I'm Building AI Growth and Retention AI Agents (Road to $100k/m profit)
March 25, 2025
Parker maps out a practical path to building AI growth and retention agents, packaging the know-how into scalable offers, and kicking toward a $100k/month profit through a focused, self-hosted stack and repeatable processes.
Key vision and approach#
- Build AI growth and retention agents that automate client-facing processes and scale with you.
- Start with high-value, low-ego client work while productizing know-how into a recurring offering.
- Use a self-hosted, modular stack (Docker + Supabase + n8n) to own the stack, speed, and data.
Strategy: map the business, fix the leaky bucket#
- The core problem: undefined customer journey leads to leaky funnels and wasted effort.
- Solution: create a clear process map from marketing to retention (marketing, sales, onboarding, fulfillment, success/offboarding).
- Use the EOS/Traction framework to identify bottlenecks, run quarterly experiments, and lock in proven marketing/sales/offboarding methods.
- Your job on client calls: extract the steps, document them as an SOP, then automate the repeatable bits.
Actionable takeaways:
- Always map the entire customer journey before proposing automation.
- Identify the biggest bottleneck and fix it first; treat the funnel as a leaky bucket to plug.
- Use discovery calls to surface real process steps, not just “we need AI.”
Offers and community: Trouble Free and the AI Pro roadmap#
- Trouble Free: a community concept Parker is refining to balance value and access.
- Pricing experiment: start at $29/month, scale with value; move away from rigid 150-person caps to a more cohort-driven approach.
- AI Pro roadmap: a long-form, value-packed video guide with 100 use cases (sales, recruiting, automated emailing, learning code, digesting books/topics, repurposing podcasts to blogs/LinkedIn, etc.).
- Content strategy: long-form core video + chopped shorts to feed the funnel; self-serve roadmaps and prompts library as core assets.
Actionable takeaways:
- Build a clear, high-value roadmap (AI Pro) with concrete use cases people can implement.
- Use a video-based sales letter and a 2-minute-per-use-case format to scale content output.
Tech stack and architecture: self-hosted growth engine#
- Core stack: n8n for automation, Supabase as the backend, Docker for containerization.
- Benefits of self-hosting: control, scalability, lower long-term costs, and easier integration of AI agents.
- The plan envisions V2/V3 iterations:
- V2: self-hosted n8n + Supabase + Next.js front-end, all containerized.
- V3: even deeper integration (edge functions, PG Vector, custom growth/retention agents) with a broader, multi-container setup.
- Key concepts: edge functions to bring compute near users, PG Vector for fast similarity search, vectorized data to power RAG workflows.
Takeaways you can act on now:
- Start with a minimal self-hosted stack (n8n + Supabase) to own your automation and data.
- Use Docker Compose to run all services locally and in the cloud with consistent environments.
- Plan for growth by architecting with modular containers (growth agent containers, school/retention containers, etc.).
Growth agents and automation playbook#
- Concept: build agents that can operate across channels (YouTube growth, onboarding, client outreach) by ingesting data and acting on it.
- Examples Parker envisions:
- YouTube growth/retention agent that can add chapter markers or optimize video metadata at scale.
- Retention/SOP agents that track customer signals, ping for feedback, and trigger follow-ups.
- Automated proposal generation from discovery calls, turning calls into polished proposals with social proof and next steps.
- The “other people’s YouTube” (OPYT) idea: use growth tactics across multiple accounts to amplify reach, while keeping it aligned with your brand.
Actionable takeaways:
- Start with one repeatable automation (e.g., turn discovery calls into a proposal) and iterate.
- Containerize growth agents so you can deploy them as needed without re-engineering each time.
Onboarding, SOPs, and retention mechanics#
- Emphasis on standard operating procedures (SOPs) to ensure consistent delivery and client experience (Chick-fil-A-style consistency reference).
- Build a client onboarding playbook that includes a full process map, expected timelines, and automatic follow-ups if feedback is delayed.
- Use a feedback ping-pong mechanism to keep projects on track without being annoying—frame as collaborative “practice” to improve the outcome.
Actionable takeaways:
- Document every funnel step and decision point; automate reminders and status updates.
- Draft a standard discovery-to-proposal flow that can be reused across clients.
Content, productization, and marketing leverage#
- Productize expertise into scalable assets: roadmaps, prompts libraries, and templates.
