The AI SKILL.md That Gets Millions of Views Per Month With $100M+ Leads w/ Alex Lieberman

Fix that. Yeah, try to go for it. If you can't fix it, that's okay. But I know we're live right now. We're it looks good. Okay.

All right, guys.

Sweet. We're live. We're live right now. And those of you that are coming in right now, just type in where you're coming from. Alex and I will get started in about a minute or so. But today we're going to be talking about just how you can actually drive more business revenue using AI because a lot of people that share this stuff on like Twitter and the ones that share it on YouTube as well, it's more like, oh, this is the craziest thing. Look at what this app I built. Oh, this is the craziest agent right now. Nobody really talks about driving business revenue, which is weird to me because everyone's talking about the AI build-out. I think it's incumbent on people like you, Alex, to be able to share what's actually happening because you're actually in the trenches right now. Maybe those that don't— I hate to do these intros, but again, Alex and I, we've known for each other, but you should intro yourself. I'll do the intro for you real quick. From my side at least. But Alex and I have known each other for a bit. He co-founded Morning Brew and then now he's got 10X and they— I'm just going to call you guys an AI transformation company. And he's been producing a lot of great content. He's moving at a very fast pace right now. And we thought we'd do a podcast here and then next week I'll be on his show. But we're just going to just nerd out on what's actually happening with AI in business. We'll talk about the latest. with autonomous models, sorry, autonomous agents, the latest models and all that. Alex, I don't know if I missed anything. You can feel free to add to that.

No, you did great. Pumped to do this. I think the number one thing, you're actually in it, but I have this text group with some executives who are trying their best to be on the frontier of this technology and building with it. I think the number one Frustration I hear from people in the field right now is that like the hype to what's real gap feels massive right now, right? Like you have all these labs that are coming out with their new models, you have AI companies that are raising an insane amount of money, but then you go into like a mid-market company or an enterprise and you ask like, what do they actually have in production, whether it's like single-player workflows or like cross-functional workflows that have truly moved the needle for their business, it's like crickets. And so a lot of these executives I talk to are just like, I want to see where the magic is. Like, I want to see the magic so I can believe in the magic, so then I can like put resources into building it for myself or for my company. And so I love that you're doing this, and I'm trying to do the same type of stuff, which is just like share the most valuable workflows that are actually creating leverage within your your work and within your company and then learn from each other while doing it.

Yeah. And maybe we can talk about these text groups and also in-person groups as well. But I think when you and I spoke maybe a couple weeks ago, you really only have 2 major roles, I think, at your company right now. So just if you can explain kind of the reporting structure for you, because you co-founded this business, but you are only focused on very specific things right now. And I think it's good for everyone to know and learn how that works. Yeah.

Yeah, I co-founded 10x with Arman Hezarkhani about 14 months ago. We had met right around the pandemic. He previously, before 10x, had started an AI fintech company called Partheon, and I had invested in it. I was an advisor to the company. We together developed a lot of our views on AI prior to starting 10x together. The way we think about the business is I basically focus my time on 2 things. One is being the CRM to the company, and 2 is co-creating our strategy and long-term vision with Armand. And the reason for that is, I think when I was a co-founder at Morning Brew, I had this ego about needing to do everything and feeling like my worth was directly correlated with how many things I was able to do as an entrepreneur. It looks totally different in this next phase, and I think that's for 2 reasons. One is, at some point, I was like, this is stupid. I should just do the stuff that's highest leverage and most enjoyable to me. But also, I think that part of the reason you find a great co-founder is so that You trust that they're going to do amazing work on the things you're not spending your time on. The way it works is like, I am deeply focused on driving enough demand for 10x such that we will never have to pay for a dollar in marketing or build a sales org. My co-founder, Arman, is incredible at running delivery and sales for the organization. Then together, both of us, like our goal is to work on this company for the next 20 years of our lives. we will pan out to talk about building this modern-day Bell Labs, which is our vision for the business. But yeah, I would say 50% of my time right now is spent creating content myself, and also building out the media operation of 10x. So I obviously draw a lot of inspiration from MorningBrew in building that. And the one other thing I'll add that I think is helpful context is we have not raised a dollar at 10x. We are completely bootstrapped. And so I think What I'm basically thinking about is how do we build a small but mighty content operation here that to the outside feels like we have a 100-person media company, but then when you actually come like behind the scenes at 10X, really it's just a few creators, Alex, the employees of 10X, and then really good infrastructure that gives us leverage to create content. And so one of the things I'm going to take you through today is this content machine, which is the most frequent directory of skills that I use in my work, but the plan is also to use it as this centralized system that any creator at 10X can use as well to create leverage for us.

Yeah, I think that's a good starting point. So maybe before we jump into that, so you mentioned, because what I'll call it here is this, you look at Dharmesh Shah, CTO, co-founder of HubSpot, he has no direct reports, okay? And I look at my podcast co-host for Marketing School, Neil, he has no direct reports. That's how he started the business, right? And Neil specifically, he's just the distribution. I'm saying just, but it's a very big thing. He's the distribution machine.

