Agentes personales de IA explicados en 30 minutos (Instinct)

0:00Instinct has become the personal daily agent that I am using every day. It manages my email, alerts me to AI trends, and also supports me with some marketing tasks. In this video, I will break down everything I have learned over these 15 days of daily use, and I also want us to see everything it can actually do, some of the daily use cases I am applying, and what sets it apart from using Claude, Codex, ChatGPT , or any other alternative on the market. What is the difference? Between a personal AI agent and an AI chat like the ones we already know. That said, let's begin. Let’s start with a bit

0:37of context. Instinct is this right here , and it is a personal assistant. Its website is so simple that it explains everything it does in three paragraphs. It is a personal assistant; you don't need new interfaces, you can use it directly from WhatsApp, and it will help you with everything you do in your day-to-day life, from looking for birthday gifts to managing your orders,

0:58to booking any doctor's appointment. Basically, it lays out that it can do everything you do in your daily life. Instinct is one of the alternatives currently standing out in the market as a personal assistant. The first thing is what differentiates it from the usual tools, right? And that is that we are used to using ChatGPT, Claude, Codex, or any of these alternatives on the market. And in this case, we are talking about the big giants, OpenAI, Anthropic, Google, and in this case, Instinct is not the same. It is a personal assistant born without having any big company behind it. This means

1:33it is a new alternative in the market. It is also true that if we compare it with these big companies, they have all gone in the same direction; they have all released their own personal assistants. This means that within OpenAI we have, for example, GPTs, which is the alternative for personal assistants, AI agents that work as a team and do everything for you 24/7. We also have Meta AI, which in this case is Meta's alternative, more focused on social networks, on Instagram, on WhatsApp. We also have Q, which in this case is the Manus alternative. As you can see, they are all very similar; they even have the same identity, right

2:08? Being a character that represents your personal assistant. If we go into Grok and X, we also have Grok Bots, but in this case, it's designed more for working with different bots simultaneously. And we also have other open alternatives that have been coming out, like Hermes, which also released Bot Mode. In short, the big difference is that to this day, ChatGPT, Claude, Codex, or any of these alternatives require us to open the computer and trigger the actions ourselves. We can take it from a lower level to a higher one, but in any case, they always have that dependency. We can work from a chat, have MCPs, have different tools,

2:45have commands, keep leveling up, but it always requires us to open the computer and launch things ourselves. When we work with bots, what we create are identities that we can somehow leave running 24/7 and that we trigger through messages in interfaces we use regularly. In this case, for example, Instinct is positioned as an alternative to be launched from WhatsApp. It is true that both Growbot and OpenAI work directly from their own web interfaces. Well, having said that and now that we've contextualized it, it's not the same; these are different alternatives, and the main thing is that in this case, they position themselves as personal assistants. That

3:26is to say, it's an AI that knows everything about you, that does everything for you directly, and that you don't need to configure absolutely anything. After trying all the alternatives, Instinct seemed the craziest to me because it can do many things for you, from presentations and sending you articles to generating artifacts, sending emails, and above all, its proactivity. This means that when you are talking to Instinct, you can tell it, "Notify me every day, keep my email in mind, and when something important arrives, totally autonomously , Instinct sends a message and alerts you." But it doesn't just alert you; it has the general context of what you

4:04usually do with it. For example, let me ground it in this case study. I published a new video on YouTube, and directly, without me telling it anything, we see here that it starts the conversation, telling me, "Your new video is already on YouTube, it lasts 27 minutes." And here you have a couple of notes with small abbreviations, content, or valuable contributions that I extract from my own content and that you can directly publish in another alternative, such as short notes on LinkedIn or Substack. It gives them to me already written so that I can copy and paste them directly. But it's not just this; if I say "OK" here, just by

4:37putting it like that, it takes it, sends it, publishes it, and schedules it for me. And this is just one of the many things Instinct has been doing for me over these last 15 days. Now, later on, I will break them all down for you because it is insane. But before that, let's get started and go step by step. The first thing is how to access Instinct, how to register, and how to

4:59use it to get the most out of it. Instinct was launched directly through invitations. This is a strategy also used by other players like Spotify, meaning each user who registers gets 10 links—I think it's five now—that you can send to people so they can register. Why? Because if you land directly here on their web interface, go to the registration section, and enter your phone number, it tells you that first you need an invitation and second, there is a huge waiting list, although it is possible that, from that waiting list, a spot opens up and they grant you access. It also depends a bit on their own market strategy. In this