- Leverage batch-style content: long-form videos plus numerous short clips to reach different audiences.
- Use prompts libraries with a quality threshold (e.g., verified prompts that have proven ROI) to attract credible buyers.
Takeaways:
- Create a searchable prompts library with explanations, use cases, and model recommendations.
- Build a long-form explainer video that maps to 100 use cases, then clip for Shorts/Fast Wins.
Next steps and personal focus#
- Prioritize productizing what Parker loves: research, build, and automate AI-driven SAS ideas.
- Keep client work aligned with profit-first goals and reinvest in growth automation and the self-hosted stack.
- Validate new offers with a tight feedback loop, then scale via automated campaigns, retention agents, and growth playbooks.
Actionable takeaways:
- Draft a simple, repeatable process map for a single client project you can automate end-to-end.
- Begin a 30-day plan to collect 30+ Q&A prompts from channel content to seed a knowledge base for your school/agents.
Links#
- EOS/Traction (Entrepreneurial Operating System)
- n8n (open source automation)
- Supabase (backend, database, auth)
- PG Vector (vector search extension for PostgreSQL)
- Docker and Docker Compose (containerization)
- Supabase Edge Functions (for latency-optimized actions)
- Warp Terminal Dispatch (AI automation orchestration)
- Andrej Karpathy on Vibe Coding (inspiration for prompt-driven development)
- Make.com (workflow automation platform)
Transcript
hey there I'm Parker Rex I Led Tech for a startup that did well over $50 million in sales when we sold we did more than that but when we got hammered by VC backed companies and after that I did a bunch of different startups that didn't work there were Silly ideas but it required learning all this technology and learning basically all the different things and so now I have been using AI pretty much every day for the last three years and then started offering these Services into businesses and I also have a community of people that learn from me so I know a fair about a fair amount about this stuff and I'm just always trying to figure out how do you push it further so example of that is I trained 50 different gpts on all the iOS Frameworks I'm using these things like crazy so on this channel I just Yap about stuff that I find helpful in the AI news world you can get that from anywhere but it's a lot of noise so I talk about that if it's relevant and then I basically just kind of Riff through some strategy on what I'm doing and I talk about builds that I'm going to do if you want to watch me actually do the builds those are on another channel so that was a long intro but yeah this is actually what I was working on today if you're interested in superbase and n8n then I just forked the NN project that does self hosting for vectorization and all the all the Tings that you need but having trouble getting it running with all the Tings so if you want to go learn that check out the Parker re builds channel so first of all Q&A let's see if we got any questions a good way to get my attention is to ask a question so good breakdown on Vibe coding an ongoing delusion is thinking that you can idiot input idiot and output working software for now it seems like the output is still dependent on the quality of the input right even a good one to check out Andre carpy the father of vibe coding aka the guy who led open AI to Victory on autopilot previous director of AI at Tesla founding team at open AI he recently posted this about how he built a little app and so you can actually see how he talks to it one thing to notice is he's starting from scratch and he's building a feature this is not a company this is a feature so it ticks things down I mean it could be a company but I don't think it's like a lot of staying power but you know what's funny is I tried to open this last night and I didn't open because so many people probably tried to open it that it busted the page okay we got one of them to open so let's see how he does it look how long this chat is so we are coding an IOS app oh wait we got the original window open let's look at how he does it can you take me through writing an IOS app I'm just downloading and installing okay so get steps on how to do it cool it's for weight loss tracking again something really really simple right he's Vibe coding a feature that will be an app these are bespoke very specific things no existing codebase calculator plus UI so he goes on I don't need to go and look at all these things but it is helpful just to see how he says okay status update this is what we're going to do next here's my idea here's what we need to do here's what I want it to do this is the primary action some math around it like what's funny is this little prompt alone is better than a lot of the prompts that Vibe coders are putting in let's look at his next one now we add two possible buttons simply one button is 100 and one is negative 100 so yeah again very very specific exactly what you want an example reinforcing what he wants so in case the previous sentence wasn't clear then he gets it then he's saying there's some some usability errors and he ends it there and then he goes over to here and then takes the existing thing so he provides context gives a big block of text and then asks for a plan tells it some new context yeah so this would be good to check out if you're trying to learn how he Vibe codes what breaks down though is this is let's see uh he has an import statement so that's where that file starts see if we can see how many files he gave this is all one file still yeah so he gave one file that's awesome it's a single file it wouldn't work as well with a lot of files so yeah iPhone 2009 iPhone I agree with you I agree totally agree check out the video that goes on Wednesday the 26th I pull up Andre carpy's Vibe code sessions it's probably pretty helpful for everyone to look at but the G of it is that if it's a single file it's going to be a lot easier and even with