Cloudflare's AI Crawler Defaults and How Agents Choose Brands

0:08Second. Yeah. All right. Hello everyone. Great to be here again. Here we are at Marketing to Agents podcast. This is episode number 12. And as always, I have Lalit with me to share his insights on last week's all of the updates around agentic commerce agents and all the progress in this space in the last seven days. Welcome Lal.

0:36Thanks Ashwani. Good to be here.

0:39Great to have you here Lith. And today we have another guest joining us. we have Vine as well with us and Vine is the CTO of Airmeet and has been instrumental in building the whole agentic commerce and video layer at Airmeet comes with a lot of experience in this space. Great to have you here V.

0:58Hey thanks Ashin. Happy to be here looking forward.

1:03Fantastic. All right so with that we have a full docket of great updates from this week. let's dive right in. So to start with first let's talk about what's happening on the what's happening more on the content side. So we just heard about we just saw what Cloudflare did and now for example there are there is a there is a limitation on what kind of agentic traffic can come to our site. So publishers are now having more control. similarly there were updates from common browser as well. So Lo let me start with you and when I feel free to jump in but would love to hear what do you think about these new measures coming in especially with we've been hearing again and again that publishers are not happy with how AI overviews how AI how LM have been treating their content and not giving them any references or any traffic. So with this coming in with some of this coming in now what do you think how does that change the landscape

2:11Sure yeah I think u so so cloudflare has been very let's say transparent about what they are planning to do they have in fact created enough of an infrastructure for letting publishers especially who monetize their traffic through advertisement u monetize their content in the new age right so they have been working on the side for let's say analyzing the traffic. So they have done a good job in terms of bucketing the AI traffic that any website is getting into four buckets which is your indexing, search agent and training right so so and with this new update what they are what they've done is they first of all enforced what they have what they were trying to do for a very long time. September 15th was the deadline. So with with that what has come to effect is that all pages which which were monetizing through ad traffic where let's say there was an AdSense or a ad widget on the page over there I think except for search indexing all the four traffic buckets have been blocked by default that is the default setting right and publishers have to go and explicitly allow and they

3:29Can when they are allowing they can also choose to have let's say a payment mechanism right so if they're allowing they can they can choose to let's say allow a certain set of AI agents where they probably have let's say partnerships or if they want to maybe have a payment for the longtail of AI agents so they can also enable that so that is what has happened so in my in my opinion I think it is a it is a much needed change because I think there are a lot of especially longtail businesses which are working very hard to bring in real insights, real on ground content experiences in digital form to the internet and for historically you know the the ad revenue business model has been the the driving force of or rather lifeblood of of let's say these publishers as you know AI searches turn into zero click searches where the traffic is dropping they needed a new way to monetize so we needed a new lifeblood in some sense So I think it's a it's a decent change that has come about. So I'm looking forward to this becoming mainstream and publishers again reviving their revenue streams.

4:42Perfect. So couple of questions on that. So first of all, so that means if I'm a publisher and I am not aware about what Cloudflare is doing or how they were the intermediaries on for my website getting the traffic, I would immediately see a drop in my earnings. Right? Right. If I was earning through ads that would have that would happen right from September 15th when Cloudfare launched this or switched this on, right?

5:09Yeah. So, Cloudfare has been communicating this very regularly. So, but still there is a fair chance that someone is not aware and if they wake up and see drop in analytics, a drop in let's say and if they're also monitoring AU on the side using airpulse.ai like products, they would see a drop in citations. they would see a drop in let's say their share of voice and that would sort of you know lead them to thinking like what exactly has happened and I'm pretty sure again you know the chatter on social media would still bring bring it bring that to their notice that they have to go and check on cloudflare but see there are two two type of pe you know publishers publishers who are not really monetizing with ad they really want marketing the content to be out their content is marketing for them and like airmeat our company you know it is falling in that bucket and we you know we were very like let's say quick in terms of going and enabling all AI traffic all AI you know training etc should be to be to be coming to us because we welcome because we we want to get discovered on AI agents we want more and more customers to find us when they are looking for high intent you know where when they are doing high intent queries on AI engines right So

6:28Quite like that I think media sites where revenue is through advertisements I think they would have also gone ahead and enabled the payment mechanism collection mechanism.

