Transcription
I've got a question for you guys: what exactly is Chat GPT? Well, obviously we know it's a large language model-based platform, and a lot of people describe Chat GPT as their AI assistant. But is it truly an assistant? I mean, everything that you do with Chat GPT actually has to start with you. Yes, it can save you time. Yes, it's pretty much magical technology, but you have to prompt it correctly. You have to give it all the necessary information it needs to complete whatever task it's supposed to complete. And, well, it just plain can't go off and do automated tasks for you. It is no automated agent; it is no true assistant.
But this changes today. You see, I've discovered a platform that uses AI to specifically create automations. It can be custom-tailored to complete whatever specific tasks that you need it to complete. This platform is, don’t get me wrong, a little bit more advanced than Chat GPT, but it is absolutely set up in a way that makes it easy for new people who have never used this sort of platform before to get started.
And I'll let you in on a little secret too: I've been tracking this platform for months and months—probably over a year at this point—and before now, before just like a few weeks ago, it really didn't work all that well. But they've just had this major 2.0 upgrade over hul that now makes this thing an absolute butt kicker. Man, this thing is amazing! So, join me today. I'm going to show you guys how to use the platform, how it works, and basically why this is a game changer for so many people if they can learn to apply it correctly.
I am very happy to introduce you guys to Lindy. No, this is not sponsored in any way, shape, or form; I'm just interested in the cutting edge of AI automation, and this is exactly that.
As you can see on the homepage here, we've got a bunch of different little icons with names in different categories, and these are all a bunch of presets or pre-made Lindys created by Lindy. You can see "meeting notetaker," "email responder," "sales meeting recorder," all this kind of stuff is AI apps that we've seen out in the wild before. But like it’s an entire thing that a company is doing, right? The whole company is based just off of being an AI meeting notetaker.
The thing with Lindy is that you can build all of these custom tailored for your use cases yourself, and it's not just meetings. Lindys can have thousands of different connections, and they can do multiple things at once and even communicate with each other. You can see I've got two Lindys that I've made on the side: I've got an adventure game and then a calendar Lindy. The adventure game is a work in progress, but the calendar one is up and working.
If I go into the flow editor, you can see kind of the diagram or outline for this Lindy, and this is where you build the Lindys. This beautiful flow editor didn’t used to exist, by the way; this is all new for Lindy. I'll explain and teach you all how it works. I know it seems a little bit complicated with all these different flow charts and stuff, but it's very simple, easy, and intuitive to use. I actually love it!
So how this Lindy works is we send a message to it, like a bot. It's a chat bot, actually. This is where the chat actually begins. You send a message, and this triggers the Lindy. You can see when I click on it towards the right, we actually get all of these in-depth settings that we can change for that specific node or part of the flow editor.
So we could do, you know, a greeting message. Since this is a calendar bot that I made for myself, "Hello Matthew, nice to see you." Sure! And you can also add conversation starters, like you can when you build Chat GPT bots, for example. But this is like way, way ahead and way more advanced than Chat GPT custom GPTs or bots. But yeah, that option is there. Of course, since I made a change, I'm going to click save up in the top right.
Once we send it a message, it's going to have this greeting message, and it's going to push us on over down to this AI agent. You can have as many of these AI agents as you want, by the way.
So if I click on this AI agent and I zoom back out here, you can see to the right, an AI agent basically is an AI that decides what to do until a jump condition is met. Don't worry about this word "jump condition." I know it sounds like some complex thing you have to learn, but it's actually super simple and is probably one of the key features that makes these AI agents so much more powerful than any other AI agents we've seen before.
Anyways, we'll get back to that, but the prompt for this AI agent is: "You are a calendar management bot. You observe, create, and manage events for Matthew." Again, this is a personal bot for just me. You keep in mind the timing of events and will check the calendar before adding new ones to observe any potential conflicts.
You can see we can actually decide what model powers this agent under the hood, and you can see if we click this, there are a lot of different models to pick from. A lot of OpenAI ones: we've got GPT-4 Turbo, GPT-4 Mini, GPT-4, we've also got Claude 3, Ha Cou Sonet, and Opus in here, and 3.5 Sonet. I've honestly been using 3.5 Sonet for a lot of different circumstances, and it's worked pretty well.
But anyways, this one is just on the default GPT-4 Turbo for now. We've got this "ask for confirmation" button here. I'm not going to explain this just yet; I don't want to overwhelm you guys with information.