something as basic as he's building his prompts are better than 99.9% of people because he knows how to talk to it because he built the thing I'm interested in learning more about how you leverage AI to help business owners with their process okay so there's a lot of different ways but think what what's happening this this goes into the strategy for my kind of different offerings too is I know how to do this for businesses so I keep begging the question of why would I not just take on as as few clients as possible that pay the highest that are the easiest to do and then use these things for myself because you gain no equity value when you're doing this for other people and it's just like strictly for cash flow so I think that that's kind of like this road to 100K I don't think I'd want to go beyond that of course it just you can't plan like Beyond 30 days right but you have these goals but just to answer your question there's a ton of ways and this plays into something that I was doing today for my own businesses which is is you start out by making a business map that's what I call it but every business has pretty much the same thing where I notice this with clients that come in especially if you go to upwork and someone has a posting and they're like oh uh need nadn Specialist or need AI or need GPT or need agent all these people that are posting on there I'd say like 99% I'm I'm still strong with the 99% most of the people in here they just hear it from somewhere that they need it and so there's like this big old wave looks horrible there's this wave happening and people hear it they're like oh I need this problem is you go and you take these calls you have Discovery call and if you just did that thing like let's say I'll actually use a real use case it's like oh I have this SAS app and I'm a a business and let's say I've raised a seed round and a seed round depending on where you are but if you're in Silicon Valley it's anywhere from two to I mean it goes up to like $20 million in seed funding but then you also have preed but if you said that you're going for one of these companies and they've reached out and they said that they need this thing and they posted something that's like we need n8n because I saw YouTube video about it and we need it because we can go get the customers and getting the customers going be awesome but if you didn't do this thing I'm about to explain of like mapping out the process of what the customer Journey looks like and the whole like pipeline of the business from marketing to sales to closing to the Fulfillment of the thing to the attention plan that you have to customer success or customer failure offboarding all these things you're going to waste a lot of time because you have a leaky bucket so in this case if I if I had a client and the client's like hey uh let's do LinkedIn scraping and we're going to throw that into cold email oops cold email Outreach and they just said that and they just said oh we want you to go and we're going to pay you to give us this big old list based on you know we want B2B SS people because we sell to those for whatever reason and there are companies that are over uh you know $100 million in market cap and we're going to pull that so you need to go out to pitchbook because you can't get that from LinkedIn and then oh we're going to take in another qualifier which is they need to have over 100 employees because this thing is going to be awesome for those types of people it's Enterprise so okay you've got that so then you end up like stitching this all together that's what they think needs to happen because then they can just go and then if they just law of numbers big numbers if you just send enough if we send 10,000 over five days and we have a 5% they do this napkin math they think it's awesome let's say you do that but then you didn't know that okay the thing that's in here like the link like the call to action whether it's uh booking a call or maybe it's just go self sign up like if that next step like if this is step one if step two is leading them to this like horrendous landing page that just doesn't deliver on the offer that was in the email or they get there and even worse it's broken what's the point like then you look bad cuz they like oh it doesn't convert you're like well wait a second your thing sucks so that's just getting to step two so it could be like step three is oh they they booked a call in this case or they did self sign up let's say if they did self sign up for this product and if they go into here and after they've done self sign up there's some other thing like there's maybe like 10 uh Integrations or 10 steps or 10 pieces of data that we need to get that's blocking them or an import with a CSV to enrich the experience that's broken like their funnel not going to do great and then again they're going to look at you and they be like your thing didn't work and you're like uh no and then the case of the book call thing it's like okay well if you're pushing him through a funnel to go book a call and your sales team has no context Beyond like you yapping at them to go and read something about the person on LinkedIn like point is is if you don't have this this map of how the business works then it just is like a swing and a mess because you're going to be just again going towards this bucket it's got a huge hole in it so no one's going to stay in here you want to fill the bucket up with customers and then keep them happy happy so you're trying to like plug that thing the whole time and there's always going to be a new thing to plug like it's just how leaky is Your Bucket your bucket will always be leaky it's just how good is it right that's what retention is so what I do is having done this the right way and the wrong way is you basically have that call and you're just trying to figure out how does this work like can you walk me through the steps for the marketing for like what are you doing now what's your proved method the sales the onboarding the Fulfillment all