Yep.

And he does go in and kind of handle strategic work and play free safety. But the rule is no direct reports. And so is that kind of what you have to do? So you're driving distribution. You have no direct reports or do you have any?

The only direct report I have is our one creator who's full-time on staff right now. So I think ultimately the only direct report I think I'll have long-term is if we build out the media operation and end up having Like any sort of kind of person running our media, they'll report to me and then all creators will report into them.

Yeah. And then what's the other 50% of your time going to? Is it mostly like strategic?

It's basically whatever the biggest fire is at the time. That's where like I'm focusing my time. So whether that is a client fire, whether that's a fire related to one of our partnerships, it's that. But honestly, most the other— so the other part of that. So the 50% is creating content. But then the way I think about it is in order to create great content, I need to both be building and be out in the field. So it's actually, if I think about it, probably 75% of my time is related to the content surface area, which is me either trying stuff out in Codex or in Claude Code or whatever tools exist. And then it's also literally, I try to reserve, I don't know, 5 hours a week minimum to just getting on calls with people who are in the trenches and have them just teach me what they're doing within their work. And then if you've looked at like my X or my LinkedIn before, that usually turns into a post about all the things, like field notes of everything I've learned from this person during our call.

Yeah, that's good stuff. All right. Well, I mean, people came here for the sauce. So what's the sauce around the content machine? Because you shared this with me and I'll talk about kind of what I did with it afterwards because you're graceful enough to share it with me.

Yeah, totally. So basically, okay, here was the inspiration. I would say ever since Morning Brew and even before this, I've always thought about How do you take one piece of content and turn it into as many pieces of derivative content as possible to basically maximize the surface area of content? And the way I actually— I think the way the framework that really stuck with me was Gary Vee posted, I think, like, yeah, like 15 years ago, the content pyramid. And basically was this presentation still online. You can go look it up where basically he shows he has an anchor piece of content. An example for him would be like an hour long keynote he did in Germany, and then he shows these levels of every distribution channel. So every social media channel, email, uh, podcast, etc., and the system he has to distribute content natively in all these channels. And the thing I've always thought about is pre-AI, you need to have a massive team to do that. So like Gary has a massive content org that supports this effort. With AI now, you don't need that massive content org. The other thing I've thought about is If I could build a really good AI-native process for creating content, there's 2 things I get out of it. One is like I'm patient zero, so if it works well for me, I can then have it used for anyone in my company who wants to be a creator. But the other thing is, as we are doing AI transformation in companies, it kind of is all the same process, which is like understanding a process step by step from beginning to end, then starting to ask yourself, if we were to burn this to the ground and build it from the ground up, where am I absolutely necessary? And we'll kind of talk through that. So before I actually, you know, work on it in Claude, I just put together an HTML, like just a website that kind of walks through it. So I'm going to show that quickly. Screen, Content Machine. Can you see this?

Yes.

Okay, cool. So basically the idea of the Content Machine is it is one pipeline that runs end to end or step by step. in the process of creating content. And what you're going to see is every step in the content machine is just the next step. If I think about any process as an assembly line and you have raw materials that you start with and you have finished product that you end with, you have all these steps in between. Every process you can think of as an assembly line. I think about content the same exact way. So basically the way that this pipeline runs is a new addition even since we spoke is creator selection. And the reason for that is I want to, I've always said to the team, internally. I want to turn 10X into WWE for nerds. I want every full-time— I want every full-time employee here to be a creator. And the only way to do that is it can't be a full-time job for them. So we have to reduce the friction of them putting out great content to as close to zero as possible. So here you can select the creator you are. So these are 4 examples right now: Alex, JJ, who's our full-time creator, Arman, my co-founder, CJ, who's one of our engineers, and Sammy, who's one of our strategists. And when you select yourself, Like when you select that you are the creator who's working the content machine, you have a profile markdown file, a style guide markdown file, and a content lesson markdown file. So basically what this does is it identifies like who you are, what your voice is, and feedback that you've given to the content machine in the past that it should keep in mind for any future content you create. Any questions on that before we go to the next step?

No, I think that makes a lot of sense. And what's interesting to me is JJ, I think you hired him as a community manager and then now he's, I guess, It's like, hey, you're community manager, you might as well just be a full-time content creator. I didn't really pick that up. That's cool.

