Getting the transcript
Reading the captions from YouTube. A video nobody has opened here before takes 10 to 30 seconds; this page fills in on its own.
Getting the transcript
Reading the captions from YouTube. A video nobody has opened here before takes 10 to 30 seconds; this page fills in on its own.

Sharbel A. · @sharbelxyz
This video has no Most replayed graph yet: YouTube shows one only once a video has enough views. These are the moments viewers replayed most in Sharbel A.'s most watched videos.
Most replayed moment at 13:32
3.9x that video's typical replay level
built-in memory, which is great. Here is how I would think about memory providers. Memo is interesting if you want a dedicated memory layer for personalized AI agents. It focuses on extracting, storing, linking, and retrieving memories efficiently. Their
Said at 13:25
Most replayed moment at 6:37
5.8x that video's typical replay level
trading strategy. Tests it. If it's better, it keeps it. If it's worse, it discards it and tries again. So, that's exactly what I built. Okay, here's the system we have at play. I gave it two years of crypto data,
Said at 6:31
Most replayed moment at 1:53
3.7x that video's typical replay level
let's install it together. Okay, so for step one, we need to actually start by installing Bullpen's CLI so that we can actually do everything that we want to do. And for that, let's first open Claude and put in the dangerously skip
Said at 1:47
The graph counts replays. It does not show where viewers stopped watching.
Words
2,293
Runtime
15:26
Speaking pace
149wpm
Reading time
10min
149 words per minute, below the 160 25th percentile of 349 measured videos. That distribution comes from the 349-video hook study.
Opening (first 30 seconds)
So, I've been building AI agents for the last few months. And I'm not talking about the obvious stuff like AI writes my emails or whatever. I mean actual agents that are making me money. And I want to show you all seven of them because honestly, by the end of this, you're going to understand how insanely broken it is that we're not all doing this yet. We're talking thousands of dollars a month,
75 words, the words spoken in the first 30 seconds at 149 words per minute.
Free, no signup. See how the first 30 seconds hold attention, with rewrites.
Sentence shape
| Measure | This transcript |
|---|---|
| Sentences | 198 |
| Average words per sentence | 11.6 |
| Longest sentence | 40 words |
| Questions asked | 5 |
| Sentences containing a number | 40 |
Most used terms
Filler phrases
26 in total: I mean 6 · basically 6 · literally 6 · actually 3 · kind of 2 · like 2 · you know 1.
A literal whole-word count of the same phrase list the Prepublish browser extension uses, so a phrase inside another word is not counted and a phrase used in its ordinary sense still is. It is a count and not a judgement.
Free, no account. See where attention is likely to drop, with a rewrite for each weak line. The free check shows the scores and the one issue costing the most. Or run it on the words above first.
Free · No login · See a sample audit first if you prefer.
What this transcript is
Every word below is the caption track YouTube publishes for this video, pulled from the video itself and reproduced unchanged. It is not Prepublish's writing, not a summary, and not a re-transcription: it is the video's own published captions. English captions, generated automatically by YouTube, in the video’s original language. Source: the video on YouTube. A channel that would rather this page did not exist can ask for its removal through the contact page, and it is removed.
So, I've been building AI agents for the last few months. And I'm not talking about the obvious stuff like AI writes my emails or whatever. I mean actual agents that are making me money. And I want to show you all seven of them because honestly, by the end of this, you're going to understand how insanely broken it is that we're not all doing this yet. We're talking thousands of dollars a month, but here's the thing that nobody tells you.
My trading bot, for example, is wild. I have a bot that's trading on Polymarket and it's making about 4 to 500 a month just by itself. But that's literally number seven because before we get there, I want to show you six other ways I'm using AI that are honestly way more profitable for most people. Let's go. Number one is the simplest, but probably the most mind-bending when you really think about it. I have an AI agent called Max.
Max is basically my employee. And here's how you train an AI employee. You show it what to do. You break it down step by step, and then you let it handle it. Here's an example of what I'm talking about. I needed to find founders who just raised money. I needed to find their CMO and then reach out to them. So, I broke it down to Max. I asked Max, "Search Twitter for raise announcement from last week. Filter for AI companies.
Find their websites. Locate the CMO on LinkedIn. Get their DMs. And then send them a message about our service." You see how every single step of any task, any goal that you want to achieve, can be broken down. And if you can break down those mini action steps into each and everything that you want, then that thing can be automatable. So, I taught Max how to do it. Now, Max just does it. And when something breaks, he doesn't just panic, quits his job, he figures out what went wrong, and then fixes it.
