The LinkedIn Parasite System (Social Media AI Growth Hack)

hey everybody Nick here in this video I'm going to build out an endtoend AI parasite content system for LinkedIn that uses make.com to Source posts from high quality or very engaged people and then repurposes them with AI into our own tone of voice to help us publish on our own LinkedIn and grow faster this is a viewer request so if you guys have your own systems that you want to see me put together in front of you then drop them down below and if it's one that I haven't done before then I'll absolutely build it

for you without further Ado let's get into the system so I want to be abundantly clear I have actually yet to build this system I haven't put any of the moving pieces together I did like a 30 second search before this just to verify that I could get the data um that you know I could put this together logically but I didn't actually like do any of the the the make.com automation stuff or any of the actual running of set scrapers um the reason why I do this is because I've noticed that a lot of

people on YouTube they will just show you the end result of the system as opposed to what went into building it you know it's almost like if you wanted to become an engineer or an architect and you just like look at a finished building and somebody says okay great now now you know what to do right you have no idea so my goal with my channel is just to show people what like an actual automation developer automation engineer mindset um is while you're actually like going through and developing it so this is the extent of

my work essentially um what I'm thinking of doing is we're going to start by scraping LinkedIn posts and I know that there are variety of scrapers available to do this because I've used a variety of different scraping platforms on LinkedIn and it's worked kind of 50-50 it's worked pretty well next I'm thinking you know we need to store the posts somewhere I don't just want to like take the posts and then you know do the rewriting and then just like forget about them it'd be really useful if you think about it for for future purposes

to just always have this big database of posts that we can refer to um and then the cool benefit there is we can check to see if a post is new as opposed to a post being old so I'm thinking of storing them in a database with the unique ID when I say database here it can be pretty intimidating to some people but it's really just a Google sheet so don't sweat it uh if it's new I'm thinking we're going to use AI to rewrite those posts in our own tone of voice and then assuming

that you know uh We've we've done the rewriting and we're happy with it we'll then post these on our personal or company Pages however we want with whatever Cadence that we want so variety of ways that we can accomplish all of these things I'm going to walk you through my mindset if I were to approach this project like for a client uh or if I were to try and build this up for myself this is exactly what I would do and I'm just going to try and like narrate myself building this so everybody uh is

on the same page everybody understands what's going on and if there are any like minor mistakes that I make along the way any syntax errors or whatever I'll just keep them in the video so you guys can understand how I tackle them and so on and so forth okay so the first we need to do is we need to scrape LinkedIn post now I've actually I've signed up to a variety of LinkedIn post scrapers in the past for um other projects typically what I'll do is I'll go over to appify if you guys are unfamiliar

with appify appify is basically a big marketplace with a bunch of scrapers that people have put together for the purposes of scraping these social media platforms and so anytime I have a need for some social media data or like some some website um data I don't try and build my own scraper from scratch anymore because that takes absolutely forever and it's super ripe with a number of problems that I'm not going to get into um just for the purposes of brevity essentially appify is just this Open Marketplace where you can build something that has value

and then you can just charge people a little bit of money to use it and they can either do so on a per results basis or uh uh like a monthly basis or like a server usage basis or whatever but it ends up being quite cheap on a per record basis um personally so that's what I go with now what we're going to do for this is I'm just going to type in LinkedIn post up here I've already made an account which is why I have my little profile picture here so if you haven't done

that yet just go through the rigor moral sign up they'll give you $5 in free credit so you can test out this flow and I see a variety of different ones here I've used this one in the past and I verified that this worked uh so this is probably the one that we're going to be going with there's a LinkedIn post information scraper LinkedIn post reaction scraper post like scraper post reaction scraper basically what I want is I want something that scrapes posts with the post copy right like I want the text of the post

because I'm going to take that text and then I'm going to use it to rewrite and then pump it into our make.com scenario um I was just on LinkedIn here browsing through a variety of people's profiles Mato just some guy that I follow and he follows me back probably because of YouTube or maybe he's in my community um but this is a good example of the sort of thing that I would want to do I would basically want to get content from people like this I would then want to um you know repurpose this Con

you know like take the copy here as long as it is we want to be able to get the full Copy Text uh and then we're going to like rewrite a little bit make it make it our own and then we're going to want to post it so we're going to need something that delivers the text we're also going to need something like it would be cool if we' got the image although now that I'm thinking about it if we do get the image it's going to be kind of like imagine you um doing this

ostensibly super unethical thing and they just using the exact same image as the person you know ideally we'd be able to modify the images a little bit I don't know if we could necessarily do that with AI like there's just such a wide problem area so no I'm probably going to keep the images out of there um it's a shame that images that are super engagement tend to do well another thing that's coming to me just as I'm I'm running through this is if you think about it probably we're going to need to have a

filter step before so it's not just that it's new it's we scrape LinkedIn posts then we determine if these are like evergreen or personal the reason why is because let's say somebody goes on LinkedIn and then just makes a post about like a a coffee chat that they had with a friend of theirs we don't want to repurpose that and say yeah had a great coffee chat with Peter because that'll just be a waste of content obviously we don't know Peter and there's no real purpose there so um we're going to have to do some

sort of AI filtering there too which uh which I'll cover but yeah so I think we have more or less everything we need we'll use Google Sheets for this we'll use AI for this this is going to require a little bit of prompt finagling um but you know anyway I think you guys are tired of just seeing me scope the project so let's actually go and do it to be to be honest I would not spend more than 5 or 10 minutes doing the scoping um on like a real client project or my own just

because I typic Ty Al learn a lot while I do the thing but anyway let's start with getting the data from our data source so I'm going to go LinkedIn post scraper and I just have a little window here I'm going to have to move to the top right uh this is going to ask us a couple of things it's going to ask us for cookies looks like so we're going to need to grab some cookies here the way that you do this on appify is you tend to use a little browser extension called the

cookie editor which I've already downloaded so if you guys don't have the cookie editor um if you just Mouse over this question mark probably yeah it includes a little direct link to some service and the way that this works is you go on LinkedIn you go to your cookie editor you just click export as Json so I'm going to go back here and I'm going to paste this puppy in and now we have all of our cookies inside um I don't know who this is probably just some guy I was scraping posts for before yes