5:38case, once you are registered and have obtained access via a link, the next thing is that you find your own interface here to manage Instinct, although you don't really need to use it for anything. I've been using it for 10 days and I haven't entered here at all. Moreover, it's something they have been evolving and developing over time. Once you have the account, you enter your phone, they send you a PIN, and you're in; this is the configuration interface. They state it themselves in their value proposition: it means they don't want to have interfaces and want

6:09everything to happen automatically. This is what ultimately breaks the learning curve so that anyone can do amazing things without configuring anything. Before, you would go to GPTs, configure them, and manage to get them to do minimal things you wanted, but they only returned text and not much else. The GPTs have already disappeared , the GMs will disappear, and Instinct is that 2.0 version, right? Where an agent itself is self-configuring. So, when we are inside this interface, the first thing you have are the access points you have for Instinct. You talk to Instinct via WhatsApp, as we have seen in this section here, and to talk to it, you have an assigned phone

6:46number here, which is where you write to it. You can also talk to it from messages and from email. If you are in the United States, they also have it enabled to make calls and for you to communicate with it through this medium . If you are in Spain, in this case, it is not yet available. Once you have it, you can also see the different

7:02connectors they provide you with here. Still, super important—it may not look like much, but Instin has the ability to work on a computer that is in the cloud. This means it opens its own browser and can log into every single thing that you are logged into. For example, let me bring it down to my own use case. I use Substack for my newsletter, and in this case, it’s

7:22not a connector that I have available. But if I tell Instin that I want it to log in under my name on Substack and schedule my posts, it does it, because in that case, it doesn’t rely on connectors, but rather directly on its own ability to launch a browser and navigate as if it were you. This is very similar to what happens when we work within Claude Chrome, where we have an extension that, let's say, operates the computer for us, but in this case, it’s all happening without you even being aware of it. All those things you see in any chat where it’s reasoning and doing things, in this

7:54case, you don’t see them, but they are happening in the background. Still, something interesting: I’ve gotten so much value out of it and I’ve only connected one single connector, which is Google Workspace, because giving it access to my email is what has allowed it to log into every single thing I’m registered for, which is basically what I’m telling you. From here, you can also delete any connector in case you’ve given it access to something you’re not interested in. Here you also have the different logins I’ve been making, for example, what I mentioned: it has logged into my Gmail, it has logged into my LinkedIn, and also into Skool, things that usually

8:26any regular agent cannot cross the border because it’s limited, and in this case, since it goes through a browser, it can. If you want to do this in Claude or Codex, you need to set up Playwright or Browser-use, any browser that does these things, but in this case, it’s automatic. Here you can also go and see how it saves things. In my case, it doesn’t have anything added, even though it knows so many things about me. Also, here comes a super cool and interesting profile, which is that my Instin communicates with other people’s Instin. So, I actually did the test with Marta, and in this case, what we were doing was

9:00managing a lodging reservation, so our Instins are reaching an agreement to make those bookings while keeping us informed. Let’s see it in the example . In this case, to see it, I have both WhatsApps open. Here we are talking about Marta's phone conversation, where the first thing it says is, "Hey, Lorena has already accepted your

9:21Instinct request and you are connected. Therefore, what can I do for you both? What things do you have in common?" From here on, we told it our situation and, in this case, we told it yes, that we needed to look for a hotel for Bali during these dates. We gave it some small conditions and let's say we delegated it to Instinct, although Marta had a bit more of a role in carrying out this search. From here, something cool and interesting is that I said to it, "Hey, can you send a message to Marta's Instinct and ask if we have the reservation made yet?" So her own Instinct answered mine and said

9:55, "Hey, Marta, keep refining the list. Here are some of the accommodations she is considering." So from then on, our own Instincts were talking about the accommodations we were going to book. The solution was that after all this, Instinct set up a favorites list for us on Airbnb, where we saw the four or five options and chose the right one from there. What's cool is how the two agents synchronized to somehow reach the conclusion of her requests plus mine, even working on it together and then giving us an answer like humans, right? In the end, her assistant and mine coordinated all the work. All this is managed from this section here, and