Before AI agents, a good VA would cost you two to $4,000 a month. A great one might cost you 10,000 a month. Max cost me about $200 a month through Anthropic subscription. And the crazy part is Max doesn't just do what you tell it to do. It finds problems you didn't even know you had and goes, "Oh, that's easy. Want me to fix it?" I mean, there are problems I've been stuck on for literally years. And once I installed OpenClaw and I started chatting to Max, Max looked at it literally once and was like, "Yeah, that's simple.
I can handle it." And it literally just did. Number two is where things get really interesting. I have two agents that just make me money directly through content. One is Sage, and Sage lives on Twitter, and the other is Nova, and Nova lives on YouTube. They're two different agents with different jobs, but they kind of both do the same thing. They figure out what content actually works for each platform. Sage has basically memorized month and month of my tweets.
Sage knows which posts get the most engagement, which ones flop, what the patterns are. Every single day he's looking for what's trending in my niche. He's analyzing against everything that I've ever posted, and then he suggests the top 20% of the ideas that would probably go viral. It also filters out the bottom 20% that nobody would care about. Then, when it has that top 20% of tweets, it sends them to me. I either approve or reject them.
You see those little green and red button. And if I reject a tweet, then it will simply ask me for a reason why I'm rejecting it so that it can learn from my direct feedback. And Nova technically does the exact same thing, but for YouTube. It analyzes every single video in my niche, finds the ones that are doing two to 10x better than my videos. Let me go ahead and show you. It finds the outliers that are performing two to 10x better than their channel average, and it figures out why they're performing better.
Whether it's content gaps, thumbnails, pacing, topics. And here's the result part. Which is kind of insane when I say it out loud. I was stuck at 800 subscribers on YouTube for an entire year. I completely plateaued. But four weeks after starting to work with Nova, I'm now at 4,500 subscribers. That's basically over 2,000 subs in a single month. I think this month I've gained 2.6 thousand. And I mean, you can just see for yourself the difference of when I started working with Nova.
That's when it started spiking, by the way. Money-wise, in the first 10 days of being monetized on YouTube, because Nova got me monetized on YouTube, I've made $253. That's about $25 a day, which if I manage to keep that on track, would be $750. Things get even more insane on Twitter, because over the last month, I've managed to make 550 last in the last 2 weeks and 830 in the 2 weeks before that, which is about a thousand four hundred dollars approximately.
So, these two agents alone, just in direct creator revenue, are making me 2,150 approximately. That's almost a salary. Depending on where you're from in the world, $2,150 every month is considered a job, a full-time job. And it's required me to literally just approve, reject suggestions, and film them, and then hit post. Number three is where the agents stopped being just about making me money and started making my business money.
The same agents that I was just talking about, Sage and Nova, but now with a different job. Instead of chasing whatever's trending, they're looking for content that speaks directly to the founders I want to work with. Unfungible is our agency, and our main service is called the timeline takeover, where we help founders build their personal brand and create viral launch moments. Oh, by the way, you can go to founderfunnel.com if you're a founder and want to find out more on how we can help you.
But anyway, the agents aren't just looking for viral content anymore. They were looking for content that would speak to founders, early-stage builders, creators trying to grow. Basically, our ICP. And it works because people see the content, they then reach out to me, "Hey, this is amazing. How did you do this? Can you do this for my product launch?" I mean, a perfect example is I built a web app in 1 day using Cloud Code and Open Claw. 1 day.
I posted about it on Twitter, it went viral, and three founders immediately slid into my DMs asking if we could do the same thing for their launches. Those became leads and those became conversations. Some of those even became customers. So, the same content agents that are making me 2,100 a month in direct creator revenue are also funnel toppers for our services. Wait until you hear about number four. Number four is where AI agents start getting really ROI positive.
Because see, I built two products with AI agents. One is called Unfungible Clips and the other is Max HQ. Both were built almost entirely with Claude code and open claw. I have another agent called Pixel that handles web development. You can see the little guy jumping right now. That's Pixel. And here's the cost comparison of Pixel. Because when I asked dev shops what it would cost me to build Unfungible Clips, the quotes came back somewhere between 10 to 20,000 dollars.
That was just for an MVP. They said and quoted that they charge another two to 5,000 for each new feature that we wanted to add going on. But the actual cost for me to build this out myself using Claude code was $200 a month for my Anthropic subscription. That's it. The same 200 I'm paying for my content agents. The same 200 I'm paying for all of my agents. It's just the one $200 a month bill that is able to make and create all of this wealth of value for me.