so I'm just going to run this as a as an example thank you very much R and then we can set a limit of the number of um posts that we want to scrape from The Source uh looks like we also have the ability to scrape some additional information I'm not really sure what the hell that means likes comments screw it why the hell not we can also get the raw data I don't know what that means but I'm just going to run this thing first before I make any more assumptions so let's click save

and start and let's just let this puppy go the idea here is I'm going to let it run wild um which actually now that I'm thinking about it may not be good because how many posts is it going to scrape a ah whatever uh we see that the LinkedIn account was successfully detected so basically just took my cookies and then used that to get in uh and if we go to Output here we'll actually see what it's what it's generating for us so I'm just going to preview this in a new tab just see all

of the data um oh sorry right over here so looks like yeah okay so it looks like um if it's an article then it has a bunch of additional information and then there's some information around the background image of the person okay it looks like there are a variety of different post Types on link so let's just see the post copy where's the post copy here okay right over here it's in this text field so right now we're using a they can do specific things like search the web or drive the car this is called

narrow a the next step is make a they can think and learn like a human this is called General a perfect so this is exactly the sort of thing that we'd want to repurpose so I can actually use this as like a test case beautiful beautiful so um this is this is everything that we need in order to do good good work uh I guess we could also check to see if there is an image if there is an image we could pass that through Ai and then we could ask you know is this image

relevant to us if it is then great if not then not the end of the world um the most important thing when you're doing any sort of appi scraping is just go to storage and what you want is you want the data set ID you just want to copy that over now in our make scenario I'm just going to type in appify now what I want to do is I want to get the um data set items down here the reason why I'm doing it this way and I just have to select the write account

because I know I have a few is this just allows me to pull all of that data for the flow that I just ran so I'm hardcoding the data set ID I'm feeding it into this module called get data set items and then I'm requesting it to give me all of this data the value here is you know we just took the data from appify and now it's in make if you haven't done you know the connection and stuff like that just remember you're going to have to do that as well super easy just a

little API key um I'll leave that out for brevity but yeah we've basically accomplished the first major purpose in my mind which is getting the data to make so now that the data is in make we can actually do stuff with it right so the very first thing I'm going to want to do is I'm just going to want to go through these fields soort of mentally here and I'm seeing that there's like there's different types it looks like there's like a short string and then there's a long string so I guess after a certain

character count make just um makes this a little toggle field instead but yeah this looks pretty good are there any images here looking for images I'm not seeing any images maybe attributes no h I'm seeing profile pictures which is cool but you know I don't really think we can do anything with the profile picture can we reactions comments looks like this scraper is also scraping for reactions and comments and stuff which is cool now this guy that I'm doing the scraping on he's not like super famous or anything I imagine you know if he was

super famous the number of shares likes comments and stuff that' be like a good piece of data for us we could use that okay cool um anyway I think we have more or less everything we need so what I'm going to do is I'm just going to start building out uh the GPT prompts so you know I uh I can verify whether or not this is a general post or this is to something specific or or particular so like I basically want to verify that this is going to be Evergreen right so let's do that

let's go to GPT and the way that I always build these things out is I'm just building them one module at a time the reason why I'm building them one module at a time is I just want to make sure that whatever I do it is um I you know if there was to be an issue with this flow I would know exactly where it broke if I don't do it this way you know I Tred just like slamming slapping like 15 mod modes on the page and then trying to test it well then I

don't know it could be any one of the 15 modules but if I test every module one by one then I know exactly where the error is when the error happens because I know it wasn't in any of the previous modules obviously the one that I'm screwing around with now so I'm just going to have two steps I'll have like a simple um filter post step and this is just going to be like a dumber sort of model it's going to be simpler so I'm just going to scroll down to model we'll just do GPT

40 I say dber meanwhile it's like smarter than my hat um the system prompt I'm going to use here just off the top of my head you are you're a helpful intelligent assistant I would just say writing assistant and then in order to set a good prompt um usually I'll do I'll do this structure I'll do system first and then that'll just Define what the prompt or what the am model like identifies as and then I'll do a user prompt after where I give it the instructions and then after I'll give it at least one

user user assistant pair so it'll be system user user assistant and then after that I will put another assistant uh sorry another user prompt where I actually ask it for the data the best way to build these flows out is usually to treat the AI as just like an API endpoint so I'm going to be sending it some Json JavaScript object notation and I'm going to be getting some Json back so I know that for sure um and if I click on show advanced settings there's actually now an option for this under response format where

you can actually just ask for stuff in Json treat it like an API call then have it return data to you but how are you actually going to do this I'll say your task is is to identify whether or not a post is Evergreen or personal Evergreen posts are about um general concepts ideas um facts movements opinions statements Etc personal posts are about are about uh something personal posts are about a specific person or about an event in the posters life or about something okay I probably go with that A good rule of thumb is

if the post discusses something that is relevant only to the poster it is personal if the post discusses a broader idea concept or subject it is Evergreen return your result in Json using this format and then I will go um ver let's just call it verdict personal okay cool that looks good to me I'm then going to go and provide a user um example I'm going to say post text or actually let's just go post and then why don't we actually go and like find a personal thing so I'm going go back on this fella

um agents the cold Callum what's your favorite combination of channels stay tuned this is him and a friend okay great oh beautiful beautiful cool so this is a great example of a personal post that I would not want um duplicated right so we're just going to paste all of this in and I'm realizing now that a lot of these are going to have tons of new lines so what I'm going to do when I feed it in is I'm actually going to remove all the new lines this is also just a good idea in general

when uh you know what I could just feed this in feed this into Ai and have it do it all for me and then um yeah the the benefit in removing the new lines is you also have slightly shorter um like fewer tokens so I'm going to go post and then I'll go assistant then I'll go verdict and I'll go personal and then um let's go user one more time and then what I want to do is now I want to provide an actual post so I'm just going to go here to where it say

post text then I'm going to close that off and then yeah this looks pretty good to me I don't see any major problems here so we're just going to give it a try um the first thing I have to do though is I just have to head back to my get data set items and I just have to set the limit to one cuz I only want to do this on one post basically I just want to see how it performs on real data okay result was Evergreen awesome what was the text that we fed