these a really good book actually about that it's called traction and it's around a system called EOS this is the entrepreneurial operating system and this is something that I preached and actually hated at the time but I've been preaching since because it worked our business Delivery Dudes we were all over the place and you couldn't communicate value across the whole or like you couldn't communicate value you couldn't communicate new releases you could get everyone on the same page without this but it also required you to go through these steps and iring them out and I mean we sat there every Monday and we'd be like I don't even know what our proven marketing thing is but let's just keep trying different ones this one seems to be working and then every quarter you go and you figure these out so that's what that's what I would do and then once you've identified which one is the biggest bottleneck like this step takes time but you're basically sitting there and they like that CU they're like whoa like they're thinking out loud they're going to maybe get like a little jittery because they might not have all the answers but that's your job is to like pull it out of them and then you'd be scribbling down the steps similar to like this like I'm doing this for my business right now which is like I've dealt with customers and clients doing it manually to Define an sop and I wouldn't go and automate this myself had I not done it and made you know like tens of thousands of dollars doing it manually so then having these systems in place for your own business if you're selling these into people like it makes it a lot easier I can give you an example of like why if you have a customer that's a friend and they're not responding to things and you're trying to get their feedback on a piece of work but then it's been 10 days and you've nudged them a couple times and then they look at the thing and it's like oh my gosh it's on fire hair on fire moment fire drill you're like it's okay like we're good this is how things go the only reason that you're having this feeling and why it feels dragged out is because we haven't had that ping pong effect and so that would be a part of this creater process map where it's like okay well you know you'd have somewhere in here where it's like after you've closed the sale after you've fulfilled the work if no feedback you know greater than 48 hours you can ping them say Hey you know just wanted to put this here want to let you know and then you don't do it in like an annoying way it's like I'm even thinking like of providing like an example with like an illustration that's like hey it's like we're playing pingpong and I just hit you the ball back and it it reframes it from being like this annoying thing that's like in their inbox to like oh man like yeah we're just like intellectually jousting right now and we're on the same team we're just practicing so we can get the best thing so those are some good questions um was there any other in there I don't think so no great question I answered this just now it'll be up on Wednesday make sure you check it out thanks again for the feedback and definitely ask more cool so next up is I'm actually working on a new offer because I've been testing things out in in the arena so I have this thing called trouble free it's a community and I'm I'm trying to figure out like how do I deliver the most value based on what I want to continue doing like how do I take the things that I know the things I already do and how I enjoy spending my time and basically like monetize them but in a fair way that everyone wins and so my first offer that I was thinking was like okay cool it's going to be like uh $29 a month and it's going to go up every 10 because that's what everyone does and I thought that I could do like I was going to fun in the first 30 people and then flip it to paid and that everyone would just get like a credit card form but they don't you have to like literally kick everyone out so I have like 40 people in there and I kick them out now so I'm like okay well maybe this new offer is for a different cohort like the original was like it's AI for everyone it's for non-technical folks and Technical meaning like if you wanted to I'm just going to put everything that I know in there and I'm going to mark it and position it for people that know this wave's coming and they don't know what what to do that was the original one and that's what I'm going back to because when I ran into this issue where I was like oh man I have to like kick them out maybe I try to make it exclusive and it's like oh it's 150 person thing and I'm like not your Guru because I was on this like idea that like that's a bad thing but at the end of the day a lot of people do want to learn from me because I have spent a lot of time so I just need to realize that it is a community and I will be the leader of it obviously and that people will also engage more if they want to share if not I don't really care like if they're learning and they don't engage that much that's great but I'll be putting into place some some basically like psychological I guess they're tricks but it's really like if you think about a video game like how do video games work you achieve something you get Stars you get gold you get a level and you feel accomplished so I'm trying to think of like how do you do that without it being like a trick it's like how do you make it is still a trick but it's it's tying it to the value so I think the offer will be forget the 150 thing I have a nice video sales letter that's like super catchy that I spent three weeks making and I'm going to put that back up I'm going to start it at 29 it's going to go up $10 everything and then in here will be a road mapap to being an AI Pro and so I'm going to just basically make a really long video where I have at first I'll have the Whiteboard and I'm going to go through like how I'd use AI for sales how I'd use AI for recruiting how I'd use AI if I want to do automated emailing