6:40Interesting. All right before yeah before I ask something when I would love to know your thoughts from from a you know from a technical point of view how are you seeing this? Is this like are other players in the market also going to pick this up? because I'm assuming top player is only certain percentage of the market only. So how does that how does that propagate across the industry then for every publisher?

7:05Yeah, in general I look forward for updates from Cloudflare that is a great signal on how industry is like moving towards with respect to content. so it is a great representation of the traffic even though Cloudflare is like we say 20% of the internet served by it. so with respect to technical things right they la made the changes live from September 15th then I'm still consuming it like as in there are settings which are applicable for free plan of cloudflare and in one line in one of the article there is a line mentioning that if you are on a paid plan and you made some ch settings already those settings will be retained that there is no effect on it so who will be affected who needs to update their settings in Cloudflare. I'm still going through.

8:01Perfect. Thanks, Van. so one follow-up question on this. Now there are we talked about there are three types of agents that are coming on a publisher's website. Now Cloudfare has stopped two of those from coming from the website. what hap so when I switch on the monetization, do I monetize all three of them or the search is still never monetized? It's only the training agent and the crawler agent which gets monetized. Is are they clear on that yet or it's still to be like to be determined and the data let the data flow in then we'll really know it or they've already explicitly suggested or mentioned this is how the mechanics of monetization will work. So you can so as a web admin you can turn off all the four buckets right but by default nobody wants indexing to be off on their website because u if search engines do not index you then search APIs would not be you know listing you and AI agents anyway will not never be able to discover you right there whatever whatever queries are being done on users behalf on let's say inference time or when let's say user is asking certain agent to do research

9:20Or fetch some information for them right that exact visit was earlier a user visit which was a monetizable event for them for the publisher and that is what the discretion is more applicable on right the training bucket is actually particularly important because you're especially you know publishers who are you know generating really high quality data just take for example there is there are there is let's say GitHub for once right GitHub has is a repository of open-source code repos right all of that code has gone into training right and there might be let's say niche GitHub like companies which are specializing on let's say you know scientific research or cutting edge like say you know life science research now the training of that is you know once once that data is information is leaked into the training sphere of things that is like a permanent loss right so that bucket is I think particularly very important publishers who are who are sitting on that data who want to monetize this data want to have let's say sit across the table with these labs and have you know let's say licensing discussions or they want to let let's say dynamic realtime determination of pricing so cloudfare is also building that infrastructure

10:38Where you can explicitly you know monetize from that one visit which is at which is being done either at training time or inference time. U the the unit to monetize also is evolving. Earlier it was page access where you would basically monetize an agent or a AI system accessing your page content but that sort of at least both the parties are like iterating on it but the and you know what we what we're reading on the internet is that the consensus is sort of converging on let's say valuable citation right if let's say a particular content or a page made it to the citation in an answer that is when I think it's worth paying for in the if if the agent is coming at inference time for training I think it is a very blanket discussion u you know there's there's not much of a discussion there but for inference time usage of content only when it makes to the citation I think there is valuable there is value to be charged

11:50Very interesting le I think breakdown the you broke it down. I think it should be very interesting for any publisher to understand and know when they should switch on and how they should monetize this traffic and the new labor that they've got. Okay, perfect. moving forward, our next item on the docket is u around agentic commerce. So we've heard with huge applaud Meta just launched Muse like a personal assistant to help you buy right from WhatsApp itself. we've heard news coming out of Shopify that whatever is the AI recommended traffic that actually has increased the proportion of search and sales which are happening. of course overall agentic commerce and the queries which are directly now coming propensity of theirs converting into a sale has increased significantly. So we are seeing a lot of lot of data coming in suggesting that search is slightly reducing in value and either agentic commerce or AI recommendations are significantly increasing the

