Parker Rex
All videos
Parker Rex Daily

Planning Agent Builds using OpenAI SDK: 3 Agent Ideas

March 14, 2025

Watch on YouTube@parkerrex2

Parker digs into planning agent builds with the OpenAI SDK and walks through turning a YouTube video into a detailed, reusable Winning Playbook you can study and replicate.

OpenAI Agents SDK: what it is and why it matters#

  • A toolkit to build and orchestrate multi-agent workflows with more control.
  • Includes capabilities like web search, file search, computer usage, and traceability of agent decisions.
  • Traces show each branch the agents took and the final outputs, making complex flows debuggable.
  • Concept to scale: think of many subagents (15–25 per user) feeding a single chief agent that coordinates everything.

Why multi-agent flows matter#

  • Complex tasks become deterministic by breaking them into steps and controlling each step.
  • The SDK helps you manage branching, routing, and outputs across many agents.
  • Your mental model should treat the system as a process flow with a chief agent coordinating subagents.

The Winning Playbook pipeline: turning a YouTube video into a structured playbook#

  • Goal: extract a step-by-step, research-backed playbook from a video you can study and apply.
  • Core pipeline:
  1. Download the YouTube video
  2. Transcribe with Whisper (via YouTube DLP or similar)
  3. Use an agent workflow (Agent SDK) with web research to pull in high-quality sources and structure
  4. Produce a one-pager JSON: name, initial niche, current niche, steps, and timeline
  5. Editor/refinement pass to tighten actions and add precise sources
  6. Output a polished, reusable playbook (with links) ready for a site or product
  • What we tested:
  • 3A: Agent SDK with web research for structured playbook output
  • 3B: Editor/refine pass to improve clarity and grounding
  • Key outcome: a repeatable, auditable process that can scale across videos and niches.

Tools and setup mentioned#

  • OpenAI Agents SDK (core orchestration)
  • Swarm (earlier orchestration concept)
  • Web search, file search, and computer use within the SDK
  • YouTube Downloader Pro (for video retrieval)
  • Whisper (transcription)
  • AI Studio (token counting and prompt testing)
  • Grounding with real sources (avoid generic links)
  • Storm (used for research-papers workflow)

Practical ideas and product concepts#

  • Winning Playbooks: a site where you pick a niche or tactic and view a detailed playbook built from real videos (timeline, steps, sources, and a readable narrative).
  • Interactive filters: by niche, tactic, and timeline; clickable steps with linked sources.
  • Collaboration/verification loop: reach out to creators for confirmation and backlinks to boost domain authority.
  • EBay agent concept (random but useful): automate listing drafts, pricing, and research by texted inputs—demonstrates how multi-agent flows can automate end-to-end tasks.

Strategy to scale toward higher leverage#

  • Focus on high-leverage, repeatable templates rather than bespoke builds.
  • Make “service as a product”: template-based Make.com-like automations that can be sold per project.
  • Ground SEO and keyword research:
  • YouTube is a major search engine; content strategy should combine SEO for web and YouTube titles/descriptions.
  • Don’t chase every algorithm factor—start with clear, problem-focused keywords and scale.
  • Use keyword-driven prompts to guide playbook generation and content creation.
  • Free content to build goodwill; monetize later with higher-value products or courses.

What’s next in the daily workflow#

  • Next video will dive into Line Rider (a planned detail) to show practical implementations.
  • Content will live on Parker Rex’s blog (parkerrex.com/blog) with a dedicated “daily uploads” resources section.
  • The approach aims to be transparent: share process, tools, prompts, and outputs so others can replicate.

Quick actionable takeaways#

  • Start small: build a chief agent plus 3–5 subagents to manage a single, well-scoped task.
  • Use a YouTube-to-playbook pipeline: video -> transcript -> structured steps -> sources -> editor pass -> final one-pager JSON.
  • Ground outputs with real sources; prefer high-authority, topic-relevant links over generic references.
  • Test multiple prompts or agent configurations and compare results; iterate to improve specificity and usefulness.
  • Consider building a public-facing “Winning Playbooks” vault with filters by niche and tactic to bootstrap value and community engagement.
Transcript