10:40let's say here you can see the requests and who you are connected with, but the reality is that it's as simple as telling Instinct to connect me. With Mike's Instinct, or Marta's, or such, you give it a name and give it access to that person. It generally goes by phone number. So, once you already have that person's contact associated, which is tied to their own phone—which we could say is that login part within WhatsApp—we can accept the request and it can be done directly in natural language. too. Well, here once you have it, you can manage it if you want to delete it or go back. And with this, we

11:13have the quick overview of the settings it has. There really isn't anything else besides inviting a friend, which is where that link is where you can invite other people and some small preferences about your own data, your phone, and such more sensitive things. That's why I'm not showing it on screen . Now let's look at the cool stuff and the use cases I've been finding in my day-to-day, because even though this interface doesn't seem like much, it's true that Instin on mobile opens up two or three more interfaces that provide great usability. Let's see it in different practical cases. The first practical case is image generation. For

11:48me, within the realm of marketing, creativity, and content, it's essential that it has the ability to generate images, whether with image models or based on code. That is to say, that ability to generate any design with code behind it, like an artifact, a presentation, or something similar. In this case, it was as simple as giving

12:05it an image prompt. Here we have it. And following this image prompt, the first thing it does is give you a check-like status, like "I'm working on it," so you know it's on it, and once it has it, it returns the response. Look at the minutes: at: 38 I already had the request, and by: 39 I already had the response here. The image is awesome. Let's say that if we compare it with the vast majority of models on the market, it's on par. The same image

12:27generated with ChatGPT looks like this. Here we have the generation inside ChatGPT. And in this case, so you can see, this is a completely free account, so you can see that it's generated with the lowest model they have here. It's the same prompt, and here we have the generated image. If we look, it's exactly the same; the variations are minimal. I would dare to say that perhaps behind Instin they might even

12:53be embedding this model for generation. I assume they have different models and will launch one or the other depending on the user's request, right? Ultimately, depending on the reasoning required. The thing is that it's the same and maintains consistency. Perfect . To also see this consistency case, I asked it to make five variations. In this case, here we have the five variations, the reference one and the ones it generated afterward. Here we see them larger, and as you can see, they are very good. I mean, they maintain consistency perfectly. I didn't perform this test in ChatGPT to see the counterpoint, but if it did the other one well, I assume these will

13:32follow the same line. I also took it a step further because for image generation, it's superb. As for small limitations, it's true that the limit is 50 images, but we're talking about the fact that right now they are in acquisition mode. Instinto is 100%free. You can try it without limits, use it without limits, 24/7, and the only limitation is 50 images per day, which is already a much higher quota than the free version of ChatGPT gives you— which I think is two or three free images a day—it multiplies it by a lot. I also said, "Okay, let's test how consistent it is with code generation, because in this case it is true that we

14:08are already talking about how those images are uploaded to a computer, it builds the code, it's HTML, it converts it to PNG, JPG, or any format, and brings it here." And in this case, it did it superbly well too. Here we have, for example, the reference, and it's all these images we've worked on, laid out within a single image. So this is pure code because, if you notice, these are the images that were input. It generated this right here, and once it has it, it passes it to you directly in a file. Something very cool that we'll see later on is that what it has uploaded here, it can place wherever

14:40you want. In this case, I went a little further. "Okay, you know how to do it, let's see how well you maintain consistency working with code." So I asked it to change the typography to the Poppins font. Here we have the result. It's also awesome. It has totally maintained the consistency of what we had done before; it changed some small nuances, right? Just as the Poppins font is a bit more robust and thicker, what it did here was also change that small line. Also, a small side note: the prompt I gave it was very vague, just directly "change the headline to a typeface." I didn't even

15:13tell it to maintain consistency or anything. From here, something cool and interesting is: "I want you to adapt this into an image generation class where we explain everything we just saw ." And in this case, I want you to apply a methodology that I had already told it about and had already given it context on what it consisted of, and that I wanted it to apply both to that context base and what we just generated within a presentation. This was at minute 48, it marked the check, and by minute 56, I already had the presentation. The cool and interesting thing is, I didn't ask it for anything,

15:46but since it's already linked to my Google Workspace, what it did was directly give me the presentation in Google Slides. So here I have it in a totally editable format, and even if they don't have an interface or that editing tab, you can come here and make any modifications. Since I'm on WhatsApp, it directly sent me an image of the cover so I could see it without it being a heavy file. And from there, it also sent me the next one and then decided to send me the presentation