13:07Commerce e-commerce category. So as this becomes more and more important I'm sure it will of course break down the search monopoly. It will also break a lot of monetization models which are around featured listings or you know showcasing yourself on the top on the SER. I know we've discussed this before around how Amazon's revenue might get impacted which is their ad revenues but in general is now we are seeing more personal assistant launch coming in. how do you see this space now changing especially with e-commerce as a focus and then again we'll look at it from a brand's lens as well but let's overall understand what's happening and then how do you see this space shifting around VA why don't we start with you this time what are your thoughts around this

13:56Yeah I think it is very evident that it is increasingly you know agents will be deciding what to buy what to do Like if it is like some activity like you know and these brands or businesses are like you know very blind to what agents are recommending and personally like last 3 days I depended on instincts recommendations for many of my daily activities and I am going with those at a 90% hit rate remaining 10 I am deciding few things just an experiment to use instinct Right. So, business is definitely you know a very valid reason to get them worried right like whether they are being recommended or not. and so what are the solutions? you know we are trying to solve the same with the AirPulse as well which we will going to talk about like what is the direction we are taking. but yeah right now it is a very you know confusing state for merchants and businesses

15:08Very very interesting yeah and yeah that's with the instinct I think we'd love to hear something more around like what kind of use cases you've been you've been able to go through and like just an example maybe because yeah instinct is one of the fastest picking up system especially a personal assistant and built outside of any of the frontier labs or or Mac 7. So great product but give us an example of some what is it what kind of recommendation are you relying on instinct for example. So yeah, I have been working in a remote setup for last 6 days and most of my use cases are around finding a co-working space with a or some cafe with a nice internet and you know proper you know conducive working environment right and just to experiment that at the same time I got a access for instinct so just wanted to try how much I should or I can depend on a personal agent as an experiment personal experiment I have just started following it pretty much up to the point whatever constraints I have given to the agent right so like distance and what are my priorities like you know stable internet is the most critical thing for me right now so pretty much on point and whatever it did

16:26Not recommend I tried to you know include in the list but then again I came to know like you know why it was not recommended like

16:36Because somebody wrote a you that you know weekends they don't allow laptops for that cafe. So I went there and then realized and I actually praised you know agent by responding in the WhatsApp message like you know you did a good job kind of.

16:54So so V that means in this example you never went on the web. This was all through your WhatsApp just going back and forth and all of your research was done. You never had to go and go open a web page, run a search, nothing. Never.

17:11Yes. that is the experiment I am going like how much I can depend. So yeah, it's going well so far.

17:18Fantastic. Thanks for sharing that great anecdote. Lalit.

17:25Yeah. So Ashwani last episode also we discussed that you know the the information access or the way we way way humans are engaging with AI right we are transitioning very fast from u our you know like from answers to action to outcomes right so we are now mentally ready to delegate our work right to agents and that is going to change the search like significantly like AI search is not not going to be the same as what we have been sort of observing in the last let's say you know 10 10 to 12 months so there are two things which are driving you know changing it first is this user expectation of complete reliance right so which means the accountability on the agent is lot more right so the the builders of agent the the way the agents have been fine-tuned or system prompted or the kind of tools that they've been given access to they have since there is going to be very high reliance that as we have seen in the case of Vine right so the accountability on let's say giving a a nonfailure experience right now the the what is expected of them is not to go overboard and do a extremely great job

18:44Not to impress but to sort of avoid the failure scenarios a lot more right so agents will turn turn out to be a lot more conservative right And then they would value a trust signals a lot more than any other thing. Right? So anything which sort of signals quality reliability of a vendor or a service provider will have disproportionate high value in in terms of making it to the recommendation. That's one. The second thing is the removing of the synchronous time boundness. Right? For example, what let me elaborate what I'm saying. So when you are on a perplexity conversation you as a user are engaging with perplexity chatbot live right and in that user experience perplexity is sort of instructed to get back to you without too much delay right so somewhere time optimization is is part of the the whole process right so the there's always a trade-off between how fast you can get get an answer to the user and how good the answer is, right? So that trade-off goes away because these agents are asynchronous fundamentally, right? So you so in case of Vinn or in my own personal usage of instinct, not yet Muse