If we scroll down, you can see the skills for this specific AI agent, and these can be applied and removed at any time in this flow editor. This is another huge part of this: it allows this agent to actually connect to my Google Calendar, create events, create events based on past events, delete events, format events for display, get event details, list best times to schedule an event, update events, view calendars, and view events. This is full calendar access essentially, and this is not something that you have the ability to do with Chat GPT. Even Google's own, you know, Gemini doesn't have access to your calendar like this.
Of course, we can also add skills, and when I click that "add skills" button, I mean there is a plethora of different applications from different companies and different things that you can add and apply to build your Lindy however you want. Again, that is the beauty here: it's very simple to use, very easy to use. You don't have to be a programmer; you don't have to be a developer. You can just use this thing.
You can see they have GitHub, Google Cloud, natural language; you can do it based on analytics, Invoice Ninja, home assistant, HTTP. I mean, if you are a developer, there actually is a lot more you could probably do than someone with less experience. But again, it's still very much open for those who are not that experienced. PayPal? I mean, it's really awesome.
The one thing that I'm really missing from this is actually Discord. I think that Discord integration would be really cool, but I bet you can still figure out some way to connect this up to Discord.
Anyways, yeah, so like I said, plethora of different apps and different actions; all these apps have their own actions that apply to just that app, and then of course, there's also basic chat options as well.
Anyways, getting a little bit ahead of myself again, I don't want to overwhelm you guys, but just look at all the connections you can make, and that is the AI agent settings essentially. So we send a message, and then this AI agent, again, with this specific prompt, this specific model, and these actions can go to work and adjust my calendar for me.
Don’t worry too much about this right now; I'm going to explain that in a second. For now, let's actually just show you how this works. So, I explained it to you; let's test it. So back here on the Lindy homepage or the Lindys page, you can see "Hello Matthew, nice to see you"—that thing I typed out is definitely there.
I can say, "Hey calendar bot, can you tell me my schedule for the rest of the week?" I'll just send that message on over, and now it's going to start thinking. What it's going to do is it's going to go ahead and connect to my Google Calendar, automatically view the events, and then give me a nice breakdown of what's going to happen. Pretty darn nifty!
"Can you add a new event for me sometime in the future? I want to go get a new TV. When do I have time for that? Keep in mind I have a pretty formal work schedule."
Let's send that off. "Here are the best times," so they give me some various times based on my schedule. It asks me, "Which option do you think works best for you?"
So you know what? Let's say I want Monday, so I select option two and add that to my calendar. No problem! It goes ahead and makes a new event in my calendar, gives me kind of the rundown, and a link to the details.
Pretty incredible! Again, Chat GPT can't do this. I don’t think I can even think of another app that allows you to connect to your Google Calendar like this and actually make events. Really cool, though! You can make this much more advanced too; you could have it CC specific emails if you wanted to, and the list goes on. Of course, it's AI, it's very malleable, and it's super powerful. Lots of potential here; this is an extremely simple example of what's possible with the Lindy platform.
Anyways, guys, let's head back on over to that flow editor, and you can see that was us just using this little node over here: a user sends a message and then it activates this AI agent that manages things until it sees fit.
Now, I want to focus on the other half here, and this is something that I set to trigger on a recurring schedule. So, it starts with this trigger, which is a recurring schedule every day at 8:00 a.m. It triggers this AI agent, which is very, very similar. However, the prompt is just slightly different: "You are a calendar-checking bot designed to check Matt's calendar for a reminder to see what is going on. Give Matthew a summary of his calendar for that day and upcoming events later in that week in a format that's easy to read."
So this one I got working with GPT-4 and it has been working fine for me, and again, it has all those same skills from the calendar. Of course, though, it isn't really supposed to update or change events, so I could remove those specific skills if I want to, but it's pretty darn smart. It's an AI model, so it's not going to add events randomly—or at least, I hope it won't!
Anyways, once it does that—and this is where we get into those jump conditions. If you remember, I mentioned that it's really not something that you have to be afraid of. I know it sounds like a word that you don't know what it is, and it's a part of this flow editor; a little overwhelming, but essentially all it is is: when should we jump out of this AI agent? This is a condition that needs to be met for the AI agent to actually leave and go on to the next task.
You see, this one doesn't have a jump condition because it's the last thing in the loop. It's just a simple AI chatbot for setting up your calendar. But this one actually does something and connects to something else.
So once this AI agent has checked my calendar, it can now move on to the next one, which is sending an email. You can see this is set up to actually specifically send an email to my email address for a basic summary of my events for the day every single morning—that's the point of this.