hey there I'm Parker X I ran product in tech for a startup that did over $50 million a year in sales until we exited after that I took a couple big shots on goal they didn't necessarily work out they're pretty silly ideas but along the way over those three years I was using AI pretty much every single day and now I teach it and I teach what I'm doing in the field as I'm building out an AI services company to fund my own SAS product s so you'll learn it from somebody that's in the field doing the stuff every day and making money doing it so let's kind of break down what we're going to be doing in these types of videos so let's clean up here this is what we're going to be covering obviously I just did this hook thing but I'm still working on the format of this but I think in AI updates plus the application meaning how do you actually use this stuff in the real world is is actually really important because like I said a lot of fluff exists in the world where you have like flashy videos that don't actually show you how you can use it so we will talk through that we're going to talk through some of the ideas that I've come up with that people can feel free to take just around what's possible now with some of the newer updates that are coming out and then we'll go through some of the strategy stuff if you guys have questions you should drop them below I'll be happy to answer them so first of all I didn't put this on the list but the agents SDK I talked about this yesterday but the open AI agents SDK is a set of tools for developers to be able to better take control of the agents that they're building and if we just come in here and look over it it basically took the product called swarm that they had released which allowed you to again like better orchestrate multi- agenic flows and it allowed it it basically beefed it up and made it better and orchestrating multiple agents is a pretty big problem to solve and there are a bunch of different Tools in here like they brought in the web search they brought in file search they brought in computer use they brought in some other stuff on the actual platform where you can see the traces of how things are going so if I go into the playground sorry the dashboard you can see traces so a trace would be like a workflow output that the agent did and then if you clicked into it you can see what happened if you had multiple agents where there was a fork in the road on a decision that needed to be made it would show you that too then it would take off which route that it went and then show you the output of it so I think that that kind of plays into this bigger idea of like things that you could do that would be much better controlled and in the multi- agentic world you basically I like to think of everything as like a process flow and this will play into the next section but in perplexity case you have the user types in this box so you have a search and that would be agent number one where it's the query and then when that search goes off you would have something that's going and it's actually kind of reorganizing the search engine results pages so that's when SCP is so that would be the second step is kind of sifting through those the third step would be a summarization so summarize the fourth step would be it typically gets that list and then it rewrites it so you would have an editor and then that edit I'm not sure because they're not h s but I would assume that they also you know once they have that then they're going and they're doing like some additional research on top of it so researcher Plus image and the researcher might be responsible for filling out all the followup and follow on questions so then that would all render to what you see on the page so you'd see you know here's the user query at the top with a little search box here's let's say a bunch of maybe links that are buried inside of some text content and then you'd have your images so with that example to kind of simplify things we can better understand how the agent SDK would allow us to get better control over each one of these step functions and I've personally run into this problem it's a big reason why I was not able to get a successful launch out of my product called map was because each one of these steps is harder and harder to kind of get like deterministic answers upon as you stack more and more and more of them so in my case we have a Notes app which the point of that is to view cursor or writing a lot of people would say well you could just do it cursor but the thing is is like you know an noral person would want to use cursor you'd have notes and then that would probably have like five to seven different ones of these and then we had health which I'll just put in a heart and you might have like five to seven of these then we had tasks SL projects five to seven of those you know the sidebar with the different projects and then you you have your to-dos so you might have five to seven of those agents and then you've have a calendar and there might be like 12 agents in there because there's a lot to do in a calendar and so even if you just had these four like the fifth one was basically or is a a homepage but then all those so let's say if on the low end we had 15 let's call it like you know 25 sub agents all working in unison around one uh like basically one vector index one vector store per user because they need all that info that they do have to have like similarity search and some of that nice stuff on it then that kind of gets tricky right so then you need to go from 25 to one like chief agent I'll call it and that chief agent would be the one that's you know on that homepage that I was mentioning so that's really like page five where you know you're going through your goals and let's say you had a list of goals so you have one goal you have two goal you have three goal and this one chief agent needs to go back in and be able to check on these 25 kind of run through this as you can imagine you know things are changing in here as as asynchronously and so this one would be the health one where it's like all of your wearable data is coming in and it's getting stored it's getting