in there let's go to messages yeah this is basically exactly what I wanted looks like it actually even worked with the new lines which is cool the reason why is because we're just feeding this into GPT so this is this wouldn't actually be like a good like um a properly formatted API call because there's a bunch of new lines in here but because we're just feeding into GPT as like content it doesn't really matter uh we could we could fix this so we could make this better and that's what I'm going to do just for

token brevity purposes just to keep the contact as short as possible and then just make sure everything's formatted in the right way I'll just go replace then I'm just going to go um grab all the new lines and I'm just going to replace all the new lines with um let's just go spaces instead and then something isn't working here why oh yeah okay great now I'm going to do the same thing it's going to give me the result probably much faster then I'm just going to check the content nice that looks pretty good to me

no major problems here at all so we now have basically a working filter this filter I've only tested it on one result but uh just based off my intuition of how these models work it's probably going to do pretty well I'm not saying that I have the best prompt ever but it's a pretty good prompt and I'm going to make sure to you know paste it over so you guys can just copy this in your own flows whenever you want so now we have a way to filter the post kind of the next step in

my mind is we need to add this somewhere right we need to um add this to a Google sheet U preferably so I'm going to go over to uh uh let's go sheets. new let's actually put together a spreadsheet that we can use for this purpose so I'm go add a row what I have to do first is I actually have to like just connect this we'll go LinkedIn post and this will just be our like our LinkedIn post database and let's think about this really quickly first we're just going to want to make sure

that this verdict is equal to Evergreen before proceeding and then everything that makes it in here is going to dump directly in here which means that now we need a bunch of um headers in our spreadsheet or in our database I should say like you know spreadsheet database same difference to be honest and yeah I can already see a bunch of valuable Fields here so um I could just map all these myself um and I have done this before but it's actually a lot easier if you just um basically get AI to do all of

this stuff for you so I'm just going to paste all of this in uh sorry hold on there might be some additional data that that's relevant to us I'm just going to go on download output bundles and you see here we basically have all the data right so I don't really care about the author I don't really care about the author well actually author profile ID yeah no I don't really care about that it's not super relevant to me so let's just do okay let's just get AI to turn this into things for us like

turn this into headings for us so I'm just going to copy all the stuff over and then I'm gonna open up my AI model which is just on my on my browser and I'll just say into um this this is Json that I'm putting into that I'm building a spreadsheet for um take all of the key names and turn this into list of headers in CSV then I'm going to paste this in and then I'm just going to remove the stuff that isn't really relevant the stuff that I don't really want um so I don't

want all this author information for instance so the reactions to be honest I don't really care yeah I don't want to like store every reaction uh and it looks like there are a bunch of reactions here let just take all those out and then author um I don't really care about that except for the first and the last name I would say so oh actually we can just get the profile ID the profile ID is is good enough get the profile picture sure um and then what else author profile you post it okay awesome so

I'm just going to pump that in it's now going to basically take this and generate just a list of headings that I can just paste into my Google sheet so I'll go back here and then what I need to do in order to split this into all of my headings is I just need to go split text to columns here and now I have a list of headings which I can just bold and then you know kind of organized in this way this looks pretty good to me I have most of what I want already

which is nice and then all I need to is I just need to make this um accessible to my other account that's actually going to be managing this so I'm just going to call myself an editor share this with me and then I'm going to go back here and now we're actually going to try adding some of the data into our spreadsheet which is pretty cool I basically use AI to do all of these little things now um I used to do it myself but if you just do the math on it uh yeah you

can save quite a bit sorry I need to go I think with shared with me because I technically shared this one I didn't make it inside of that account so I'm going to go shared with me and then I'll go LinkedIn posts perfect and now we're going to go to sheet one inside of sheet one we're going to have a bunch of fields now we just have to do the particular laborious thing of just matching them all so let's go over here and let's just pump it in um is repost Ur this is going to

be our this is going to be our unique ID actually so now now that I'm thinking about it we should probably change this let's just drag this over here and then this is going to be here um this doesn't work like this says is repost but it's actually just column A so I know Ur n is going to be in column A is repost is going to be column B from here on out we can just proceed and what I'm doing is I'm just pressing tab here and then I'm just clicking on the thing that

I want um this is just going to make it substantially faster attributes and comments probably technically won't work to be honest because these are nested but uh that's okay um can post comments can share commenting disabled allow commenters root share share audience other profile ID that's what I want so here here here here here here here and then post it at post it at ISO and then input your awesome so now we we're going to get all this data and it's going to pump into our spreadsheet so I'm going to run this it's Evergreen and

voila now the reason this is bolded is just because um you know the first header row was bolded and it's just going to do the same thing that it sees I don't like how this is spread on all these lines so I'm just going to go into my Google sheet and then I'll just go clip so I don't want this to be wrapping then I'm also um hold on a second why is this so annoying and big I think I need to do this I don't like it okay there we go uh which is nice

and then really thing that matters for us is this Ur n Li activity this is going to be like our our unique thing when you see you when I see UR and I'm thinking Universal Resource number or something um earn meaning in code Universal resource name permanent identifier right which is basically exactly what we want okay so now we have um our database but let's just make this a little bit prettier because I like when this is prettier nice nice and blue you know what let's let's change the text enter that's good okay great um

and what do we have we have this text column which is the thing that we really want this is what we're going to be using to like generate our data so uh we can do this in a variety of ways we could do you know what we should probably do we should store okay now I'm thinking in my head I want this to be like a good future solution future proof I want us to be able to use this for as long as humanly possible so if I wanted to make this really really future proof