But this AI agent is actually able to tell that, "Oh, I've checked the calendar, I've checked Matt's calendar; we're ready to move on to the next task." We can continue on and do this jump condition and then move on to the email sending. It's really incredible! This, you know, moving from point to point to point used to be a huge struggle for these AI agents, and now it pretty much works flawlessly every single time.
The jump conditions are a huge part of the reason this flow editor is actually able to operate and work in the way it does, and I'm really impressed.
So yeah, that's my little calendar bot that I built yesterday. I think this took me like maybe a half hour to figure out and build.
Anyways, I've also been working on this adventure game bot, and this is completely unfinished, but I thought I would show you what this node looks like because you can see how many different flows you can connect to actually produce the experience that you want to be produced with the Lindy.
You can really connect them together; you can make something incredibly advanced and incredibly complicated. So I think hopefully some of you guys are now starting to see the power potential of the Lindy.
And remember, it gets even deeper! Like I said, two Lindys can actually connect and talk to each other. So why don't we go and try to build something today? In this video, I want to see if I can build an automatic AI news scraper or gatherer to summarize maybe AI news every single day and send me an email or something like that. I think that would be pretty cool.
Obviously, you could let your imagination run wild; you could build some really crazy things. Honestly, I would love to hear your ideas for what you could build in Lindy or maybe you actually, after this video, go off and use the Lindy platform to build them and then share them in my Discord server or something like that because you can actually share Lindys as if they were GPTs or something like that from OpenAI.
Alright, so let's make a brand new Lindy. Again, they do have these templates that you can kind of start off with. You know what? Let's try the web monitoring—real-time alerts, stay updated. I think that could be a pretty good little basic one to start off with.
Oh, it has notes here to actually explain how it works, so we could set it on a recurring schedule, or we could have it trigger when a user sends a message. I like that you could just combine it into one. I didn't know you could do that! Already learning.
So I'm not going to set this timer up just yet, but we’ll do it when a user sends a message. We want to extract a web page, extract text from web pages and PDFs. I guess we’ll keep it on 3.5 Sonet, and we can have the URL of the web page or a PDF.
I just had an idea: we could literally just have it as, like, a Google search result for latest AI news, right? And then it can just kind of gather all of this! What happens if we just do that for the—set it as the Google search indirectly using Google?
Alright, we'll put that in as a URL. Then it's going to define what pieces of content you're monitoring and paste a Google Doc link to track changes. In this scenario, for example, we're going to need permissions from Google to authorize it to work, but it can work on your behalf. It can work automatically; you just have to give it permissions at least one time.
There we go, we connected the account, and now we just have to provide it with the ID document name or the link to the document. So let's just go make a brand new Google Doc called Lindy News, and we'll just use the name Lindy News. That's the one that I want you to use. There we go, and we'll save that.
Now you can see we actually have a full condition node here. You'll notice that this agent right here actually isn't using the AI agent step like I was using for my previous calendar bot, so yeah, all of these different actions and stuff can be set up as their own node, which can be very useful in certain circumstances.
So the condition here is only continue on if you just extracted the exact web page text from a web page. If you found any new content since the last time you checked it, then continue. Now it's going to update that same exact document.
You see, first it had to browse the web, then get the document, figure out if the document needs to be updated, and now finally update the content.
This can be specified in three different ways: you can either manually set this AI prompt, which the AI essentially will fill out the field using the prompt in the context of the previous steps—basically like an AI agent just figuring it out—or Auto AI will fill out the field using the context of the previous steps.
So the difference between AI prompt and auto is essentially auto will have an AI prompt it for you, while AI prompt is you filling out the prompt for the AI to use, if that makes sense. If you set it manually, it's going to manually put whatever you put in.
I don’t know if I'm explaining that properly; doing my best here. Lots of different terminologies!
Anyways, so it's going to update the document with the content found on the web page and the most recent content essentially. You can actually set it to different modes too for just the Google Doc, so you can either have it replace everything in the doc, prepend, or append it. So either put it first, replace it entirely, or put it last. I'm going to set it to just put it first.
Finally, we have it set to record the changes to a Slack channel that you choose. I don't have any business Slack channels of my own, so I'll probably just remove this. But yeah, you could have it send off to a Slack channel.
Again, I wish it had Discord integration because that would be so cool! Imagine a whole Discord server of just Lindys automating and doing work and communicating! Crazy, right?
Alright, so let's remove the recurring schedule because we're just going to have this thing work when I send it a message for now. But of course, we could have it work on a timer and be completely autonomous.