20:04But Grogbot, I would give that instruction in a chat message or a voice note and then I would, you know, let's say expect it to respond asynchronously. 5 to 10 minutes and even 30 minutes of delay is is fine with me, right? But that's not the case when I'm actually on perplexity right so the same cafe query if winner would have given on perplexity perplexity would have had less time budget less compute budget in finding the best answer right so what these two things have an implication on the AI discovery is that first of all you have to focus a lot more on the quality signals and go as a business you have to go an extra mile to get as many quality signals as possible right second is you have to be lot more explicit about what your offerings are right for example if one of those cafes did not even have free Wi-Fi you know as a very clear benefit right or someone if let's say vine had already mentioned that he needs a high-speed Wi-Fi right now someone has a cafe has Wi-Fi and they probably had already invested into a high-speed connection and they don't bring that up onto their website or their listing pages is then that's a shortcoming that you know that would sort of you know inhibit them or rather

21:22It would become like a fiction point for them to get make it to the recommendation. So an extra effort has to be then put in terms of bringing out all the let's say value proposition right that the business has right what are what are the attributes which make that business the right choice for their right customer right those two things are are going to be significantly more important because as I said agents now do not have a time constraint do not have a compute constraint

21:55Very interesting ing points. I have a few follow-ups on or just three observations from what you just noted. So number one from what I realized is for this is and this is important for all the publishers, all the brands out there that means negative feedback becomes even more critical. So it's no longer an average rating now. That means negative feedback like we had NPS scores. We had of course we had first average scores then we had NPS but now looks like your deterrent or the negative score will be their importance gets amplified even more. So just like I think in V's example just one bad feedback or or some bad reviews could completely remove you from a recommended being in the recommended answers. So I think just for businesses I think that's one of the key points key takeaways as well. negative views are going to become even more critical now. And then couple of things interesting came out. So number one is now so there was a data published I think a few months back by chad GPT that on an average the users are qu the user queries are about three queries three to five queries per day on chat GPT or it was some some kind of that matter but three was a number between three to five but now with this becoming a personal assistant coming on WhatsApp coming on

23:13Voice I think the number of queries are just going to explode right so you're going to just use this throughout your anything that you need to do, any search that you need to perform, you actually are doing all of that here then. So what that also what that also means is that eventually this whole then snowballs into much larger number of AI queries to be handled, hence the more need for more AI compute, hence more data centers, hence more chips. So I think this we are so the buildout that we are seeing from the hyperscalers from Nvidia all of that that means they're already seeing this data and that means they're already believing that this will this will like snowball into maybe 100x from where we are today hence so so that I think the form factors are changing earlier it was you have to go somewhere and search now it's in your WhatsApp in your browser so I think that is a great insight and that's how the usage is increasing as well.

24:08Yeah. And then of course love sorry go ahead.

24:11Yeah and I think see the the drastic reduction of friction with which let's say users can interact with AI assistant would just open the gates for more and more things that they would get help with right and that would again let's say result into more and more let's context being transferred to the agents. So agents would know that you let's say you have a family you how many kids you have which schools do do they go to what is your preferred travel route what allergies and then you know what what let's say favorite places for destin destination typical budget favorite airlines all of it will be sort of you know will be then absorbed by the agent as a context and your understanding and then just imagine how precise would be the choice. right now if it knows you to that level right it can very precisely recommend you that okay go with this option this is the best right so I think the second point that I mentioned was where I think the the the catalog the quality of the catalog of about like for example being very explicit about the benefits that you offer right who is it that you are serving and being very let's say verbose in terms of their in your

25:30Features benefits to that customer becomes even more important because someone is coming to to your website with a lot of context about who they are representing and they want to really make a really really good choice. So if you have the quality signals and a good quality catalog page, I think these two combination is a sure shot recipe for getting into the recommendations.

25:55Perfect. Great insight. Yeah, I think for a brand that becomes extremely important. Yeah, just being very very clear on your target persona and then ensuring that you're comprehensively answering their needs or covering everything around that need because someone coming with huge amount of context will choose you only when they see that when they see that preparedness in the brand.