uploaded non-stop so it just gets a little bit complex so I think with this process in mind like if this can be your mental model for how these agents work then you'd probably better understand how this toolkit will be really really helpful um and that's what I want to talk about in a sec where it's like I need to kind of start building these live on these daily uploads so you can get a better idea of how it works and so I can get a better idea of how it works and then you can get better outcomes for businesses I can see how before the barrier to entry to build someone a custom agent be very high because it was difficult and hard to maintain but I think with this set of tools it would be much more PL for you to spin these up in just a week or two so the next part I want to talk about and this could play into the agents SDK because you can just chain these things together the like glue these pieces of tech together is that the Google launch yesterday allows us now to enter in YouTube videos so I was researching the guy Neil Patel I'm sure you've seen him online you look him up I came across a video yesterday where he talked about his journey from zero to100 million guy's insane by the way like he made like his first 100K a month I think at the age of like 18 which is just insane and in this video he breaks down his Playbook and what I traditionally would do is I just sit there and most people when they're learning they just take notes and I'm going to take a a sip of this cof real quick but time for money when I was 16 years old highly recommend him uh highly recommend watching this video but you still should take notes because you better register the information but afterwards it's you know if I wanted to take what he's done and then try to emulate that in 2025 and Beyond where I want to go from you know 10K a month to a100 million dollar a year but I want to chunk that up to like let's start with getting to 100K a month and then let's start working our way up how would you do that and before I had a way to do it right so it's like without this new Google Update that lets you put in YouTube videos you can just go into warp and you can just say hey look you know I made a directory called winning playbooks and I said we need to download the following YouTube video and then transcribe it with whisper into a text file in the same folder so then I pass it that it checks to see if I have the tools required so there's YouTube Downloader Pro I think is what DLP stands for but I could be wrong and then you have whisper which is free tool that open AI lets you use that transcribes things and it gives me the command and I get it then it runs and you can see it downloads it great it creates the transcript so now I have this transcript and what I could do is then feed that into a an llm and have it extract it to give me the Playbook and I did do that because I had not seen that this new update come out had come out it didn't come out until like three o'clock but the way that I did it was with notebook LM and I fed at the video but now oh by the way it's my birthday that's what the Ballo are for thanks but now ai studi kind of taking over what this product does where there's restrictions on the amount of prompt that I can put in here so I couldn't put in the biggest prompt that I wanted so AI studio now does this and I find it it's pretty helpful so let's see now when I come into AI Studio I can just drop in the YouTube video and it'll immediately tell me how many tokens are in the video like so let's just do an example I'm going to grab this video on the five notebook LM hacks that will blow your mind and see just how quick this thing goes so let's create a prompt and I can go down to the plus sign and hit YouTube video and then as soon as I paste it it should just show me the amount of tokens yeah so in like a second it pulled the amount of tokens that's because I'm assuming it's pulling it off of the subtitles not the transcript or maybe it is the transcript but either way fact that this exists now makes so much sense because Google is the most well-positioned company to make the best video AI generated content on the planet and in terms of like like just think of the amount of information like let's get a visual of it how many minutes of YouTube content are uploaded every day so let's search that up just so we have an idea of it because it's overtaken Netflix watch time it's you know the biggest media people on the planet are out they're like basically beating news outlets like the the folks that just literally jump on like this and talk so this is the future we're here but this is such an old article but it does say like even like as of three years ago 500 hours are uploaded every single minute that's nuts but we do want to find something a little more accurate because I want to kind of drive this point home as of March 133 2025 my birthday how many minutes of YouTube video are uploaded every single day I want to prove a point about how large the ecosystem of YouTube is for research paper so we're gonna fire that off and we're also gonna open up Storm because this is an AI research paper thing it's kind of insane I'm writing about the breth of YouTube and its impact on traditional media and we need to Steelman the argument that YouTube is the future and have as much quantitative and objective data sizing up its impact on the world in terms of viewership watch hours comparisons of it versus other channels and more so we're going to get an article written about that oh no okay research the size of YouTube in terms of its watch time and viewership versus traditional media so let's let that one run it's going to take three minutes and we get a Stanford research poer and then looks like 4 43 million minutes of YouTube are uploaded daily so if you just think about that that's that's kind of insane let's look YouTube users I know this is a little bit of an anecdote but that's what this channel is about is like we're just rabbit hoing this info and then we'll tie it back so interesting YouTube stats 2.7 billion people okay what's the global population as of March 13 2025 that are connected to the internet I just