I would probably store the outgoing posts as well as the incoming posts what I mean by that is this might be like the source posts or scraped posts and then this might be like the generated posts or something because what we're going to do is now now that we have this data we can feed this in AI we can have it rewrite it what we can do is we can add it to generated posts as well we're going to need different um headers than scrap posts and we're also going to have to change the sheet

name here because right now it's sheet one but what we needed to say is we needed to say scraped posts otherwise our automation will break um they use the sheet name in the API call uh and then what we're probably going to want to do is we're just going to want to grab the text of it and then we're going to pump it in generated post alongside this Ur just in case we ever need to reference the two back and then I'm just thinking what else I think that's all to be honest you know maybe

we'll do user as well yeah so let's do this let's um let's grab all of these headings then I'll go Ur and then we'll go text and then we don't need any of these because these are all likes and comments and stuff we can get rid of all this stuff all this stuff let's just keep the author information profile ID profile picture I mean we don't really need the profile picture do we I do I do like having the profile ID though the idea is just we could very quickly and easily reference these between each

other uh profile URL sure author you are inure post it at timestamp post it at ISO cool we'll just go here paste these in and then let's just turn these other ones from Blue back to White awesome okay great now we have basically everything that we need in order to actually go and have ai generate the post for us and this is really where the magic is going to happen so why don't we do this we'll go Evergreen first of all oh you know what I'm realizing now I think I missed the step didn't I

yeah because we need to we need to store them in a database with the UN and then double check to see whether or not these are actually unique every time we add them see I wouldn't have realized this had I not done this so here's what I'm going to do right before all this and this is going to save me some tokens I'm going to go search rows and before we even do the AI filtering we're going to just check our database to make sure that this doesn't actually exist there's a quick and easy way

to check databases in make.com um H I guess would it be my drive I don't know I may not be able to search things that aren't technically ours uh what was this called LinkedIn posts yeah okay yeah we can and then we want to we want to check scraped post so here's the quick and easy way to do it all we do is it gives us columns that we can filter by so we check to see if Ur is equal to this and if it's the same post as before it's not going to proceed okay

so basically what we're doing is we are filtering all of the rows that we are returning and then I'm just going to check to see if this returns more than one bundle if not if it's zero then we can proceed if it's not then we have to stop so let me show you guys what this looks like in practice we're just grabbing the same data over and over and over again with the same um Ur here so because of that we're running this through our Google sheet filter step we're checking hey does any of the

UR NS equal to the new Ur that we're pumping in since our sheet currently already has one that does it's going to return more than one bundle and our flow is only going to proceed if we have zero bundles or more so I'm going to click run once voila and the total number of bundles is actually one because the total number of bundles is one um it's not proceeding beautiful beautiful so now we we've actually accomplished sorry um basically all these steps so this is good this is good this is good and now we're on

to this one so if NE we use a to those posts in our own ton of voice let's go back here and let's go rewrite post let me just drink from my cherry bubbly I can tell all of the carbon dioxide is currently being wasted these things are goddamn incredible do you guys haven't tried this on your own o man bubblies it's like zero calories not that I give any shits to be honest but zero calories it's just like um it's like the oil from an orang orang peel and then yeah helpful intelligent writing assistant

uh and then it like satisfies that craving that I always have for that that hit of something okay so now it's time for us to design our prompt I'm going to be using the same prompt design principles that I used before where I had the system prompt up at the top your helpful intelligent assistant that just defines it the next one is going to be basically just telling it what I wanted to do what I wanted to do is I wanted to rewrite some posts for me in order for prompts to be in order for

uh yeah prompts to perform as well as humanly possible generally speaking what you need to do is you need to actually take the data that you want to transform and actually do one of these Transformations at least one time so it picks up on your tone of voice picks up on what you're trying to do AI is never going to be perfect so this is not going to be like the most amazing Post in the whole wide world but if you follow along with a couple of the little tweaks that I'm going to make here

it will get you 80 to 90% of the way and a lot of the time that is sufficient so let's say your task is to write a post based on the text in another post you should be heavily inspired um I'm just going to call I'm not going to say copy or rewrite because in my experience when I've used the terms copy or rewrite tend is do a really job so instead I'm going to say hey you should write a new post based on the text and Sample post you should be heavily spired by the

sample post and your output should be of similar length to the input okay great now I'm going to go over to user and assistant the issue is return your post in Json using this format let's go replace U lines with backs slash n this is just for us so we'll go uh post post goes here and then I'm going to actually um the issue with writing all this stuff ourselves obviously is going to take forever so what I actually do is I get AI to write me the first example and and then I just make

a bunch of tweaks the example so I'm just going to go post and then I'm just going to feed in the text from um you know the I guess module number three get data set items here now I'm noticing that I'm going I keep on trying to do the same thing I basically am trying to feed in the same variable that I was feeding in here with the replace and stuff like that I don't really want to do that I don't want to be in the habit of just like constantly feeding in the exact same

thing over and over and over again generally speaking when I notice that I do something at least once I will just copy this and then I'll just add it to a set variables module instead so I'm just going to go over here just going to go set multiple variables and I'm going to stick that puppy in here instead what I'm going to do is I will say uh sanitized post and I'm just going to feed this in so I'm going to replace all the Texs with a new line you'll notice that my filter is now

not really in the right place I don't want this to run three times before it determines whether or not it's new I only want it to run twice uh because I could say one and a half Ops every run right so I'm actually just going to replace this filter I go back here I'll go new then I'll just make sure that's equal to zero and then I'm just going to delete this make.com currently has a little bug where you can't see the text unless you wiggle this around a bit so that's where that's coming from

awesome um and now we'll just say sanitize text and now instead of the variable that I was using before I'm just going to use the variable from this module and this does technically add operations but it makes the flow more maintainable so instead of me um having you know if I wanted to change this and I forgot I have to go through every single one of these modules and double check and be rigorous and then forget half the time if I do an update um I I know now that like hm there's a purple module

here let's say 3 months from now H there's a purple module here well I let me check the variable let me see what's going on oh okay this is what's grabbing the text so I need to you know if I'm going to make any changes to the post copy I'm going to do it right over here okay great so we have the rewriting um why don't we actually run this through now and yeah let's just see how this goes uh what I want to do is I want to I just want to see how it

rewrites on its very first run and then by knowing that I'll be able to tell uh you know I'll be able to basically make my changes and stuff what I'm going to do is I'm going to change a limit from 1 to two the reason why is because I wanted to go through one more and I want it to add an extra new line for me so I'm going click run once cool so it stopped the first run and just went through all of the steps on the second and you can see that you know

we now have the same thing we have another row added with its own Universal Resource number which is beautiful and we actually went through and we we rewrote the text okay great so let's see the source text let's see this and let's just compare them show your resume to 100 people and you'll get 100 different opinions at Ai and applicant track system and feels even harder it's a lot right but in the end a person will hire you do not rewrite your resume 100 times the resum is one job to get you notice and help