26:16Absolutely. And in fact, you know, the other day in another discussion that I was having, I think the designations on testimonials on a website, right, that also is going to matter a lot more, right? For example, if a agent is representing a director of marketing at a let's say enterprise company, right, then if the testimonials are from startup founders, right, on that marketing website, it would probably, you know, sort of dep prioritize that and then maybe prefer a vendor which is recommended by similar profiles. So every pixel, every word is going to matter. So we have to be more precise about it, more thoughtful about what goes onto the website and what doesn't. And I 100% agree with your point on that the negative reviews, anything which is not aligning with you you being projected as the best choice is going to matter a lot more.

27:11Perfect length. And there was another related note on that. So I think if that happens there will also be there will also be implications on number is so for example how much traffic comes to your website because now that the traffic is not coming so I'm I'm also now starting the question will website remain a main like a key source eventually especially in the in the businesses where it's not PLG or where the users are not expected to come and those are more information based sites. does the form factor even remain a website? But that's I think that's a completely different discussion. But because the form factor there also could change like you could just if you're if you why don't you build something which is purely for the agents to consume. They get the best material best content in the fastest manner and just that recommendation flows in through the users to the users and then gets there. But yeah, this could be a a complete different set of

28:19I think just on that just on that I think I think you touched upon a one of the most important points I think you know the website as a form factor right because the consumer is changing and so two things in the ecosystem which will become very huge going down the line one is something called webmc that we have discussed earlier where any visiting agent can basically very efficiently in a very token efficient manner can access more information from your website can take actions on your website. So the the user actions that were exposed through buttons and clicks just imagine those also being sort of declared in the markup and any incoming agent can just you know call it like add to cart or you know find more information check availability of this co-working on this date right u and you know apply a discount code if whether check whether this discount code is still applicable or not so all of those things the webmc has to address the second is going to be there's something called as ARD agentic resource discovery right now this is something that is actually very nent which I think a consortium of large cloud providers are

29:38Championing right now so what it does is basically allows a business to expose a lot of the agentic services in their own catalog so for example there if imagine there's a travel agent longtail travel agent in in a small city in let's say Spain right and they have a catalog of let's say experiences and they expose that catalog in the agent you know agent resource discovery so they can host their catalog on their own website in a JSON format where it can be discovered by any agent and agent is reaching to that catalog via a mechanism of federated search so there will be let's say a lot of these registries which would keep cataloging basically indexing these various services and cataloges and would expose a search for any agent to look for a service. So the problem today with AI agents is that when you are building an agent the tools that goes along with the agent they are somewhere shipped by the developer right now that limits the the what the what the agent can do right so ARD essentially decouples the agent runtime with the tools tools could be discovered in real time depending upon the task depending upon the user input right depending upon how the user is

30:56Steering the agent tools can be discovered real time that is the paradigm which ARD is going to champion DNN right very similar to DNS like whenever you type let's say Airmeet.com from anywhere on the globe your browser is able to find Airmeet.com website right so there is a federated DNS infrastructure that is there and ARD is going to be an infrastructure just like that so these two things are going to be very important how your website or catalog of your services getting discovered through ARD and the federated ecosystem and then once the website is discovered the agent when it comes to your website do you have enough content content and tools exposed via things like webmc or natural web NL web there's also a Microsoft product called ML web NL web sorry natural language web these are the things which on the website will be the 2.0 version of the website and would be really efficient for the agent to transact or get get more information.

31:57Fantastic. Yep. I think u a lot of these new things will I think enhance the experience and u change the way we now start thinking about presenting information about our brands or our services. Perfect. moving into our last item on the docket usually which is around our search discoverability on AI and again like every week we're getting more and more data around how more how there is a widening gap between what AI is recommending and what AI is referring to your site eventually. So we are seeing more and more brands coming out or more and more studies coming out which are clearly showing a distinction between if you are getting mentioned but if you're not getting mentioned that's a different matter. Citations and mentions are completely different. getting more citations is great but it doesn't mean that it'll translate to more traffic for you. And again getting more mentions means your brand queries might increase but may not directly lead to traffic coming to your pages as well. So there is a lot of still investment and lot of research coming out of the whole AI visibility space. So would love