how I'd use AI if I was learning the code how I use AI if I was learning typescript or Swift or a new topic how to learn anything with AI how to use it to congest or uh digest new books new topics how to turn podcasts into blogs blogs into LinkedIn like all these use cases there's probably going to be 100 of them and then I'm going to record a long video and if each one takes 2 minutes that's great it's a 200 minute video just me yapping like this and then I'll chop it all up and then it'll be the you know step to AI Pro road map something like that and then I can just take clips of that and then go market them against uh a new audience on shorts and it's going to be really easy for me because I have a gillion different ideas and it seems like every single day someone's ask me how to do something so I just need to productize it and package it that goes into the next part which is strategy so I just was asking myself today after I've done some client work that's not so fun and I've launched this school but not really like brought in I followed my initial plan which was like get a bunch of people in there and then I went to flip and realized I couldn't so I'm like rethinking I'm like okay well going into next month how do I do the most amount right now such that I can be like blasting off in April and I just basically write down like you know you are the product of the things that you like spending time on the things that you do spend time on and so right now I really enjoy researching and building with AI I like that's just my jam like I really like doing that I like building SAS products but they have to be in my stack so if I want to actually accomplish something I need to know how to do it right so jumping around and working on other people's projects for client work when it's a total mess or if it's some technology I don't know that's not profitable and I'm putting profit first like I need to I need to take that in and then I need to reinvest it so I can get growth and I can get more momentum so I just kind of start thinking of like okay well how do I take all these skills that I have and then leverage the crap out of them so when it comes to the client stuff you know that's it ties into the question from the guy earlier it's like okay well when comes to this so far I've just been doing it all manually and I did a process map of all the things that I'm doing right and then all the things that I could be doing to handle it better so again I've done it manually but now it's time to put those into work because I've made T of thousands of dollar doing it I feel comfortable now to start automating some of the pieces and then it's automating a process that I've already defined the way that I speak the way that I act the timing of it all that and the V1 is just just to get it all lined up and then it's going to be about The Tweak the tweaking to make sure that it's like okay if it hits some certain threshold I don't want anything to go out because it's going to put my ass on the line if it's not right so it's just getting it lined up so that those automations and those things that kick one another off it's like dominoes that they end up coming back to me but a lot of the leg work's done a good example of this is I stole this from Nick serve he's a fantastic guy he's literally the reason I started doing daily up s he's the reason why I have this thing this scribbly thing I'm like a Walmart version of his account literally but I'm learning a lot from him he's like if you want n8n no if you want make.com like how to make money with make stuff he's like the the goat at that so shout out to him and one of the automations that I took from him that I'm going be building out and enhancing for my stuff in nadn is this proposal thing so I did this manually in that I didn't have any sort of workflow kicking it off but what I did do is I just had it so when I did a call the discovery call which is like you're going to always have that until you're like super famous and people trust you so much that they'll just like enter into the stuff uh without a call but I like talking to people is you fill out all these like things while you're on the call so you're like okay here's the one liner about the business oh there SAS business just wants to drive business to their app and here's the process map and here's what I think they should do based on it but I I'm not totally sure so we should probably maybe think of some stuff to brainstorm and here's a list of all the problems and what they're doing now and their budget and their business size and all this and then his one had I think it's like five steps my one has a bunch more steps not meaning it's better right this is a different use case because I'm not building out the same stuff but mine uh will essentially do the same exact thing where it's like okay you end up with this like proposal that has like all your social proof on it I have like oh like at the beginning of these videos I say oh I did you know XYZ before this and then I put my client stuff on there and that's just like one of the like 30 things where it did the admin work of taking the call and then turning it into the proposal so I don't have to go do it so it's just going to compound and I'll probably do like one or two at a time but that is dependent on me getting naden dialed in with superbase cuz I think that's another skew I could offer it's like it's going to be sick and I'll talk more about it in a sec the way I think about it is Chick-fil-A if you live in the US if you ever been to a Chick-fil-A they always act the same if you go to one in a mall in the middle of like Kentucky or if you go to one in Miami Beach they're going to have the same thing that they say that's because they're well trained they have standard operating procedures Sops and that's what allows them to do it so on more of the strategy stuff with school I've been thinking like how do I leverage agents in these nadn instances and the superbay stuff that's gonna be just chalk