you get an interview hash resume this one to share your business idea with 100 people and each will have a different perspective toss and market trends and Technology shifts and it could seem daunting overwhelming isn't it yet ultimately it's the customer who validate your idea instead of tweaking your plan inly focus on its core purpose to solve a problem and catch the interest of your target audience okay great now that we have this I can actually take this in here and I can use it as an example okay so uh I'm going to add an

assistant call here and then I will just say let's say sample post for future reference I'm going to paste this in and then this sample post actually needs to be this beautiful cool so now I have U basically my first system prompt my user prompt my other user prompt and then my assistant prompt so sort of like the last thing I actually need and I'm I'm going to go through and I'm going to touch so I'm going to make it a lot better but the last thing I just need is I I just need you

know the actual data itself if you think about it which would be uh this variable over here sanitize post now you know it's going to read through the text that I'm going to touch up and make better in a second it's going to get like the idea of the tone of voice and stuff like that and then it's going to use it to write um a response to this which is beautiful okay great so now that we've done this I'm just going to click okay just to save that because I've had a nasty tendency of

um exiting out of prompts that I've worked a very long time on and I'm just going to go through this and I'm just going to use my own prompt engineering sort of like best principles um and then try and make this as high quality as humanly possible now keep in mind I'm like I have a writing background and so I used to do freelance writing it's actually I got started in the AI space I uh you know was doing some freelance writing right around 2020 when gpt3 came out I was like could I automate my

job and the answer to that ended up being mostly so like I'm very familiar with how this stuff works and a lot of things probably come intuitively if something doesn't um if you have any questions about what I'm doing or or maybe I don't fully explain myself just leave me a comment down below I'm W happy to walk you through it but okay let's actually run through and do some of this editing so this is the sample post this is the so basically what I'm seeing that it's doing is it's changing resume for business ID

I don't want that I wanted to talk about resume so if you shared your resume with 50 people each would have a different perspective then consider market trends and Technology shifts then when you consider market trends and Technology shifts it can get maybe maybe we should say add market trends shifts and it gets even more daunting I'm just going to remove the overwhelming isn't it one thing to note ultimately it's always the customer who validates so I guess instead of customer what we want is who is it the [Music] person oh sorry hold on a

second I'm actually out of the loop here each would have a different okay if you shared your resume with 50 people you'd get 50 if you asked for feedback if you ask for resume feedback from 50 people you get 50 different perspectives let's go ultimately uh if you ask for resume you 50 people you'd get 50 different perspectives there are tons of conflicting there are many conflicting beliefs and not all of them are valid ultimately you only need one person instead of trying to satisfy every Everybody understand that ultimately you only need one person to

like your approach in order to get hired that looks pretty good to me I I think this is a good inspiration it's on the same subject but it's not just recopying it and pasting it um so that's pretty solid you know some of these posts are going to be pretty timely they're going to be on like events that are really really hyper relevant maybe some like news thing if you just copy and pasted the same thing you you get in a little bit of trouble but if you just took slightly different angles to say the

same concept I would say um then that'll generally be a lot uh a lot better of a strategy than just copying and pasting same thing so show your resume to 100 people you get 100 different opinions at a and applicant tracking systems and it feels even harder it's a lot right but in the end a person will hire you don't rewrite resume 100 times your resume is one job to get you noticed and help you get an interview if you ask for resume feedback for 50 people you get 50 different perspectives there are many conflicting

beliefs not all them are valid instead of trying to satisfy everybody understand that ultimately you need only one person to like your approach in order to get hired right for that one person cool looks pretty good to me um I'm going to go down here and I'm just going to decrease my temperature a little bit on the am model I found that that tends to improve prove the output I'm not going to muck around with any of these other settings I'm just going to click okay and then what I'm going to do is I'm going

to go back here and I'm just going to no we need to do one more actually we need to do um three instead of two so I'm going go here I'll do a limit of three and then I'm just going to make sure that I've set everything up right for this final test and based off this test I'll try giving it some higher level instructions oh let me do one more thing 20 of voice guidelines write simply write with simpler shorter sentences be Spartan in [Music] tone shorter straightforward sentences let's do a b Spartan in

tone use what are these called asterisks for emphasis that's sort of like how I write use use asteris for emphasis if necessary the issue with writing this is it's going to include asteris in every single post maybe I'll say flip a coin the coin comes up heads use asteris from emphasis if not if not don't we're having to run a little simulation here it's not actually going to flip a freaking coin it's just I don't want it to write with the asterisks every single time and let's see what the post result is so the message

that came in was before asking for a referral pause and think why would somebody refer you is they seen you work do they know you have you worked they trust you recently receive Val advice from experienced professionals blah blah blah blah okay cool and then the result before you ask for a referral stop and consider why would they refer you do they know your work have you collaborated for is there trust I recently learned from season Pros about the power of jine relationships a referral isn't a simple ask it's a testament your abilities and characters

here's my strategy building real connections showcase your work and let it make an in Impact engage sinere with others add value every time solid connections grounded in trust and respect focus on these and referrals will come naturally this looks pretty reasonable um I don't like how there aren't new lines so I'm just going to go back and just do one little tweak here where I have a couple of new lines jumping in we'll go [Music] NN then we'll go NN cool and now if I were to delete this and then run the same thing on

the three posts it'll grab that final result and it'll rewrite it but it should put in some new lines because my example contained new lines and it did right so voila we have the new lines which is cool I don't really like how it's parsing the freaking new lines uh ah whatever okay so now that we have this um I just want to update this to the database for generated posts and then once it's updated in the database for generated post I'm just going to go through and then um we'll actually like hook this up

to LinkedIn and then pump this out and then I'll just set uh this appify Runner or sorry I'll set a schedule in appify to have this run every so often and then maybe we can just get the data set items I don't know like every 3 hours or something and that way I'm just going to be posting stuff every 3 hours that you know somebody has posted previously in in the last 3 hours if I assemble a list of very high Roi um LinkedIn channels or LinkedIn Pages LinkedIn people I should say because we're looking