33:16Your thoughts on that. Vanel anything that you are observing or anything latest that you've noticed in the space and which can help our listeners for example especially brands especially publishers what can they start thinking about and figuring things out better as they want to get more visibility into this

33:38Yes I I'll jump in I'll maybe just maybe kickstart some thoughts here so I think our docket clearly talks about data which is interesting for e-commerce players where a brand getting cited versus a brand getting mentioned right so a a brand getting mentioned is basically you know getting twice the number of visits and versus let's say brand getting cited right so businesses should optimize for not getting not just getting cited but also getting mentioned right so you should be part of the recommendation that the agent or the answer engine gives to your to their to the to the user who is visiting So that's one very interesting thing and u so that would be from my side and vine what to do to you. Yeah, the same lines like I agreed that citation not equal to mentions, right?

34:30We also have our own example, right? where before we started AirPulse like when I generally asked perplexity about one prompt to check I was surprised that Airmeet is one of the citations some listical was you know crawled and cited but it mentioned some competition five of them but a not mention was not mentioned in the final response when I probed further it said like you know some review GT review mentioned pricing is little complex somebody mentioned it. So that's why you choose not to give in the top five or top three. So yeah this is a very real challenge like you know knowing whether you are mentioned or not then how many citations all these together form your you know set a yeah a yeah a yeah a yeah a yeah a yeah a yeah a yeah a yeah a yeah set

35:25Absolutely and I think maybe you can talk about the dev tools ecosystem v where I think the penetration of coding agents is anyway like very high and people developers are discovering tools services right when they are you know in the coding session. So what is working out there? Yes. again one thing one clear signal is that there are tools that are recommended by cloud code or codeex right which one of the projects basically it went ahead with one vector DB where I did not even hear the name before. there are like quadrant and there are some other tools which I might have chosen but when I just let clot code go it made the decisions then I reviewed okay then made the decision to go along with clot code suggestion which I would not never use the database so very important for them like you know getting discovered by this cloud coder codeex for particular use cases writing content about which use cases are served by particular you know solutions. this is happening across tech stack

36:43Not just some database not just some monitoring tool. so very interesting like to experiment like you know to come to know about these tools which are existing for a while but just recommended and just came to know about this now.

37:03Yeah. And u u so one of the things which is which is a quality signal when it comes to dev tools is how many opensource repos on GitHub actually use that dev tool right and and that's where I think dev tool companies are investing heavily in hackathons. So you basically go and sponsor hackathons. Much of those projects will be let's say student projects or you know just done for the sake of winning a prize. But what you end up getting is a large number of repos which are using your dev tool right and like this is what is working and because coding agents obviously crawl the code bases and then there's no better proof than an actual project using a dev tool right so I think in e-commerce also there will be in not just e-commerce let's say e-commerce particularly but in other categories also there will be new modes of signaling quality right where we are already seeing that let's say in podcast genuine mentions not sponsored but let's say genuine mentions of products and services is actually u you know is is really very important as humans we might have stopped consuming a lot of podcasts because it's exploded like like a billion minutes in an hour is what is getting uploaded on

38:21YouTube so we definitely cannot consume but AI has the bandwidth to consume it right and a genuine mention there matters a lot more right I social signals in SAS companies how strong is their community right which is their YouTube followers Twitter followers do they have a slack community or a discord community right and the bot might attempt to let's say hit their discord URL and see how many members are there right those are lot of signals those are those sort of account to some set of high leverage quality signals right there may be other quality signals also more explicit let's say a review on G2 or any new website or a case study being published. But much of it could be fabricated today, right? You could, you know, come up with a really fantastic case study in some corner of the web which you might use any any name on the internet, any company as any company logo which which is which may not get refuted because it doesn't matter to a large company. Maybe they don't even have the bandwidth to go and refute come after a small company, right? But an AI agent can confuse that to be a real case study and maybe you know give you the weightage. So they are looking at more hard to do stuff stuff which is you know not easily replicable by just you know generating content something that requires let's say

39:40Human work and a threshold of of let's say actual validation right of let's say customer trust or consumer trust or community trust in e-commerce per se I think what is really needed is I think you know a lot of Amazon reviews trust pilot And these are so any platform which has which has sort of made reviewing frictionful so that only a human can come and review that will actually be very important. So the social proof is becoming ever more important because I think that will be across categories that will be a great signal. So social proof community community verdict on that brand I think has will carry a lot more weight again because it of course it's harder to harder to maybe you know maybe fudge it or harder to maybe you know fake it. So if it is a larger community a broader community which likes the tool I think that is a better sign for AI to start entering that into recommendations.