want to see as percentage how many people on planet Earth that have internet so 5.56 so what a little under half the people on the world that have internet oh man there's more people on YouTube than there are people without internet that's good so 93% of population is online wouldn't that be it's funny it doesn't really I guess that's Regional but that stat would conflict with the one above it so anyways let's check and see if this research paper is on I need to share the motivation we're steelmanning an argument that YouTube is the future of media by comping it against the previous companies or media Outlets um and you can see I mean I I hope I seeing how I kind of work you can see the breadth of tooling that AI has like we went from let's see how many tokens they have to oh I think this is you know really meaningful for them because they own YouTube because they own so much video it's probably somewhere deep in a terms of service that allows them to do whatever they want with that uh so the summary it looks like it launched an' 05 yep 2.5 billion monthlies 1 billion hours watched every single day that's as of two years ago so majority 18 to 34 that makes sense I'm 32 now and these are all 2022 numbers so I should have mentioned that it should be later anyways back to the point I think that this is going to be really meaningful um because now I can start to query against this so I made a a prompt and by the way I put all these resources at the end of the video I put them in a spot on my website under the daily uploads so if you just go to my website and go to the blog you'll see yesterday's in here so yesterday I talked about client and then I have all the resources mentioned were they linked out it includes the Whiteboard it includes prompts so and we're going to use this to make line riter in a second so back to here sorry back to my Pro prompt on turning the YouTube video into a winning Playbook to analyze What winners are doing and then turn that into an asset that you can study so if my brain would just turn on it'd be fantastic if I go to the playground then I go to prompts it's because they change the UI a little and I can see that I made this prompt that basically says provide a detailed analysis of the journey from of an entrepreneur from being broke to becoming rich using a YouTube videos reference additionally research and more information where necessary to understand to enhance understanding I wanted the time frame I wanted their Niche because I think this could be a product that would get really popular it's something that I want where I want to be able to go to a website and then select the niche that the person is the winner in and then see how that works for them like what did they do what are the steps step by step what was the timeline so that I can get a realistic understanding because they say takes three years if you're building a brand this isn't just from making money but if you're building a brand takes three years to get real momentum where you're using all the same branding people really start to recognize it you're everywhere you're on all the platforms takes five years to go from having momentum to like being an industry player where like you're getting invited to stuff and you're influential and then it takes 10 years for you to basically like have staying power so it's it's good to know when you're looking at this stuff like how long is it going to take so you can set expectations and so you get that then you get the details so the step-by-step summarization in one to two sentences and then an in-depth breakout of each step so for example an analyzing Neil Patel's video my story from 0 to 100 million you should identify distinct steps you took summarize them and then research further details such types of Articles of not specified in the video so then you give the steps so these are the things that it needs to do transcribe the YouTube video of the entrepreneur and identify all key steps mentioned summarize them into one to two sentences for each summarized step conduct additional research to provide a more detailed analysis UB web search to do this pulling the highest ranking I have more coffee I need a chance highest ranking links and returning the that's basil by the way for the three viewers yeah happy birthday thank you and verify the sources returning the outline of the source material actually I'm just going to say pull the high or reference referencing inciting the highest ranking links verify the sources if they don't have sufficient detail and then outputs some one pager so this would be really helpful right and I do think that this you know output format it would be structured output with Jason and then you could just like plug that into front end but let's take that and in the old world what I do is I like not old world literally like as of yesterday this thing new if I didn't open AI I would say input is a transcript attached here and I just grab that transcript that I had manually output so let's and I'd grab this and I'd put it there and then I can just run that let's see which model it's using O3 Min High reasoning ey cool but with the new version in AI Studio we can kind of comp it see how the chat playground with open AI Works using a transcript versus not having to do all those steps that I did before the video should I be just put well I kind of this prompt so I'll grab this and I'll put it over here actually I'll put it in the system instructions but I don't need there it is I don't need this part so I'm GNA peel this out because it's going to be using the built-in transcript and put the video in your message just put this here and we're going to use their flash thinking because this should be the best one basically the newest oh what's learn lm5 that's new learn LM I love this say all these are just free it's kind of crazy multimodal understanding realtime streaming latency use case processing 10 L thousand lines of code yeah so we do want to have a tool call so grounding with Google search and we're going to put the temperature down because we don't want any creativity we want it to just do the thing we're asking it so let's compare these