for people we're not actually looking for pages I also find that Pages tend to have lower quality content then we can realistically have a post come in like every few minutes which is pretty powerful so I'm just going to go back to our spreadsheet and I'm going to go from scrape posts to generated posts and then the values are going to still be the Ur So this is going to be the UR of the source post so why don't I go like Source Ur and then the text is going to be this right here author

profile ID it's going to be this sorry there's two author profile IDs where are there H we don't actually need that did I include the same thing over here no looks like I just did one must have just screwed up when I created this okay uh so C is going to be auor profile d d is going to be author name let's go author name so that'll be here then the next one will be author profile Ur on it's kind of annoying that I have to scroll down every time sorry author profile URL I should

say there you go this one is going to be the UR n and this one's going to be the timestamp iso and then input URL I'm just going to delete all of the rest of these because they're not relevant to us boy was this a lot of data awesome looks good and now I should have basically everything I need in order to go and actually post it um in order to do LinkedIn posts variety of different ways to do so um you do need to make a connection with your LinkedIn that's kind of like the

big um hurdle for a lot of people so I'm going to go create a user text post then I have to well I'm not actually going to like do the whole connection myself I don't think I mean maybe I will I don't know I just don't want to like show any API details here if it's just a simple ooth it should be pretty straightforward you know what I probably need to change my password too oh actually okay yeah that looks like it worked okay great uh and then for Content what I need to do is

I just need to put the actual Post in so let me just see to escape reserved elements use the backslash or set of the IML function replace with search string so it looks like there are some elements that we cannot include here and I think one of them is a bracket it looks like we can tag people that's really interesting we could totally use a to do that now that we have the um yourn identifiers wow anyway um we could also make an article I'm just going to do only my let's do only my connections

because I don't just want to be pumping this out to many people yeah and yeah I don't want anybody to see this because it's just going to be a test I'm just going none um and then we're just going to give this a try see how it goes let's go back to my own LinkedIn nice that it was just um oo eh a lot of the time you got to do much more than ooth for this it's super annoying and then I just need to find where my posts are I don't think I've actually made

a LinkedIn Post in a really long time I see some featured stuff here oh activity yeah sorry it's right over here so let's do this and then let's actually give this a run in order to do this we are going to have to go one more Let Go limit four might be a personal post I don't know it's an evergreen perfect we're going through our filters and I'm making the post so now I'm just going to refresh this puppy and then I'll go to post and then I'll go back to my own profile do walk

in interview still happen yes they do the special place H world that might seem outdated cool I think this is more or less the same thing that this guy wrote just rewarded to be honest a little bit better um I think our English is probably just uh a little Cut Above which is actually kind of scary that you could create a campaign like this and then just proceed to absolutely MOG the guy that's doing the content um I think it'd be cool if we had some new lines here so I'm kind of thinking like hm

how are we going to add some new lines looks like we did post content but I don't know if the new lines worked so let's go back here let's refresh us puppy no the the new lines definitely aren't working let me delete that that's embarrassing you just know there's some people that are like looking at these being like Nick is really screwing up his automations here uh yeah there has to be a way that we can get these new lines in here I don't really understand why ah okay cool that one worked so before I

ask for referral pause and think why would some refer okay cool cool cool I figured it out so uh what we have to do in order to make this work after some handy dandy debugging was the new line characters that come in here need to be replaced with a new line from make.com because there's some string formatting thing that it does before it sends the request out to strip all of these so if it doesn't have a new line it's not going to work um so basically what we've done here is we you just changed

the text so that instead of just result post we're replacing all instances of the new line and result post with new line which will make things just look a little better um which should solve more or less all of our our concerns now here's another issue um if you think about it the way that this is going to work right now is it's just going to scrape everything and the second that it scrapes the post it's just going to make the post so we're going to be creating like 50 user text posts every time we

receive 50 records we don't actually want to do that right what we want to do instead now that I'm thinking about it is we want to add away to like schedule a post um and so the way that I'm I'm going to I'm going to create my scheduling feature is I'll go to generated post and then I'll just go Um scheduled post and then I'm also have a column here called posted status maybe I'll go like post status what scheduled post is going to do is we're just going to create another scenario and that scenario

is just going to look through this scenario every now and then basically every hour or so to see if it is time for us to do the post if it is time for us to do the post we're going to grab the data from here do the posting math and then we're going to update the post status um this way we basically have like a closed loop and everything in our system is like self self-managing so uh I'm going to go for input URL here sorry I'm going to go scheduled post to J so what

I'm going to do is we need to have some some way of scheduling a Time hm so I'm thinking about how we're going to do the time scheduling um we might need I ideally I want this to run automatically so what I want it to do is I want it to run completely autonomously and I wanted to distribute the posts so that I never have to worry about doubling up on a post I never have to worry about having multiple so what we're going to do is I'm G to use this as an example I'm

going to go to scheduled post I'll go uh let's say December 8th 2024 and then we want like um 14 this should be a format that It'll recognize as a date so I'm going to go to format number we'll go date um no we need a date time sorry uh date time there we go and then what the hell's the format for the daytime uh I go back here it says 926 2008 um 15590 so right now would be 1208 2024 1208 2024 4 um I want 14000000 is that right yeah okay that's right so

if I wanted to make this 15 then I would just stretch this okay cool all right so basically what I'm going to have this do is I'm going to have it schedule a post every hour the way that it's going to do the hour scheduling is it's going to return all of the posts on the spreadsheet that have a date later than write the second so I'm going to sort those to get the latest one it's just going to add an additional hour to that and then schedule it that's how our schedule is going to

work now in order to actually make this work obviously we need to make sure that whatever we're providing is in the exact format that um you know the thing needs so uh what we have now is we have month month so I'm going have to parse a date basically okay actually there are a couple more steps that need to occur before I do this what I'm going to do is I'm going to separate this and I'm going to add another scenario here uh first I have to update the status right so post status let's just

have a couple of different post statuses one will be um let's just say uh yes we'll just say scheduled for now okay yeah we'll say scheduled this will be called schedule post add Source post schedule generated post um and then what we need to do is we just need to search the rows one more time now what I need to do is now I need a filter that says and we need to refresh this now because we're not getting all the fields we want so I'm just going to go back to LinkedIn posts and just

