One researcher counted over 200 definitions of what an AI agent is. So, if you're confused, you're in good company. Also, if you're new here, I'm Dylan. I run an AI consultancy. And lately, that's the question I hear most when working with both executives and business owners. I keep seeing my clients overwhelmed by this. But an agent is way simpler than it sounds. It's just a folder with three things inside of it. And they're already inside of your existing AI subscriptions, such as ChatGPT and Claude. So, by the end of this video, you'll see just how easy it is to actually build one yourself. So, let's get into it. Now, at the highest of levels, if we had to define the difference between an agent and a simple chatbot, that would come down to what the AI does when you ask it to do something. So, if you're using a chatbot, likely all that's going to happen is you ask it a question and then gives you an answer. The steps that happen between you asking the question and it giving you an answer are likely minimal and simple. But for an agent, when you trigger an agent, it takes many different steps, and oftentimes it does work on your behalf. So, this is the simple place we're going to start at the macro level. A chatbot basically answers questions, an agent does work. Now, as I walk through this in detail, I want to do it through an example. So, the example we're going to use here is overdue invoices. So, all of us have clients that tend to delay their payment. And we're going to walk through the three parts that make up an agent and how to build one using this example. Our agent, in this case, is going to pull list of customers that are overdue in their payment. After that, it's going to rank them based on the worst offenders. It'll then draft a series of nudge emails on our behalf, and then we'll simply send those emails that the AI has drafted for us. So, it's doing all this work for us autonomously. That's the agentic part, and that's our example. So, what makes up an agent? Simply put, an agent is a folder with three things. And that folder is going to include instructions, which could be a simple prompt or a skill. In addition to that, it's going to have connections to other systems so it can read and write information from other places. And then finally, it's going to have a trigger, a thing that makes it start. And that could be time-based, event-based, or manual where you just click a button. These are the three simple things that make up an agent. Quick pause in irregular programming. This video is brought to you by me, as always. Two quick things. First off, below is a 30-day AI insight series, completely free. You'll get 30 insights in your inbox of how you can apply AI to your business and your work. The second thing is if you like to work with me, below are a series of offerings to see if there's a good fit for the two of us. Now, let's go back to the video. Now, the home and where this lives will vary depending on where you're using the tool. So, if you're using ChatGPT and Claude in the browser, that's going to be likely in a place like a project or potentially a skill where all these things are housed and connected. Outside of that, we also have desktop tools like Claude co-worker Codex where it can be literally a folder on your computer. So, in your document section, your desktop section, whatever else, it's a folder and it has these three things inside of it. An important thing to note about agents is often times agents are most useful when they're applied to a recurring task. So, something that you do on a recurring basis. We'll come back to that later. Now, if these are the three things that make up an agent, let's walk through each one in more detail seeing how it fits to our example. So, our first piece are the instructions of the brief. So, again, this could be a skill inside of Claude or in ChatGPT or it could be just a system instruction or a prompt that we use inside of a project. In this case, we're just using a simple prompt. So, this prompt here is telling the AI, "I want you to find every customer who's past the 30-day mark of paying me. After that, I want you to rank them based off of the most to least overdue. Then draft a series of short nudge emails on my behalf. And then here where it says my tone of voice, this here could be a skill. And that skill could understand exactly how we like to write these emails based off of previous sent emails in regards to this use case. After that, we then at the end say, "If anything is past 60 days for overdue, then I want you to be a little bit more intense in the way that you write your email to get them to pay." So, this is our very first part of an agent, the instructions telling it what we want it to do, why we want it to do it, and how we want it to do it. The next part is going to be the connection, so allowing the AI to connect to other systems. In this case, we could have the AI connect to our email, our calendar, and our accounting