16:15already exported so I have the result. Something cool is the proactivity it has to give you things in multiple formats in case you want to edit it, see it quickly, or download the presentation itself. Here we have the presentation ready to download directly . In this case, as I was telling you, this methodology is applied to each of the images. I mean, it seems crazy to me because the prompt is so simple, so vague, we haven't connected anything in the background, I didn't even tell it to use Google Slides, it did it super well. Whether you want to use it for images or generating visual content, I

16:48recommend it 100%. When should you do it here? Compared to GPT, Claude, Codex , or any alternative depending on the level of customization. Here it serves for something quick, practical for day-to-day use, and getting by. If I have a system that is very well set up, orchestrated, tailored to my brand, and built out, then it’s true that you would make the jump to Claude or whatever other tool you're using. Let's go to the next use case, case 2, and in this instance, it's the typical

17:15scenario we all face: email management. In this case, I gave it access to my email and used it for four different things. The first was to unsubscribe from everything that was spam. So I asked it to search for all marketing emails and make a list for me. Based on that list, I would decide what to keep and what not to. That list, once again proactively, it loaded into a Google Sheet so I could mark what I wanted to

17:38unsubscribe from, keep, or review. Since I was on WhatsApp, it also decided to give me the same list directly in a PDF in case I wanted to view it in a more accessible way. After that, I reviewed it, completed it, and then asked it to unsubscribe from all of them. Suddenly it unsubscribed from all of them, including some that sent you form after form after form and made you jump through three steps to unsubscribe. I also gave it a list of what I didn't want to unsubscribe from, and it kept and respected that very well. The second email case was maintaining alerts for different newsletters. This is a typical scenario

18:15we can handle with ChatGPT Codex using any routine, making it very easy, but in this case, I like the proactiveness it has and how you can interact with it so comfortably. So, every morning it sends me a message like this one. Good morning, here are your news newsletters . I highlight this; I don’t use it just to summarize everything, but so it has the judgment of what applies to me and what doesn't. So in this case, it gives me two news items; I think there could be more interesting ones, so it refines different stories a bit more for me. This is a very typical and common case, but it can also be done

18:44from [it] and it is very cool and convenient. Continuing with the same routine and expanding it a bit more, it also has control of my email regarding collaborations. The thing is, generally when you create content, you get a lot of sponsorships, many people who want to talk to you or create different things like partnerships or similar. And in this case, the vast majority are agencies, there is a lot of spam, and it is totally overwhelming. You can have a routine that summarizes it for you, etc., but it usually ends up going nowhere, and having this distinct proactiveness here is insane because the first thing it tells you is what

19:15interests you and what doesn't. What doesn't interest you doesn't even bother you, and what does interest you, it tells you directly who it is, verifies if the company exists, and if it's an agency or not. And then it tells you, "Hey, be careful with these things, as there might be something suspicious or hidden here." And furthermore, in case you are interested , you can also tell it directly, "Okay, generate the email and from there, send it or handle it." It fully manages the conversation with that brand based on the context it has about me. If, for example, at any point it requires my attention, it sends me a message. If I

19:48don't reply, it decides after a few hours to resend the message and tell me , "Hey, you haven't looked at this, but I think we can do this other thing this way." It has a proactiveness that works very well, and for these types of scenarios, which are boring to manage and deal with, it ensures they don't fall into a void, because any routine with ChatGPT Cloud, in my case, ended up going nowhere because I’d eventually move on, stop replying, wasn't interested, or never found the time. And in this case, it’s not just that you don't find the time: it’s that it reminds you, insists, handles it, and filters it for you. Also here,

20:21the fourth use case is that, continuing within the same Google Workspace, for the little things that require something from me, what it does is generate a task for me, and that task is something that links to my calendar, and I do set aside some time to check it because it needs me, because one of the problems with this is that you aren't checking WhatsApp 24/7. So, in the same conversation, you could be talking about many other things. Well, for example, here I was talking about that same trip to Bali and how it's organizing things for me. So, if it tells me about an email now, most