two and we already have one output scooch this over okay below is a onepage analysis of an entrepreneurs Journey from struggling to Poverty to building and multi9 figure digital marketing Empire based on a detailed review of a representative YouTube video transcript yep overall duration approximately 10 years from the Breakthrough days is a 16-year-old in low-income circumstances to establishing a robust business generating 9 figure annual revenues this alliance with industry Insight showing that the sustained traction maturation and scaling often require a decade of focused effort see the Harvard Business Review see what this linked out to okay so it didn't give me a specific link so that's kind of lame initial Niche the entrepreneur began by creating a job Board website modeled after platforms like monster.com and he experienc with experimented with basic content creation and digital marketing tactics to drive organic growth he now operates within the digital marketing and performance-based advertising space running a full service agency I see I would want to have like more information so initial and current niches is current Niche with basically business kpis if applicable if possible and then we'd say current Niche oh initial Niche at the start of the journey we want to swap those around because that would make more sense I want to see where he started where he ended I want to see the path I think also a good one would be uh a a timeline of basically the duration so timeline displaying years in business and revenue or profit generated by year example year 0 through one 100K a year year one to two 250k a year and so on and then because I put the following structure there I need to have this where instead ofing fors five I'd say the last step is create a well researched timeline of the entrepreneur journey by including the each year number and the revenue profit generated during that time now I want to keep going through this but as you can see this is kind of the workflow right where it's like I have two of them going and I have the first come back I'm going through it I'm trying to make sure like does this do the thing that I wanted to do like is this helpful is this not helpful and I just noticed that my camera it's on here right but it for for whatever reason it's not on the screen share so sorry that you didn't see me for a little while I don't have any clue sure you guys missed me but let put it back on there we go now I'm here but as you can see this is kind of like what I'm doing is anytime I'm trying to make a prompt or solve a problem which should happen like hundreds of times a day you get a fire alarm test one okay camera's good again so what I'm doing is I basically I run two the ones that I think are the best for the task and then I fire both of them off and then I review the results and then I tweak so we've already made a tweak where we added a timeline and now I want to go through and keep finding if there are issues in here so for the links I know that we didn't actually have the proper ones at least for that first one yeah like it's just it's just linking out high high ranked High domain Authority websites so I'd say if there's something I want to find like where the links are referencing the site yeah referencing and sighting the highest ranking links so use web search to do this only include links if they are specific to the topic at hand not generic URLs like inc.com or hbr.org we need links to actual to real and uh relevant articles cool um early life and motivation yeah that's a useless piece of research this is where it gets interesting so he took an investment and he built out a website and then he chose to learn how to do digital marketing and then he immersed himself in research experimenting with SEO and content tactics so this would be another one where it's like I don't know what that means like SEO and content tactics well what it was it specifically there probably somewhere on the internet that he mentions that so in the breakout I would say to PR provide a more detailed analysis do not use General tactic language use uh specific key actions to take so that the reader can replicate the entrepreneurs playbook for example do not say experimenting with SEO and content tactics instead say he uh leverage on page SEO by conducting keyword research and publish in five high quality articles 1,00 to 15 or 1,00 to 1500 words per day okay we'll see if we can do it the rest of this is going to be bad because we already found like the majority of what we didn't like in here so let's fire that off again um we can just I think there's already a Rong but let just click there cool and now let's check out um Google's we'll take the updated one because I have a feeling yeah first of all there's no links which is strange you'd think there'd be links so before I even read this I can tell that this one's it seems like it's worse um here's the analysis the entrepreneurs Ryan do oh my gosh we have the wrong video that's fine so let's remove this let's remove this and we'll give Google an advantage we're going to put in the good one and we're going to remove the transcript so input is removed all right input YouTube video attach and then let's grab the Neil Patel video and make sure it's not with any sort of like seconds starting you'll notice they add like an equals sign if you have a start ad date and then I'll just run that we'll go back and see the update the reason that I'm getting so uh specific on this is I would want this to be ready for a product like the idea would be that this would be a um like I'm I'm kind of getting into this this section here where it's like how do you take a new launch and then you turn it into an actual product that would be called like winning playbooks where anyone could go and on the website you might have like a bunch of different entrepreneurs in here and then you can select by the niche you can select by maybe like the I don't know tactic that use to grow that might be an interesting one because if you are getting an SEO then you be like oh I want to do the ones or I want to see he did it with SEO so you could do it by like growth tactic you