full of information about me and what I do and I can train literally all of my YouTube videos on it and my voice and it's G to just have PG Vector Galore which is like you know vectorization that helps you with Rag and whatnot um it it's just going to be doing a lot of stuff around me well cool it's going to know it's be like a clone of Parker and then I can start doing growth agent work where I saw something that was really interesting where talked about in a previous video where it's like okay if you go to a YouTube video and they don't have chapter markers set up an automation that does that and then just say like some little tag at the end so that at the top of every single video it's just like oh Parker maybe it's like Parker trouble free boom there's an ad because the top comment on a video is hey where this the chapter guy and it has 50,000 likes how many people saw that how many people saw and didn't like it I think it's going to get millions of Impressions so I think one is setting that up testing the limit with it see how many different accounts you can do based on their upload schedule such that the account doesn't get flagged or marked as spam or whatever but I think that's just one I think I call it uh it's like other people's money like that's what they you know OPM but I think opyt other people's YouTube so what are the other accounts where it would make sense for your face and your brand to show up because in the age of AI brand and Community is going to be everything so basically you by personal branding you make your face the asset I'm fine with that I think that that's a really good defensible thing so instead of being like faceless I'm totally being face and I want to just have more things like this or like if this is idea one I want to be coming up with a bunch of different other ones like I immediately think for the Parker re builds Channel I should be on the cursor forum and even if I do this manually if I'm making a practice to answer 30 questions a day these are such easy questions for me to answer and this could be across all the different channels so it's like do it a bunch of times manually train it have it go back into super base learn the answers to the questions then you can set up the automation after you've done it for 30 days I mean if you did it for 30 days for 30 answers you have 900 different uh queries that are answer uh questions and answers so that's a fair amount of data then train on and then you could train it on the docs so you could have a doc rocker that goes out and is like okay these are the things that this channel the Parker Rex builds Channel answers questions about and then you just give it an array yeah you can give it an array of different URLs that you wanted to go and monitor and pull to check to see if there's any new questions you could be the first to answer on every single Forum relevant to your Niche that's sick so yeah I need to I need to really get going on that I think the same thing with retention agents where think about this when I open the floodgates they're not even floodgates when I kick everyone out of my group called trouble free and I invite everybody back in I don't want to deal with like the limitations of school's Tech I love I love school for the fact that they exist it's nice but I dislike a lot of the stuff that they have because it's limited so how do you get around that well I think you could get around it by I invite myself as an admin on a different account and then I can control that account whether it's through puppet I don't know any one of these like testing Frameworks or like scrapey things like you could literally do it you could Target you could pull you would you do the same thing that I did up here for opyt of like going and answering it manually for a bunch of times but I could go and I could do it in my own paid community and I think this is is going to be a trend moving forward but I want it to seem like me so I can't have it be like a robot so be maybe I just say it's like Parker's brother or something and then I put like a little like gray thing over it it would confuse people but it would be cool because like what what is this because it would answer almost instantly I probably crawl walk then run where it's like I'd go and I do the thing manually I'd be training it I'd be sending everything back I i' treat School in my Community i' exfiltrate all that data back into my database build it up once there's like a thousand rows or something in there then I can start to walk which would in this case be probably getting a human in the loop so like Hey we're thinking of answering this question but if you could just take a look at it that'd be great and then I can give it a yes no either via text over twio or whatever an email and then after that this thing will just get better that could be a skew for someone else but I don't want to give this Alpha away I could talk about it but I need to actually prove it what else is in the onboarding well it'd be that same process map right so I need to do one for this it's like okay whoops like okay people come in from YouTube here they are they go into your school what do they get how do I keep them how do I make them happy how do I make them feel special well I was listening to someone that has a a consignment store school which is super Niche but they're making like 15K a month off of it which is great for her and what she does she just tags everybody in there and then does like contextual tags so it's like yeah they get tagged in something but she had gone looked at the profile looked at one of the links that existed on someone's profile and so you just list that out you'd be like okay uh list profile links or like get them scrape page so maybe it's like oh new person joins we go we learn about them it's just what would you do manually you'd go and you'd look at the person you'd see if they have a GitHub you'd see if they have a Twitter you'd see if they have a Blog you'd see something relevant about them maybe