another 2-3 hours. We haven't eaten or roamed around yet. We might even miss our flight tomorrow morning. Done bro! Thanks alot. Thank you It was a crazy experience. Let's clap for ourselves. Claps for all you people, for being so patient. I forgot how many cameras are here. There is a press feed video. Same press feed video. We should give this view. Yeah let's wrap. See you. Bye bye