So, Unfungible Clips is now generating about 30,000 a month for the business and it's not done. Every single day Pixel looks at the app. It figures out what we could do better. It scrapes competitor features and updates. And it suggests improvements. I approve them or I don't. If I do, then Pixel builds them. Next day, we iterate. Same thing. New features. They just keep going and going and going. Look, a human dev team would cost us thousands thousands of dollars a month.
Pixel costs basically nothing. And he doesn't sleep, never burns out, never leaves to take another job. Is it perfect? No. Does it get you 95% of the way there? Absolutely, yes. Number five is where things become a little logistically insane. Hear me out. We have a ton of clients. And clients need a lot of manual work. I mean, there are client portals, campaign reports, meeting scheduling, guest booking for our spaces, launch research, all of it.
So, we built agents to handle a lot of the grunt work. Before AI agents, we could have one customer success manager handle about 20 clients. They'd be completely slammed at that point. Reports writing alone for 20 clients would eat a whole day for that product manager. After we automated all the grunt work, one customer success manager can now handle 35 to 40 clients. That's almost double the capacity without hiring anyone new.
I mean, think about that. Our team stayed the same size, but we doubled how many clients we could serve. That's just money on the table that we weren't picking up before. Number six is something my co-founder Leon built out. Lead scraping. It's us basically searching Twitter and LinkedIn for founders who just raised money, founders who are hiring, founders who are announcing launches. We find them before they're flooded with agencies.
Then we scrape all of it with agents. But here's the cool part that most agencies don't do. Every single sales call that happens, our AI scores it. It scores the closer on a scale of 1 to 10, and it gives a very specific feedback. You missed the budget question. You pivoted too early. You didn't ask for the deal twice. It gives actionable feedback for the closer to learn and improve every single time. The closers get better because they know exactly what they're doing wrong.
The hit rate goes up, and the deals get bigger as a consequence. Leon actually has a YouTube channel where he breaks down how this works. So, definitely go check him out. The guy's insane at sales. Okay, number seven. This is the one that seems to always get the most attention on my YouTube channel. I have two bots on Polymarket. One trades 15-minute markets, one trades 5-minute markets. Both are looking at the same thing.
Bitcoin, Ethereum, Solana, up and down markets. Very simple prediction markets. These two bots alone make about 4 to 500 a month, but here's the thing nobody talks about. Those two bots are actually the least set and forget bots that I currently have. I remember when I set them up, I would just obsess over every single trade they were making. I would check the bots constantly, stressing about whether it was going to blow the account or not.
So, I ended up building another agent to monitor it every 4 hours and alert me if something's wrong. Here's the agent, Knox. You can see him hopping, bouncing, enjoy having to do all this glorious work. But, then I also built an auto researcher that looks at market conditions, historical data, sentiment, and constantly tweaks the strategy that we got going on. That agent was inspired by Karpathy's Auto Researcher. Just keep improving, keep learning, keep adapting.
That's the whole goal of that agent. The big lesson here is the first and biggest win with AI isn't some insane wealth hack. It's automation. It's freeing up your time from stuff that doesn't matter, so you can focus on stuff that does. Every experiment on top of that is literally just a bonus. So, my friends, to recap, AI agents. Number one, employee. Saves me thousands of dollars a month and solves problems I didn't even know I had.
Number two, content agents making me 2,100 a month in direct creator revenue. Number three, same agents pulling inbound leads for our business and services. Number four, web development that would cost 10 to 20,000 dollars happening for 200 a month. Number five, doubling our client capacity without hiring anyone new. Number six, better sales through AI feedback on every single call that we get. And finally, number seven, trading bots making 4 to 500 a month.
That brings our total to something around 30 to 40,000 a month total revenue being powered by AI agents. I'm not a genius. I'm just building tools and letting them work for me. If you want to learn how to build this stuff, I have a whole bunch of tutorials out there on my channel on how to set up these agents. So, subscribe if you want more. Hit the bell and, you know, all that stuff. And if you're curious about any specific one of these, just drop a comment.
I'll make a deep dive video on whatever you want to learn about. I'll see you in the next video.
The words are the caption track's own and nothing is reworded or re-transcribed. Paragraph breaks are placed between sentences so the text reads as prose.
Free tools for your own script: paste a draft and see where it stands before you record it.
Paste your draft and see where viewers are likely to drop off, with a rewrite for each weak line.
Paste the first 30 seconds of your own draft for a hook score and rewrites.
Check your draft against YouTube's advertiser-friendly guidelines before you record it.
Read this channel's public videos and transcripts, and download a writing brief for it.