what they're trying to accomplish is found and then those are all in a database somewhere on super base and then it does a it basically does uh similarity search using those vectors to see oh hey some and so has done done something like this tag them it's so much better one thing she does that I would have to do manually like maybe you could get away with like using something like Tavis is a personalized video so the personalized video she does is like hey you know Parker this is so great for you to be here and you do some like 60c thing that's like undeniable that it's real but maybe Tavis is good enough where like that first part is like hey blink is done with Tavis I actually met these two Founders they bought a bunch of furniture for me in Austin which is really funny Hassan shout out Hassan and Quinn they're worth hundreds of millions of dollars in market cap now paper gains email flow future unlocks this is what I talked about a littleit uh earlier about like kind of figuring out like okay well what's the offer and in the school you just treat them as SKS so you'd have the road map and that's waved so that's zero bucks that's oops that's included and then I had this idea for verified prompts and I didn't end up working on it cuz I'm like this is too much work to do for $39 sorry but a verified prompt stands out because you could only have it be in this database if it's made you money and then there's some like classifier on there it's like okay if this prompt has generated somebody over 10K they get to submit it and then you also need to provide the evals so like examples of it it doesn't have to be like running some fancy tool it's just literally like a g sheet and then you need an explainer how to use it and then you need need the use case and then you need the model now this that's actually worth a lot like if you know that you could buy this thing that's sick and then maybe there's some way where it's like okay you can either unlock that by doing a number of points or you pay $1,000 or something and then the people that are in there that made up a decent chunk of this um prompt Library would get a profit share that could be cool because then it doesn't put it all on me and that was a big part of why I didn't want to do like the normal school thing of oh Parker is Guru but like guess what I'm gonna just throw that word in the incinerator and I am laughing Guru no I I want to just I'm big on the idea of like the shared pool of knowledge is way more powerful it's just undeniable but I trying to figure out I ultimately need to seed the platform I need an incentive to seed said seed said platform so lastly I wanted to talk about this before I actually got to go do client work I fixed A vibe code app and I got to go explain it to someone to a client so it was cool though thousand bucks for a day NAD and super base build so what I did was I realized that the normi person which is fine the average n8n user will go and they'll get the Cloud version that's fine the problem with it is a couple things I want full I want it in my stack I want to be able to do a bunch with it so when I look at my stack and I'm like okay in order to accomplish all these goals that I have and in order to set up this awesome flow for clients where I only do the things that I really like and it does all the filtering for me to find what I like and all that like it needs to be in a system that's like super versatile and I think that it also pays to to know how to do this so case for that would be like the typical one would be yeah n8n and yeah it's more expensive to self host or Cloud host but that's not really the reason it's more that I just need to know how all of it works so I can push it to the limit so if I wanted to do self host like this is like V V1 this is V2 and then this is V3 of where I'm going so V2 is what I I started today it's what I actually roll I didn't roll I just uh cloned it down and what it had was a Docker see you need Docker and has a Docker compos file what that does Docker is if you've ever seen like the little whale it's like a whale right and the whole thing is that if dockor is running it puts whatever you're working on and what's called a container and the container is it's called containerization so you can take that container and they can stick it anywhere because the container is running on its own operating system and the container can be put onto Azure Google Cloud AWS can be run on any computer anywhere and you don't get the excuse of like oh it doesn't work on my machine so that's Docker so what I was doing was I was like okay well what's the stack out there well they have the n8n kit which has Docker running it through yubu Ubuntu ytu is Linux if anyone didn't know it and Linux runs the world so at least all the computers and data centers VMS vpses all that so I take that and that was included and then they just had an Alpine postgress which is great what wasn't great was was cudr and O Lama and I get it these are nice little tools not little they're nice tools if you want to like be local and I don't know it's just not for me so I was looking at this and I was like okay well to my point earlier about your stack guess who's a master of super base me I've spent way too much time with it but it's great because super base can just take these three and smush them together so that's what we're doing so instead of it having be in this like one Docker file where the docker composed file basically declares all the services that are in it and it spins them up and then you can hook that into a cicd or continuous integration continuous development pipeline AKA when you push new code to change stuff in there in the container when you're doing it locally and you push it to GitHub it will run a doctor compose on a machine hosted by iub and then it'll run through like all the different things same thing applies with if you have like a nextjs server which we're going to talk about in this V3 if you have a nextjs server and you run mpm run build