do it by Niche and then you'd be able to see all these and then if you clicked into Neils you might see like I don't know like the timeline cuz that's flashy like a you know where you see them going up over time maybe you could have little interactive things where when you click on them it shows you the tactic used but I think this would be I know this would be really helpful for me and I know that there's a lot of interesting growth stuff you could do around this where if I made this really fancy page for Neil and it was like winners Playbook or whatever then I could call the email him and be like hey I just wanted to let you know I built this thing because I thought you were awesome could you just see if this is even close accurate and maybe if it is and maybe if he does respond then they get like a check and it's like verified and then if they end up writing about it then your domain Authority would go up because you're getting backlinks then the same could go where it's like later down the line if you wanted someone to be able to add one they could go on here and you know maybe these are like paid where it's like 19 bucks or something to get access to it you give some free ones but if you wanted people to seed the platform by literally just uploading a YouTube video then you could give them discounts you could probably just build this out of a pipeline as well let's look at what it came back with and then we're going to move to the next one because I'm not actually going to build this out otherwise we won't get to any of the other content so kind of disappointed in this I don't understand why I can't get links in here I don't know if it doesn't support it but let's read the quality of it took roughly 16 years I specifically askd content marketing and paid advertising um that means we should refire this one updated prompt step one understanding his background um yeah okay building leverage he worked at this making this he decided to copy that paid someone yep we know that he found this thing they spent a lot of hours but he got no results decided to learn so he went from 0 to 60 he started hitting up other websites in the space and begging them to link to his website he ranked higher on Google he was taking AP and I classes during high school and doing nighttime see this is much better and much more specific based on a couple really quick promp changes his first speech was on how Google algorithms work okay someone in that class was taking was looking for someone like him they worked at a power supply company got his first contract making him 5K a month see this is way better because it is doing what I wrote in my notes which is like that's all these models like you want them to to give you the output that you would have done yourself like that's the goal with all of these right like everyone is always struggling to get it to act like them like that's what you want you want it to act like you so the way that you get it to act like you is you just try over and over and over and over again by tweaking the promp tweaking the promp tweaking the promp and so when I look at this this it's really similar to what I was writing what I'm again not seeing is the links so it's grounded but I don't say links so um got his first contract and then he was making 20K a month then he started crazy egg I don't see a time oh there's a timeline there's a rough timeline yeah so this still didn't go outside the box but it's good so I haven't really used this very much I don't know if I can sa this as a prompt that would be really cool um can I save a copy oh it looks like it is already saved okay so let's call this winning Playbook Playbook prompt and let say an easy way to feed a YouTube video and get a one-pager explaining the Playbook that entrepreneur used to succeed okay I think that's good do think that's good so let's do one last review and then we'll move on because we've got to build some stuff um okay additional Insight this formative stage was about mindset and it was about resource allocation princial detailed in so we got a high ranked SEO page but for some reason it sent me to the Spanish version that did not give me it looks like that article might be outdated I'm looking at the bottom left and I'm seeing link he hir a developer and build an MVP expecting natural traffic Discovery however the lack of integrated marketing marketing led to minimal traction teaching him the crucial lesson lesson that great digital products need strong marketing day one I've learned that too uh for replications research how to launch a job board using Niche and effect framework again bad link after a costly and in effective Outsourcing experiment he dedicated himself to learning digital marketing SEO achieving Breakthrough by immers himself in online resources yeah so what I think you have to do [Music] is oh this is ironic it's linking himself that means he's a winner I I think the next step for what I do here is I I probably wouldn't use the Google one which is funny like I'd have to do this extra step so I like this but I think in order to turn this into a product you would have to do a couple things I think what we always do when we want to make something so if this was a product which I think I'm going to make just as I'll like include this as a part of um The Daily upload but it'd be like winning Playbook a whatever you can go by the domain I don't care I'm not going to get it and then step one would be YouTube download step two would be YouTube to text and the tools you use Here YouTube DLP and Whisper and this could be done in code this could also be just done on make.com or nadn because it's selfhosted and you know or not it's Cloud hosted you don't have to do any anything really um and then after YouTube to texts The Next Step that we did was We did an API call so this would be an 03 Mini I prompt or better yet what I do think this would actually be so this gave us like a really mid answer um so this would actually be an agent and we could probably use agents for all this but it would be agents SDK with web research search CU I think it'll give us better outputs with actual links and then once we get that