run my way top to bottom uh sorry scraped post is not we want generated post and then I'm going to scrape this top to bottom for uh SK post it's date time operator I believe and we want it later than I think we can just say now I may be mistaken we may not be able to say now but I'm I'm just going to try and say later than now I'm going to run this see what happens um we get one right which is obviously the one that's for 2 p.m. so why don't we make

just a couple as tests here we'll do one that's in the past do one that's 16 and then we'll do one that's 18 let's go back here now and let's see if we could just get all of the rows and we got this one which is scheduled for this which is nice then scheduled 16 which is nice 18 okay cool oh I I screwed up I should uh we didn't get the one that was 10 right just to clarify 14 16 perfect perfect so the system is now working well uh what we need to do

is we just need to sort these now any way to sort these automatically no I'm just going to aggregate these into an array first and then sort them based off of that date time so I'm going to go um aggregator because we have multiple bundles here it's opening bundle one bundle two bundle three if I don't aggregate these results every module after this is going to run the same number of times as there are bundles so if there three bundles this will run three times that'll run three times there are 20 it'll run 20 times

to eliminate this I need to select this multiple bundle structure then I'm just going to um map all of these fields and put them in an array instead the benefit to sorting based off of an array too is now I just go back here I just need to test out this specific flow if I run this now this is going to Output um array with three bundle or sorry three elements then we'll have the schedule post as one of those variables then I just want to set a variable and then I just want to call

it sorted scheduled posts then I want to sort this array and what I want to sort it on is I want to sort it based off of this value um where it's a scheduled post here which as you see if I Mouse over this there's this little raw thing that's popping up that raw that nine that's actually the key name so I want to post this based off the key name I just want to see what happens I don't actually know this is going to work so we're just going to give it a try Okay

so this one was 14:00 this one was 16:00 this one was 18:00 and this is how they were initially formatted so I I want to just change these a little bit want to say um 15 19 let's do 16 first so 16 15 19 should be 15 16 19 so we're gonna run this one more time we're gonna see whether or not the sort actually worked okay it worked 15 16 and 19 wonderful and now that we have this what we can do is we can just take the latest date and then we can just

add an hour to it and that's how we're going to schedule our post that way everything that comes in is always going to be one hour later than the latest Future post essentially um oh boy actually hold on a second now I'm seeing an issue the issue is we have have uh turned our things into an array which now means that these elements are no longer accessible to us which kind of sucks H so if we're going to want to schedule our post we're going to need to do so either earlier which wouldn't make any

sense or we're going to have to do so um we're going to have to pull out another iterator the data that we would iterate off of two probably be this data here right so I'm actually going have to pull another iterator now because I kind of screwed myself up with the structure let's drag that puppy there the array that I could feed in um oh hold on a second no no no I lied I lied my bad now you can you can totally do this I just didn't um have the variable on the right module

if we connect these now and I go back here now we should still have this successful okay perfect yes we do and then this is now going to be with status scheduled perfect what we need is we just need a time so I could set the time through another set multiple variables but I think I'm just going to do it here I'll go um actually yeah let's do the let's do the set multiple variables one more time this is going to be sort um scheduled times and this one here is going to be calculate new

scheduled time and then what I'm going to do in order to do this is I'm going to grab the last element of this array which is going to have the um excuse me the latest time so we'll go latest scheduled post I'm going to grab the last element in this array if I just show you guys what this looks like as a test one more time we're going to drag this back over over here and then run this puppy and now I have the last scheduled post and what we want is we just want the

time so I'm just going to get and then what I want is I want the name of this so let's see here oh this is going to be annoying because actually the name is also just a number I think I can get nine it should work if not then we're screwed I'm do a different approach here okay no we did now what I want to do is I just want to H could we parse date um the date format is going to be oh boy uh mm SL oh hold on a second DD seeing here

that this is just an eight that's not good MMD d y y y y and then hour hour minute minute and then second second I think maybe it's XX I don't know let go integrat datetime operator go to the tokens here and we just want the seconds so how the hell do we format seconds again I think it's SS that would make sense uh it's HH for 24-Hour time good SS with leading zero start actually we're going to have to go lowercase so [Music] mm yeah so M HH Capital HH mmss like this knowing that

we have to do it that way but whatever so basically what we're doing is we're we're getting the date we're parsing it to get this in makes um built-in datetime format which has it in our built-in datetime format and then what we're going to do is we're just going to format the date after we've done this math oh but first we have to add an hour um add hours and [Music] then I kind of screwed up by add this format date thing let's just get rid of this and see how many semicolons we still have

yeah uh I think this is probably sufficient right all right so let me just run this one more time make sure I didn't completely screw up my place every now and then I have to do this okay and then what we want to do is now that we've add the date we want to add an hour just going to add one and then you want to format the date you want to format it in the exact same format that we just did unfortunately we have to you know do this little bit of math in order

to get this back in the format that we want um I have a feeling that this is going to be a different format but screw it we'll add it anyway and then I'm just going to remove the actual posting thing here for the purposes of this example now let's go back to our sheet let's delete this let's run this on actual live data now how many records are we returning four okay let's let's only return two let's run this and let's see if there are any breaks this is uh I consider this my like end

to end test oh sorry the reason why this isn't scraping anything is because I did this on the generated side I didn't do this on the source side it's working through searching the rows scheduling the time okay great so schedule it for 1,700 should be 1 hour than the 1 hour later than the latest post doesn't look like this made it in Why didn't it make it in oh yeah because I didn't feed it in my bad we actually got to feed in the latest schedule post let's just delete that puppy one more time run

it one more time uh right got to delete this too and with this we should now have everything we need in order for our system to function this is a time this is a time good wonderful cool uh now all we need is we have one side of the LinkedIn scraping system done now we just need something to post and uh the way you actually do the post uh let's do LinkedIn scraping system add to sheet scrape and add the sheet here the way that you do the actual posting is we just need to create

a new scenario and then we'll have LinkedIn scraping system to which is just responsible for checking that generated post sheet and then actually putting the post out so my bad let's paste this puppy in just copy this your um title and we'll call this uh post schedule post from scheduled database beautiful the way this is going to work is we are just going to Google Sheets we are going to search Rose um we're going to do this every hour well I guess we can do this every six minutes now let's do this every hour we