which is the process of taking like all the different files that you wrote in next which is a framework had a framework around react it takes that and then it basically puts them into the HTML and the JavaScript and all the stuff that is only needed like the build files right so it makes it a lot less readable but it it smooshes it down and if you did that on a normal VPS without doing it through GitHub if you did like a $5 VPS literally it'll choke the thing because it requires a significant amount of cbaa CPU relative to a $5 VPS so it literally just kills it I don't know if that's a a hate on next or not I like next but point is with V3 where we're going is I can have my own n8n and super base and that'll run on its own little container and that little container is independent of other things but it includes all this and what each one does is really special because with superbase you get Edge functions and Edge is basically the network uh location closest to the person that's making the call so if I have website that I visit that's using an edge function and I'm in Cape Town South Africa it's going to try to do it closest to me rather than in New York and what that unlocks is oi calls which is huge you can do it elsewhere but it's pretty convenient with those you have Edge functions you have off social you have a DB which is postgress and within those it does all the awesome postgress stuff right you have triggers you have uh RPC functions an RPC is remote procedural uh something but it's basically like a a function that lets me put the calculations on the database because the database powerful as heck like if I wanted to do an upload of let's say 18 months worth of calendar events which is what I did for map that's a lot of information so you wouldn't want to be putting it all on one network call across the line and you do it in batches you could do an RPC call so there's all stuff you could do in here and in there PG Vector that's an extension and that can basically add a little row that's like you know like the AI glitter right but in there it's all the numbers so eight one negative one it figures out where that bit of information not bit but piece of information in that row exists in Vector space so similarity search Rag yada yada y point is that'll all be set up I can do unlimited stuff if it's on a b machine you literally would not be able to break thing and then you're going to get better at devops and then you can hook into other things so that's when I say okay I want to have the growth agent growth other people's YouTube agent and I can spin that up right as its own container that's pretty sick and then would communicate back to this Tada same thing is if I have trouble free which I'm in dire need of redoing our site it's so bad but if I have trouble free and I want to be able to communicate with my database about whatever maybe training something on coming up with new ideas or I don't know eventually not being on school and being on there that's important and I need to be able to communicate there another one would be a school agent that kind of retention one and it needs to be able to communicate back and basically I want to say exfiltrate but it's like clone the data so like pulling it all out of there and storing it so we can train and come back and take action on it so that wouldn't be able to happen over here I mean it would but it wouldn't be in the stack that I'm familiar with because super base supports all this stuff then you get all the goodies you get claw you get you know whatever ah just made me think I guess I have time for one more I realized today that by hosting on here and by taking advantage of this thing called warp dispatch you could create a devops flow that deals with auto updating because once a month is what's recommended by super base for their upgrades is when you'd go and upgrade your stuff and when you're self-hosting this is what everyone's all all like scared about is like oh like you're going to have downtime well guess what there's ways around it and just set something to go do it for you so this is just like a total unformed idea but it' basically be like you'd have a set of agents that just goes and does that I don't even need to draw it out but like you you list out like what are the top five devops tasks is it like snapshots well guess what you don't need an agent for that is it like backups which is kind of the same don't need agent for that you could just do a Kon you could do a PG boss that's another one you put into your PG boss but I I need more research on this but it just made me think of it because this warp dispatch thing it's it uses 01 and then it comes up with a plan and it's trained on bash so it's like you take advantage of whatever their like trained fine-tuned thing is and then run an instance in here so if they have a Mac app you'd need to figure out how to run a VM with Mac on it I don't know for another time but yeah I got to get to it I got client work to do and it's super late but oh man I had this one too this is for build ideas I just love scribbling down ideas when I see them was like okay I need to have a checklist for new ideas this one's dumb but it's fun it's like oh I'd have a fully customizable quote react package because every single website on the internet has like testimonials and I don't know if you could make it better and make it look really good and then just open source it that'd be cool but why would you do it what's the value so I said like oh for my new ideas I need to have a checklist like what's the oneliner what's the Y combinator one liner no fluff why would you do it pretty good reason I don't know you should have a reason to do it what's the value what's the probability that it's going to have a good outcome that's it that's the video for today whoa my hair is crazy it is almost 10 oock on a Monday and I still have so much to do oh well without further do go ask a question and like And subscribe and uh oh my gosh all right I'll see you tomorrow