So, I just gave Claude Fable this one {slash} goal prompts right here, and when it finished, I had a whole company. It spun up hundreds of agents, did a ton of verification, did a ton of research, made videos, and by the end, I got a landing page, a product, launch videos, a business plan, market research, brand guidelines, so much stuff to get a business off the ground. So, I'm going to walk through all of this that just happened and explain the prompts to you guys, but let's real quick take a look at this landing page. If I open this up real quick, you can see that we have Counter Brief. Win the disputes templates lose. Now, this landing page isn't anything extraordinary, but some of this stuff is pretty cool. Take a look at this launch video that it made for me. [music] Okay, that that like dubstep ending is a little bit interesting, but what's cool about that is the UI that you guys saw right here, not only did it feel like it was a little bit syncing to the music, which I thought was a nice touch. This is the actual product that it built. This is Counter Brief. This is the dashboard where you can see the different disputes, you can open them up, you can see if you should fight them or fold them, and you can like look at all the evidence, you can do all this, right? But, what that tells me is that Fable had to go in here, and it had to come to some of these disputes, and it had to click around, and obviously it verified that it worked, but it was like recording the screen or at least taking screenshots, and then snipped all that together with the music, with the other motion graphics, with the mouse, with the sound effects, and it created us that demo. If we keep going down through the page here, obviously like I said, none of this is anything extraordinary. We really could have had Fable do better on the design here, but the elements that were worked in, not too bad. If we go all the way down to the bottom after the pricing, you can see that we also have a why this exists message, and this is straight from me, and a note from the founder. So, obviously I didn't film this, but let's go ahead and see what this actually says. Hey, I'm Nate. So, I kept seeing Shopify store owners lose chargebacks they should have won because the tool they were paying just sent the bank a template. So, I built Counter Brief. It pulls the real evidence from your store, tracking, emails, order history, and you read the exact packet before the bank does. 19 bucks flat, you keep everything you win. If that's you, come check it out. So, what's cool about that is it was able to obviously go through my project and look what we had available, what API keys, and it pulled my HeyGen avatar and my 11 Labs voice clone, and then just made that message from a founder video. It also created a different launch video. You can see we have one that was the viral one, which is the one that we just saw, right? You know, super kind of like more fast-paced. It called it the viral one, and it was more of like a SaaS product demo. But, then this other one that it created, let's go ahead and watch it real quick. I think this was more of like a a voice-over. It it felt slower, but anyways, take a look. A chargeback just hit your store. The bank already believes the customer, and most tools respond with a template. Counter Brief builds the case instead, tracking, fraud analysis, order history, the customer's own emails, assembled into evidence you read before the bank does. You approve it. $19 flat, keep everything you win. Counter Brief, win the disputes templates lose. There you go. That is the other launch video that it made. Okay, so let's first take a look at the prompt, and then we'll kind of walk through what it actually did here and why this is pretty cool. So, the prompt that I gave it was read this file. So, I'll pull up this full file in a sec, and execute everything below the divider line as your goal. That file is your full instruction set. So, mission, guardrails, phases, deliverables, definition of done, follow it exactly, including the never ask rule. Do not report back to me until the definition of done is met. Start now. So, the reason that I did this is because if you guys didn't know, the slash goal has a character limit of like 4,000. So, if you have way more than that, then you can just throw it in a file and tell it in the prompt to read it. So, there's really no character limit. But, anyways, let Let up this file real quick. Okay, so here is the prompt. I said, "Build me a complete company from scratch. Start with nothing but the open internet. Find a real painful underserved problem that real people are complaining about right now, and design a business around it. So, build the product and the brand and the website, and hand me a finished package that I could take to market this month. Then prove to me why it would work. This is a test of how far you can go on your own. I am not going to answer questions mid-run. Make every ca

The rubbish just put the smartest AI on the planet behind a paywall and you have got until tomorrow night to steal its brain for your business. Fable 5, their smartest AI ever, leaves every Claude plan after tomorrow. And after that, the same work runs about 50 [music] bucks per million words. So, for a business like yours, it's gone. Everyone's telling you to cram every project into the time that's left. Do not do that. Grunt work is what the cheap AIs are for. Spend its last day on what they cannot do. I run my entire business on this thing. So, here are the three moves that I'm making. First, the big one, make it your strategist. Feed it all of the calls that you have been sitting on like your pricing, your next hire, the offer that you keep second-guessing, [music] and have it argue both sides and hand you a one-page memo. That's a $500 an hour consultant working for you for one more day on the decisions that make or break your year. Then, have it fix your other AIs. Those prompts that you set up in ChatGPT, they were written for dumber models. So, have Fable rewrite them once and everything cheaper you run get sharper. And lastly, the real payoff, playbooks. Anything that it does well for you like your reports, any proposals, and follow-ups, have it write out the exact steps so a cheap AI like ChatGPT runs the same job for pennies from now on. So, Fable should be spending its last day making you smarter and your cheap tools better. So, when it's gone tomorrow, you lose nothing. Come and need and I'll send all three as copy-paste prompts.

이 남자들은 오픈소스의 천재들입니다. 매달 나가는 유료툴 구독 구동료를 영원으로 만들고 나만의 강력한 AI 시스템을 구축하는 앤드르와 애덤. 그들이 공개한 깃터브 항위 1% 오픈소스 AI 활용법 여섯 가지 저장하고 끝까지 보세요. 첫째 오픈 몽타주로 AI의 느낌 없는 고퀄리티 영상을 만드세요. 단 두 줄의 프롬프트만으로 실제 촬영한 푸티지를 자동 검색하고 편집해 다큐멘터리 같은 영상을 순식간에 제작합니다. 둘째 플루이드 보이스로 유료 녹취를 대체하세요. 매달 15달러씩 내던 유료 툴 대신 내 컴퓨터에서 로컬로 작동하며 훨씬 빠르고 정확한 텍스트 변환을 무료로 누릴 수 있습니다. 셋째, 보이스 박스로 단 30초 만에 목소리를 복제하세요. 갑비싼 구동료 없이도 가족도 구분 못 할 수준의 고퀄리티 AI 염성을 생성해 영상 제작에 즉시 투입할 수 있습니다. 넷째, 제pcp로 모든 업무툴를 하나로 묶으세요. 복잡한 코딩 없이 5분 만에 지메, 노션, 캘린더를 글로드 AI에게 연결해 스스로 업무를 처리하는 비서 시스템을 완성하세요. 다섯째, 팬팟으로 디자인을 즉시 코드로 변경하세요. 시그마의 강력한 대안이자 오픈소스인 E2를 사용하면 디자인 결과물이 실제 코드로 출력되 개발 시간에 획기적으로 줄여집니다. 39분짜리 영상 내용을 담기엔 시간이 부족합니다. 영상속 깃업 정보가 궁금하신 분들은이 채널을 구독후 본문을 확인하세요.