I'm assuming if we fire that off and we use the agents SDK and we should get better links we already have the transcript here we'd be prompt injecting the transcript and then we should get the output just like this on the right so then we take the response and we could Plus up so we might um do like after we get the response then there be maybe like this is 3A so 3A would be agent SDK with web research so that gives us the initial version of it and then 3B be really like refinement so agent SDK edit SL plus up so if you came back and you saw that it didn't have the stuff you want like I see here then I could put in ay read through you you're the editor of a draft for a website called winning playay books.com and you're on deadline you need to get this out the quality is not quite there your Chief editor is waiting for you here's the following things you need to do here's the expectation here's the tonality all those sorts of things that would be kind of like the editor and plus up let's just think through this so now you have the edited version I think that it does give you structured outputs so then you could probably serve that back to a H you then do an insert so you take uh you like store the output as structured Jason Json that's JavaScript object notation so that would look something like just to think at it'd be like name Neil or uh NE would be string we're just doing like the domain and architecture like just thinking it through we have name we have initial Niche is the string we have current Niche is a string we have steps and this would probably be an array of strings and we'd have could include links I could include all this stuff and then we'd have timeline and we could just have that be another string cool so in this actually exercise what we did was we basically showed off that the new chat GPT release seems to be better than the Google one but it would require more steps then we kind of ideated across what the steps would be for multi- agentic flow taking one YouTube video and turning it into winning Playbook of exactly what they did so entrepreneurs can study it learn from it I think that's pretty good I think that would be really really helpful for a lot of people so that's that another idea that I had was as I was looking at the agents SDK I was really curious I'm like what's like a really random but useful thing that you could build that would automate maybe even like an entire job um or like most of it and I thought of one that was really random which would be like this eBay agent because I had bought a bunch of keyboards and anyways I didn't use one of them and I had this fancy keyboard that was literally Gathering dust and I'm like okay it's time to sell this so I can get a better camera and the process of doing eBay has a few steps in it that I think you could just make agents for so let's scaffold that out real quick and the way they I do this when I'm thinking about it is it's just standard it's like what are the things that I do when I want to get something done then you just translate that into Tech so for the eBay agent oops need to remove that button there's something on this little pen that really screws me up sometimes there a button on it that's I don't know why but it like locks itself into scribbling so if you did an eBay agent um come on we go if you did an eBay agent and there's entire like stores I don't know if you guys have these where you live but there's like stores that still exist where it's like oh I'm an eBay guy you can go buy but you basically create listing copy and that includes the title that includes the description that includes the um it's really B and then you'd have other fields that's like condition you have photos you have price so there's really I mean this is kind of it and if you wanted to just automate most of this you could probably get away depending on the access in their API like a best practice when you're selling something to figure out the market price is that you go to sold so you go to sold items and then you can see what's actively trading so in the case of like a keyboard I have like a moonlander those were trading at 250 bucks so you could probably based on their API if not you could use computer use you could go and see like ideally I would just have some sort of interface where I'm uploading a photo like not even an interface I think even better than that it's like I ideally want to text the assistant photos and I name and then I should get back the draft I should get a draft back basically with the price and it should link out what the competing prices are so that I can know pretty much exactly that so instead of scribbling I'm just going to do like ideal flow would be text the eBay agent receive a response with the listing draft price researched and then you'd say yes or no if no what changes and then confirm go live I think that'd be a pretty pretty good example where it's like I was trying to denote that these were different people but yeah so I think that would be actually really really helpful where it's like you do this in like two seconds it just be done so that's an idea for you guys to take if you want what else so on the strategy side um getting to 100K a month how do you do that so I've done different types of work some with lower leverage than others and I've been trying to think like how do you get the most amount of Leverage and especially when you're doing like contracted work because you're still trading time for money but you're getting a massive multiple on it and what I mean by that is like when you do it by project because if you're really good it should take you way less time so like my kind of V1 of this was custom software custom software Dev and when I was doing it it's like high paying like you can make 20 to 30k in you know a little bit over a month even a month probably you could probably get away like 25k a month depending on the customer but the leverage was really low meaning like as soon as I went to a new customer you start from zero which is not good then now the V2 that I'm looking at is like okay well maybe the make.com and Ann world it's good or maybe you could even templatized meaning