40:46Yeah absolutely. Makes sense. Perfect. Lalit, when any last thoughts on anything that you've heard or seen this last week? would you like for us to discuss here? Bring it to our listeners.

41:05Yeah. So let's do a simple you know like like an exercise by going going in a circle around us. So why don't we start with Ashwani and Ashwani just talk about one of your last prompts on charge GBT or perplexity which you can share should not be too private but how long was it what was it?

41:26So let's for me so again we at our company are trying to build a community as well. So my first my latest prompts were actually two things that we are working on and prompts were around that. So the community prompt was was very short and just a exploration. So that was on charge GPD just an exploration of you're building a community and this is this is this this is the goal this is what we want to do how should we structure it. So I think this was very businessoriented, very specific. On the other hand, there was a client requirement, a customuilt requirement which was there. There the prompt was a few pages where I put the full requirements, uploaded the full client transcript for the call, screenshots of what I want, what kind of a product I wanted. So that was to cloud design and that got of course in the first go itself that was a great output and some minor modifications needed but that prompt was way more done because I wanted a very very expected very specific result out of that. So yeah that was my exploration latest. Yeah

42:34V how about you

42:35I just checked my chart history. So similar to Ash's use case only. So I'm working on a project with a client and I wanted their design system to give it to my coding agent. So I gave the URL of the client and it went on to their website prepared a design docket kind of thing their color palettes logos downloaded logo from their source this thing and their font file everything into one Google drive folder which I just gave it to another agent to create some mockups to showcase to that client. So it did like I think took 5 minutes kind of what about yours?

43:23Let me launch your audio.

43:25Okay. Yeah. So I was on mute. Sorry. So my my very recent prompt with which actually I just gave while I was in breakfast. my daughter was actually reading about whales, right? and and then we were with the discussion moved towards how whale oil was so important for let's say lighting lamps and then suddenly petrochemicals and the the kerosene industry sort of you know replace that and how before that large number of whales were getting killed so I I sort of just had a a longish voice conversation I you know get with GPT where it sort of took us through the history of how the whale industry has evolved and at what points what regulations sort of came in to protect that specy. So pretty interesting. Yeah. How it is becoming part of our day-to-day life.

44:18Right.

44:18This format looks good. I will implement this in our theme standards as well.

44:25Yeah. Awesome.

44:27Very cool. Yeah. Alth just one question. Your output of that chat was just a just a text based output or maybe a voice based just to understand the topic more or you actually got them to got the system to create a visual story line as well of what happened what was the history.

44:44No, this was on the breakfast table. So it was very much about them narrating it. and GPT live is my go-to model for any conversation. In fact, I've I've u see today if you go to Twitter, right? every day there are at least 50 new things that you you come across, right? And then you don't have the time in your day to go deep into anything. So like I think 2 days ago there was this new model which is like a classifier like a new model like Java model, right? So I I had less time budget to go deep into it. So on my commute I just had a voice conversation with GPD live you know got to understand what it is what's the architectural breakthrough and what are the use cases that is valid in and GPT live searches internet on the behind on the back side sorry behind the scenes and then does a pretty good job of having a really intelligent assistant who is you know always by your side is thorough in information access etc and you can ask them questions and interact and go deep and come back come out come come out with a really like whatever understanding whatever level of understanding you wanted to get into you can come out with that.

45:59Fantastic. Yeah, this was an interesting addition to the podcast. Love to hear what everyone is doing with the with the power that has been given to us.

46:10Yeah.

46:10But another great episode and thank you so much Ve for joining in for the first time. We'd love to see you more often on the podcast and LA as always great to have you here and great insights for this conversation today. Thank you. Thanks. Thanks. Thank you.

46:27Thank you.