Let's see if this thing can actually go search for the latest AI news and put it into this doc. Say go!
Alright, it's extracting the web page text. It's getting the document. Okay, it's all successful completion so far. Now, is it going to be able to fill out that document?
Extracted the text from the web page, so it met that condition. Now it's going to go ahead and edit the doc. Oh, it updated the document, or it's updating the document! There you go; it's actually putting the news in.
And there we go—the latest AI news. It actually puts the date, the correct date in, and then gives me some of the latest pieces of news. How freaking cool is that?
That was so awesome! And then you could set it, you know, to email you. You could essentially have this bot email you the latest AI news, tracking the latest news from Google every single day.
That is, oh, it's freaking incredible! I love the AI automation. We’re really starting to enter that era. You could build so many different cool things with this. You could build different entertainment apps; you could build obviously plenty of stuff for business or personal use just to organize your life. This is incredible!
Okay, so at this point, I do just want to kind of go over some of the different actions and stuff that you can have with Lindy. I want to go over some of the more notable ones that I've seen.
So let's go with actions here. With chat, you can either have it send a message to the chat or observe messages from the user chat up until that point. So if you already have a conversation going, you can have it observe those messages and do something with them. Some of these different actions have different splits in the nodes, which is pretty crazy.
So, you know, for this case, it sends a message, and we can have it, after the message is sent, change to another step or, after a reply is received, change to another step. So we can have it do two different things based off of what the user does.
The complexity here—I mean, my brain is racking; my gears are turning. I can see how complicated this could possibly get!
Let’s talk about another one. Obviously, I showed you guys Google Calendar, I showed you Google Docs, but we also have Google Sheets in here. So it can create a spreadsheet for you automatically, track data points for you.
Gmail—this is a huge one! You can have it pretty much do fully automated emails for you. It could respond to every business email you receive. It can evaluate them and determine their importance, and then forward them to you via a text message if you want something like that. If you get a lot of emails, this could be very useful.
It could also filter out all of the spam and the junk that isn't caught by Gmail's automatic filters. And don’t forget, like I said, it can connect to itself! So Lindy has its own stuff: add a Lindy template, delete a current task, modify Lindy memory, search the Lindy store, get a conversation, so you can have pre-made conversations if you want.
We've got HubSpot on here. Again, like I said, Slack, list channels, send channel messages, send direct messages, get messages. You can build a complete Slack bot with this if you really wanted to.
Again, we do have the ability to search the Internet—probably not as good as Perplexity or the new Search GPT that OpenAI has—but essentially, it's like a basic web search, a basic web scraper; still very much useful.
Again, we have the Lindy Mail, so it can send emails for you, which I have been using. They also have the built-in meeting recorder. They've got a bunch of AWS stuff, so this fits more in for the developers, and you know they have a ton of different things that I've never even seen before, to be honest.
I mean, apparently they have DHL for shipping; get tracking details for a shipment! I mean, if you use DHL for your business, let’s say as shipping, you could build a very, very customized bot that can not only place orders but cancel orders for other people and help them out with their orders.
We’ve got Google Tasks: create a task, get a task, update a task. You could probably have this thing work just based off of Google Tasks, so you could remotely operate it from the Google Tasks app. That’s a possibility.
Google Cloud Storage, natural language, Google Translate; we've also got OpenAI in here. Some of these are a little outdated. I've noticed that I've had struggles with the vision prompt, for example.
You can also do custom API calls, but again, I'm not really a Dev, so it's been kind of difficult for me to figure out how to get this thing to work properly. You can also have a basic web hook received as a trigger to begin actions. I think this is a little bit more advanced.
Again, I'm still pretty new to all this, so forgive me if I'm not the master. But I mean, yeah, the triggers are incredible: update a Google Sheet, Gmail, Google Calendar, Lindy Mail, Stripe, Telegram. This thing is malleable, man! It is a very, very versatile platform already.
Alright, so let's edit this. I want to make my web monitor, or essentially my news AI news gatherer bot, a little bit more advanced. So we have extract web page here; I think I would like to make it extract more than just one web page.
So what we could do theoretically, right, is add a step in here. It's going to perform another action and just extract more web pages essentially. So we’ll have it search the Internet, sure!
And again we can have it so the AI automatically prompts this search for latest AI news. Oh, we can change the search engine too! Alright, let’s use—well, this time we’ll use Bing. That’s pretty cool!