20:52likely it didn't interest me or it doesn't matter to me. So, if it considers it important and that I'm going to leave it hanging, and that the time to deal with it isn't today at 8:00 p.m., but tomorrow at 10:00 a.m. during work hours, it leaves it for me in those written tasks, all with that level of proactivity. So much for this use case as a personal assistant in Gmail, but now we're going to see some that are even deeper, and that is that I wanted to put it to the test and see how far it was capable of going. It also gets into video format and does it

21:19very well. In this case, what I wanted was to ask it to take my long videos and get short videos out of them, for it to edit my shorts. So what I did was directly send it a YouTube video of mine and ask it to make a short video from it; understanding that my YouTube videos have the least editing in the world, that they are generally poorly recorded, and that they also usually have few visual resources. But, without giving it anything, I wanted to see how far it was capable of going, what it could pull out, what it could edit. To break down the process, to get this

21:49video, it needs to download it. After downloading it, um, transcribing it, that transcription, cutting it importantly, then editing the video, removing silences, packaging it, rendering it, exporting it for me, and then sending it to me here. All of this , generally in Cloud Code, you ask for it and it keeps asking you for many things, it needs you to do

22:08installations, dependencies, et cetera. Here, everything happened automatically . One of the small limitations I encountered was that YouTube blocked it because it was a bot. And in this case, I told it, "Hey, well use this tool that generally works very well for me, and from there you'll be able to download it." So, here it brought me the video. It made two versions. The first one is this one here where, being hosted in the United States, it took my video in English, so it transcribed it and edited it all in English. This is the video and here we can see how it has been cropping it. From there, I

22:38gave it two new premises. First, that it do it in Spanish, and second, that it doesn't leave me in such a short format, that it crops me well, and puts me in a 9:16 ratio. With these two small natural language adjustments, it redid it for me. Like this. Here we have it in Spanish already. And in this case, as we can see, what it did was adapt to me, crop me better, and adjust things so they look good. Something cool here is, for example, in this case , different illustrations are appearing , while in the original recording they all appear together. Well, what it did was cut two, then another two, and then

23:11another two, and it edited them based on what I’m saying. That is, it is trying to unify that image, understanding that I am on the right side; it is cropping me and, from there , placing the available illustrations above. The result is superb, considering we didn't ask it to go any further, and from here, it just gave it to me so I could download it. But we can also take it to the next step, which is: publish it directly to YouTube, or refine the design, or directly every time I upload a video to YouTube, do all this repurposing automatically without me having to review anything. Let's go to the fourth

23:48use case, and in this case, we're talking about the model's own creativity—something many usually fail at, or in finding real references or going a bit further. The thing is, I have a newsletter, and in that newsletter, I usually make memes; memes generally require me to think of how to make the joke, and second, to look for real memes. Third, to edit those memes,

24:05and then upload them to the newsletter. That is to say, it's quite an extensive job that I used to do through different APIs, combining them and editing with Claude, but now I’ve gone a step further—tell it to do it and see what comes out. So I gave it the draft of my newsletter, and from that draft, I asked it to say where it might be interesting to add a meme, and from

24:27there, to create them and upload them. The first test went wrong. Obviously, many things don't go well at the beginning, and it had a tendency to generate the actual image. In this case , it would use a generative image model and imitated something real, which is this meme here, but it generated it from scratch. I told it I would prefer it to do it with the actual meme, since it looked for the reference, not just to regenerate it, but to use it and generate based on it. So here it took the next leap, it looked for the real meme, which in this case is this very

24:59famous image, and it made that joke based on what I was talking about in the newsletter and set it up here. The next step, cool and fun, is that it did the same with another example and then uploaded it to Substack and added it into the corresponding space. In this case, well, here it put this one for me to open and as you can see it adjusted the entire screen and also, well, down here it put the other one. Here we have it. In this case it is interesting because to get here the model itself

25:28had to navigate through many pages. Reach my own profile, sign in to the browser, from there look for posts, then go to drafts, in drafts open the article it is working on, and in the article, look for the space where it belongs, and add the image. That is to say, it has quite an extensive navigation job in a tool that has no API, no MCP, and that makes it very difficult for a bot to register and log in, because in this case, when you try to do it on your own, it often triggers many verifications and controls. And now, let's go to one of the last use

25:59cases and in this one we are going to see the different interfaces that Instinct actually has. The first thing is that when you ask it for something very long and that long thing cannot be housed within the different tools or connectors you have given it, it has the ability to create a kind of artifact that gives you a link so that