like 8 to 10,000 a month um or per project and then you could pick those off and I do have templates for that so it's just kind of a thought experiment where it's like you know how do you get like you basically want Max leverage where if these are projects and you're jumping from one to the next you do you basically never want to have it be like totally custom where you're not using anything this even if you got like just 5% is reusable that's still horrible versus what seems like the highest leverage stuff is really like these make slnn where your Leverage is you might be putting in a ton of time up front but after you do it you you know if I complete the project at 100% then I go to the next one in most cases it's going to be like 90% of the work's done because you've built the automation out it's a very deterministic thing it might be like um turn website leads into proposals where maybe the user goes and they fill out a little form like a quiz and then instead of getting a phone call instead of it just sitting there instead of putting them into a funnel and then later hitting them with some custom proposal they get a panda do I stole this from Nick serve I think is his last name but I did this and it was fantastic and all you'd have to do to change this is basically like the inputs so instead of pointing like your instead of pointing qu or having questions are specific to one Niche you just change it so that's where I'm going now is like going after those I also recognize that when you're early on a platform need to kind of Pick Quick wins so you can get the ratings because they have their own SEO and that brings us to the last piece on here which is SEO keyword research so something I learned just from that guy Neil that we were talking about was how SEO goes beyond just Google Google only accounts for 17% of searches which I thought was crazy because search exists in other platforms like Amazon has a search engine um kind of crazy to think obviously there's Tik Tok obviously there's insta and naachi really chipping away at it but what he mentioned is this quote from the uh CEO of alphabet where anytime you type a keyword in in a anytime you type in a keyword as Google we have gone and stored billions of copies of pages in our index anytime you type in a keyword as Google we have gone and stored copies of billions of pages in our index we take the keyword and match it against the pages and rank them based on 200 plus signals things like relevance popularity and how other people are using it Google's looking at a lot of things so when you're doing keyword research don't worry about the 200 factors I know a lot of seos talk about them I'm going to break it down so it's as simple as possible and you don't have to worry about each and every single algorithm change or factoring the algorithm cuz it's just too complicated and it's just overwhelming as I mentioned picking the right keywords will make you more money and picking the wrong ones will hurt you my last startup I was getting roughly four times more traffic than my competition they're worth $800 million during their Peak and we weren't even worth one tenth of what they were worth the reason being I picked the wrong keywords that just shows traffic is everything now most people start off with Google Keyword plan so what I think is different about this is a lot of people think that seo's dead because oh now we have ai well no that's wrong SEO as I see it today is the starting point for all your content YouTube's the second largest search engine think of the amount of time that Mr B spends the largest YouTube in the world that he spends on writing titles that's the same that he's doing SEO for YouTube It's a mixture of psychology um and SEO like you can game it one way with search intent where you're answering questions to people's problems and that's an okay strategy I think that you know if I want to figure out how do I connect this Bluetooth device and there's someone that answers that question it's really well done I might subscribe probably not but I think it's important and what I'll be doing moving forward is kind of nailing these keywords out because there are automations that I can use with my you know hitl human in the loop where I can check to make sure that the automations are making blog posts making scripts making um outlines that sound like me but they'd be grounded in the keyword research so I think I'm gonna actually well I was said I was going to wrap the video but we're gonna have to do the Line Rider thing tomorrow um so that's something you can look look forward to I think we did cover some good stuff we went through a couple different multi- agentic flows we went through this idea of Leverage we went through a couple pieces of news and then what I'll do for you is I'll put all this stuff on the site like I mentioned so um make sure you go to Parker rex.com blog you can see all that there and then if you want an inside look into some of the stuff that I'm doing you can go to uh School page it's free looks like we got another member that just joined Coban O'Neal um and I'll be putting this will be free I think the way that I'm going to go about doing this is that this will be a totally free resource but then when you go into the classroom this is how I make money this is where you know if you want the better stuff then you can pay for it but by making it free I think I'm going to build better Goodwill with an audience um and then over the long term my brand will grow a lot faster it's literally what Neil Patel said where he's like the way that I blew up my brand was I made tons and tons of valuable content that was free and my free stuff was better than what people charge for and then he also had an info product business he did mention that his info product business Peak at 600k a year but like if I get to that I can worry about it then and then take that 600k a year and put it into um different different Avenues so that's it for day uh that's it for today and I'll see you in the next one yeah oh I got to make a thumbnail uh the thumbnail is going to be me looking at an agent which is going to be like right here so