This time, we’ll do another extract web page; this time we can use this AI news website, artificialintelligence news.com. And you know what? Let’s do one more: I wanted to search YouTube as well, obviously, because I offer for YouTube AI news search engine.
We’ll set it to YouTube and we'll set the result to video. And we’ll let the AI just search for AI news. Cool! Alright, so now we've hooked it up to some more resources.
Again, we could have this run on a timer and email us or even make tweets about it or something like that, but for now, I just want to be able to trigger it manually. “Find me some news!”
Alright, let's extract some web page text! Let’s do it! So it completed the first one. Now it's going to search the Internet. Oh, it’s having a little bit of trouble searching the Internet.
So yeah, some of these API queries and stuff can have errors, but it usually does a pretty good job at just moving on or figuring out how to fix itself, like it did in this case.
Alright, it completed all of its different searches on all of those platforms including the YouTube one, which is pretty funny. Now it got the document; it met its condition for extracting the web page text next, and now it's going to go ahead and apply it all to the Google Doc.
And there you go! Would you look at that? Updated the Google Doc! So now we've got a plethora—way more latest AI news—and it actually was able to recognize and format it all correctly, which is pretty awesome!
Of course, we could have another Lindy also be able to access this page and then do something else with the AI news. We could also have it be the same Lindy. I mean really, the limits are your imagination with this platform.
Alright, so now that I've kind of given you guys a taste of what's possible with Lindy AI—and this is Lindy 2.0 again with the flowchart and everything, that’s all brand new—this is still very much kind of early access stuff. It's being improved over time. If you see issues, you should report them because it's in development.
Now that you've got a taste, let's talk about how it works in an economical sense because that's going to be important.
So you guys will see, yes, Lindy has plans, and unfortunately for a lot of you, it does work on a credits system. But I do understand it because this is expensive. Using large language models to do automated tasks is not cheap at all.
Of course, you can actually optimize your Lindys to be cheaper and work more efficiently and faster, and that's sort of where the development stuff comes in, right? Easy to use, difficult to make work perfectly efficiently.
So if I open an incognito window so I’m not logged in, we can go ahead and check out pricing. I will say what's good is they do give you a free intro month period, up to 400 credits, or as they say, up to 400 tasks and a 1 million character knowledge base, and you get that just for signing up, which is nice.
They let you get a taste; they let you try it out for free. Always like to see that! But then it goes to different amounts of credits per month.
So, you know, whether you're a business or just a person, I mean I have the $30-a-month plan right now that gives me 3,000 credits or 3,000 actions per month, and we’ll see long term; I think this should be enough to actually use my calendar email bot thing.
But I’m probably going to try to build more automated bots in the future; we'll see how long these 3,000 credits last. Of course, they get replenished every single month.
So the knowledge base indexing goes up to 20 million characters with this. Every million characters after you've got to pay five bucks, and then you also get access to the premium actions, which is like the Lindy Messenger, Lindy Meeting Recording, web hooks, prospecting, and a few other things.
I don't think the pricing is bad; this is easily the most powerful AI automation building platform I've run across. I know Zapier has some AI connections, and I know that's a great automation platform, but I don't think it's going to be as good as something that is from the ground up, AI-first, and actually has access to those incredible AI agents like Lindy.
Yeah, I would love to see what you guys get up to with this. I would love to see what you build, so send me your Lindys! You can actually share them, like I said. You know, I’ll show you how to do that real quick: like, if I wanted to share this web monitor one, I could just simply click the share button in the top right here and click “anyone with link” or I can publish it to the Lindy store.
So yeah, like I would love for you guys to share your Lindys with me! I want to play with them; I want to use them! Share them on my Discord server especially!
I don’t know; it’s a really cool platform. This to me is like the first kind of inklings or beginnings of really being able to build fully autonomous AI apps that can actually do real work for you completely on their own without you needing to constantly have oversight.
You know, check them over and make sure that they’re doing everything properly. It’s going to take a little bit to trust them, of course, but this one’s seriously good. This is seriously the most capable I've ever seen—point blank, period.
Oh, and guys, I did forget to mention one last thing to you. Lindy actually has an iOS app and a desktop app for both Mac and Windows, meaning that pretty much on any of your devices—except for Android just yet—you can access your Lindy on the go or from wherever you are. Especially important for me on my phone, you know, making new calendar entries by just texting Lindy!
So yeah, thank you so much for watching this video, everyone! I hope that this was insightful for you; I hope you learned something! Stay tuned on the channel for more in AI. Peace out! [Music]