should really start this only on yeah we should only do this every hour if I go every day can I set the times no specified day pretty funny uh no we'll run this every 60 Minutes what we'll do is we'll start this on uh let me see 2024 12 8 2 o' p.m. we're not going to run it yet what I need for this is I need to grab the sheet ID which is called LinkedIn posts is right over here and then now that I have that I'm just going to go sheet name um generated

post we're just going to pull from this every hour what we want is we just want this scheduled post to equal to and we can do this with either text or datetime operator let's try datetime operators try datetime operators uh no it let's not do that well actually let's just grab them let's see what the the format is and then let's see if we can format our own date to be that run we didn't get anything why didn't we get anything H right right so we do have a filter in place so now we have

our post with the date and time so all we need to do is just unfortunately it's 2 p.m. oh just missed it we had the perfect time to test and I let it pass Us by what I want to do is I just want to format the date the date that I want to format is now and I want to format this into the same format that we had before which was H uh sorry was um year uh month month day day year year year year space H H mmss I believe we're going to give

this a try let's just see what we fit in right so this is running at 14131 so that's actually that's uh not going to work because this is going to grab the exact minute and the second so this make scenario is going to take probably more than half a second to run or something so it's it's not actually going to be correct so instead of now what I want to do is I want to set the minute of now to zero I also want to set the seconds of now to zero as well then I

want to test this okay so what we fed in now is we fed in 1208 24400 let's just go to our sheet and let's go uh 140 let's just see if we can grab that data going to run this little test with the current hour huh this is actually not pulling it for some reason why because this one says eight right yeah this one says eight this one says 08 so we used a leading zero I think for the day we don't actually need the leading zero for the day so I use DD we just

need the D right yeah so that's what we need to do my bad we um we formatted it wrong because we slapped an extra zero in there pump that in now we do get the data there's one more thing I have to change though um I was doing the same encoding with the scheduling generated post I was doing it with the uh with the D so we'll do this D back we'll do this D back this should now work awesome and then basically um this this is going to run every hour so every hour starting

from 2 p.m. so 3 4 5 6 7 it's going to grab all of the posts that have the hour equal to the current hour which will presumably just be one if you're you know pumping these puppies out and because we're searching it we're going to have access to the data which will include the text we'll replace any new lines with new lines which is nice and then yeah we should be able to pump this out and then I will just turn this puppy on and then all you need to do is uh basically hook

this up to auler oh we can remove this now hook this up to auler and then voila you have um a pretty well functioning like LinkedIn parasite system that actually is a little goes a little further beyond what I was initially planning on doing when uh the view were sent in their request I was just going to have it like automatically post immediately but this is actually going to smartly intelligently schedule based off of the latest last post um theoretically you could create a Content bank that has a million posts here and it would just

constantly turn it out every hour also theoretically you don't have to do it every hour you could do it every 15 minutes if you wanted to you could do it every 4 hours 6 hours 12 hours I just chose an hour for Simplicity reasons like you could schedule it to run every three hours but I just wanted it to run every hour um just to show you guys what's possible scheduling wise okay so the last thing that I want to touch on is the beginning of this if we just leave the scenario as is um

it's just going to run every 15 minutes and grab the same data set right so there kind of a couple things that we could do um what I recommend we do is we go to appify again and go watch actor runs and we should make this the trigger of our flow and now instead of running every 50 minutes it's going to trigger only when web Hook is called basically and then under get data set items what we're going to do is we're actually just going to feed in the data set ID of the run that

just finished um so that should be sufficient now we do need to set up a hook so I just have to go into I believe it was this account and then it was the LinkedIn post scraper right uh right here and then I'm going to click okay and now this is going to be tied to the runs of that scenar of that actor in appify if I click run this module once and actually go back into appify here and let's just go back to our LinkedIn post scraper and then why don't we just set the

limit per source to be five or something instead instead of whatever it is right now we'll turn this on it'll run and you know it's going to go through and scrape the post voila just finished um that just turned green so a few seconds later this is now watching it it's finished we got the default default data set IDE of the new run and now that's good to go so we know now if we if we turn this on basically and we turn this one on we have like this the self functioning perpetuating system the

only thing we need to do is we just need to schedule an ABY just go to schedules create new and then you know however often you want to run the the original scraper uh because we are checking to see if the post already exists in the database you can just you can run this as often as you want if if somebody runs and does like 500 million LinkedIn posts a day um anytime that you want to change who it is that we're scraping uh you can do so just by going back to our actor settings

here and then um changing the or adding different URLs I should say you can just do all the stuff in make too you could run this scenario um using the appify uh run an actor module and then you can actually just feed in like a dnamic list of people so maybe instead of you having to do it in appify what you do is you store all the URLs in a Google sheet then anytime you run the thing which is once every day or something you pull all of the people from the Google sheet feed that

in as input variety of different ways you can tackle this but yeah that's the completed system so I'm going to save These Blueprints and then make these publicly available to you guys so I'm going to export that number one I'm also going to export this one number two I'm going to go back to number one rename it oh apparently I can't use this uh because of this colon and I'll do the same thing with this one just remove this colon and I'm going to make it publicly available so anybody that wants this uh can feel

free to to use it I don't know why my M Dash isn't popping up and then if you guys have any questions about hypo the St let me know hope you guys appreciated the video had a lot of fun putting it um in place uh as I mentioned previously this is a viewer request and I really like doing viewer requests because a lot of the time they ask for things that um weren't entirely that I didn't fully entirely understand before I actually did the thing for instance I haven't actually done the scheduling flow before where

you um sort a list based off of the last date and then use that to pull uh and schedule future dates so if anything it was really just me honing my skills a little bit more and I really like doing that uh if you guys have any questions or suggestions for future flows just drop them down below otherwise if you could do me a big solid just like subscribe to my channel I have a I've been growing a lot faster recently as I'm sure some of you guys have noticed but as a result I have

the lower concentration or lower ratio of watch time from people that are subscribed to unsubscribed so a lot of like new unsubscribed people are watching me if you're one of them and you like the stuff I put out just subscribe really helps me out