26:15from there you can access and see it. Why does this happen? When you ask it for something huge, in this case I told it, "I want 100 video titles, 25 per content pillar." What it did was generate this for me here. It generated a kind of artifact for me, it looks great in mobile size, it is responsive and you can read it well. And from here

26:32what it did is add that list for me. Why? Because it is very extensive and if it did it for me via chat it would exceed that length, but it does it totally autonomously and it doesn't ask you to tell it anything. Next case, which has another one of the interfaces that also makes your life easier, and one of the things I liked is the security. You can't directly ask it for things or give it a password to sign in or access things for you, because then it remains open and, let's say, the model is drinking and inputting that password into its own LLM, so it

27:02remains exposed. So, something it does very well is that it gives you a link that it creates and generates on the spot so you can enter passwords securely. For example, let's suppose I want to connect any of the tools we've seen. In this case, I'm going to go with Notion because I haven't connected it yet. I say I want to connect Notion, and the first thing the model will do

27:21is generate that connection URL for me. Here we have the URL, and what it does is that if we click it, it directly opens this connection window for Notion . Since I'm already registered, it opens my own profile, and from here it tells me, "Hey, which pages do you want to grant access to?""What do you need from here?" If what you need is a password, it gives you a secure space to enter the password so it doesn't get exposed. Everything involving granting access, passwords, keys, or permissions is done directly in a URL generated on the spot, and it will never ask you to pass them directly through a chat. It

27:54gives you some assurance that the data is being handled, at least clearly, by the model correctly. And so far, we have seen some of the daily uses I've been coming up with to test Instinct that have served me in my day-to-day life. Obviously, there are many more, from asking it about the weather to having it manage a restaurant

28:11reservation or handle any type of task. It does the basic and common things superbly. Now, the important part: when is this worth it, and when is it worth going to other tools? Exactly in this case, where I show you the 100 YouTube video ideas, what I did to set it up and make it work well was to tell it to analyze my entire LinkedIn channel, analyze my entire newsletter, read all the articles, and also pull in all the videos I have already published on YouTube. I did that by giving it a prompt like this one here. And the reality is that it did that analysis part well; it extracted my metrics,

28:46pulled out each of the things that were working, including which keywords my audience is finding me through. It did very well to base it on internal data because it used my Google account, went into YouTube Studio, and pulled all the metrics from there, metrics that are generally not public. So far, so good. What happened? That this result is actually quite bad, and if I compare it with what I have set up within Cloud

29:11Code, it leaves a lot to be desired. Why? Because here it is only taking into account the memory that knows me, that has analyzed my videos, but it is not taking into account how the different elements that I do have set up in my YouTube harness are linked. Let's say that here it searched for case studies, analyzed pillars, thought about the audience, and extracted different things. Perhaps that prompt could have been improved to make it go deeper, but let's say that in this case , we are already entering an expertise where what I set up in Cloud Code is 10 times more useful to me. Let's say that

29:42I wouldn't get a single idea from here that serves me or works for me, because in the system I usually use, which is the one you have here on screen and which I also have in this other video I'm leaving for you here, what I build are all the methodologies for how I handle every single nuance: strategy, analysis, competition, content, and also how that links to memory. And not all memory works the same way. Each memory has a small base and connects to

30:06certain tools, and each skill does too. The data comes from the YouTube API, so it is real-time data; it doesn't just look at mine, but also evaluates market size, and let's say the complexity is much broader. It would be useful for thumbnails, for quick ideas, for making those little Shorts—things that are not core and where I don't want to, or haven't, set up my own harness with my own methodology. Things where you already have that expertise, you know your way of working functions, and you

30:35want to adapt it to your own style. That is where, for me, Cloud Code—or in this case, Codex if you are within ChatGPT—becomes the tool to use. The difference is that Instinct or any of the alternatives we've seen are ideal for day-to-day operations, and the rest of the tools where you set up your own systems and harnesses are better for everything that requires more expertise . Now, a quick side note: this is today , and surely in a few weeks or a month, this will have to be re-evaluated because the market will have changed completely. Anyway, I'll see you in the next video, and if you want to try any

31:12of the different agents I've left around, or if you want to know my opinion or take it a bit further, let me know in the comments and I'll keep it in mind for future videos. Yeah.