Yeah. And the way we think about content creators at 10x is interesting, which is historically in a media company, to underwrite a content creator, basically you have to believe that you are going to be able to sell enough ads against them or drive enough subscriptions against them to justify their salary. At 10x, it's a little bit different because there is so much appetite for AI trainings and workshops. And that is a fair bit of what we do, at least in the early days of relationships with customers. JJ has become like a 5-tool player where a lot of time outside of creating content and building that he does now is teaching, like leading workshops. And so there's this interesting model, which is if we have enough demand for companies to pay, let's just call it $25,000 to $50,000 for a workshop, you can very quickly pay back the cost of a creator. And then anything they create and any leads they drive on top of that is effectively free marketing. Um, so that's creator selection. Step 1 in the process now. So I've selected myself, I'm Alex. The next is what I call the Oracle. And I think this is one of the biggest value adds in the Content Machine, which is helping you find ideas. One of the hardest things that any creator will tell you that they struggle with is day after day thinking of content ideas. So right now, as we're recording this, I just started a new series called 30 Days of AI, where every day I'm gonna share an insight from the field. Coming up with content ideas for 30 days straight is hard as hell. The Oracle is what helps me here. And so I'll show it in a minute, but basically what the Oracle is, is I have my content machine hooked up to all of my data sources at 10X. So Slack for internal comms and client comms, Notion for call recordings and like our internal documents, Gmail for emails, and then also Internet Reader. So Internet Reader basically goes and looks at a curated feed of X accounts and websites that I've told it will have valuable information around AI. It goes through all these sources and then it provides a list of 10 spikes of what it thinks are good content ideas to work with. So just to like, I can just show you what it looks like in practice. Um, let me share this screen. Uh, here we go. Uh, window Claude. Okay, so if I just do new task and I do, um, Run the content machine. So what it's gonna do first is, again, it's gonna go through the first step, which is like asking me the question of which creator are you? So we'll answer that in a second. And again, as new employees join 10X, they can literally just be added to the content machine and we can have it start by their content. Here we go. Which creator is this run for? Alex, what do you wanna do? So to ask me, like, so say I already had a content idea, so I didn't need the oracle. I could just do develop my own idea, but I want an idea, so we'll run the oracle. So basically what the oracle is gonna do now, like I mentioned, is it is going to look at all of the different data sources we have at 10X, and it has a scorecard for what it considers to be a good content spike. And a spike is literally what's an idea that it thinks would be interesting to my audience. And then it's gonna provide a list of the top 10 spikes from the last 24 hours that it found. And someone just asked, is all of your work through Claude, or do you have any agent harnesses? This is all through Claude. There are no agent harnesses here. Like, again, this is— It's a harness. Yeah, I guess, I guess it is a harness. You're right. And so it's searching all of these sources. The other cool thing is something I've thought a lot about is when it shares these different content ideas, if it shares 10 of them, there's only gonna be one that I talk about today, but there could be 4 other good ideas that I wanna use in the future. So what I also built into the content machine is something called the Vault. And I'll show it in a second, but the vault is basically just a Notion database where every content spike from the day gets added into this vault. See here, it says, let me dedupe against the vault and grab a couple of the freshest angles. So then what can happen is any of the creators at 10X can then go into the vault and search for a good idea, even if I only used one of the 10 that I was served.

Got it. That makes a lot of sense. And this is cool. It's, by the way, I think it's good to call out for those of you that can't see this. I mean, Alex is sharing a skill file right now that's pretty portable. It doesn't have to live within Claude Coworking. You can bring it to something else if you want.

Correct.

You can use it in Codex as well. And that's what I think is really powerful. And even in that chat group this morning, I think the fact that we were just nerding out on how it's really important to have a skill academy or a skill dojo, you can talk about how you guys are doing that too, because I think it democratizes these processes for people. Because fact of the matter is when you're a company, people talk about processes and repeatability, but nobody reads the processes. But people are forced to use these skills in a sense if you really want to get everyone up to speed with AI, I think.

Yeah, 100%. There is this article, or not this article, this video interview that Claire Vo, who has the podcast How I AI, was interviewing the CEO of Sendbird. I was watching that episode and basically what they have at Sendbird is, because one thing I think about a lot is how do you turn AI into a compounding asset in your company versus a lot of people just building in silos because that is what it is in most companies right now. And I think one of the ways you have to do it is with what Eric is talking about, which is the Skills Dojo, or the guy at Sembird has like a similar type of thing, which is what is a central repository that has approved skills for your company? So let's just say your company has 30 salespeople and there's a seller that has built an exceptional skill for, I don't know, sending follow-up emails or scanning scanning interactions in Slack to measure client health or anything else. What usually happens in a company right now is it's not centralized. Other salespeople build their own versions. They're not as good, and they're getting less performant skills. So the idea is a central repository that has your skills. What the folks at Semper did that I think is really interesting is they have these things called quests where any employee in the company can post a process that they have an issue with or they just think is not good. And then what happens is anyone can go in, any engineer or any person in the company can go in and accomplish a quest, like basically build the agent, the skill, the plugin that solves the problem that the person posted.

Yeah.

And then when that quest is complete, a few things happen. One is that is then posted into the repository, into the Dojo, so anyone can use it. Anyone can improve the thing in the future. So like future quests can just be add-ons to previous quests, and the company built kind of like governance, data controls, and security around it. So even if you're not an engineer who doesn't understand, like, you know, concerns around data leaking or cybersecurity, it's been hardened so anyone can build, and kind of this like package is wrapped around anything that anyone builds within kind of the central repository.

Yeah, I think the walk and the run before you go through the screen over here on that one is like, I was just talking to my team this morning about gamifying it, right? So you can have the quest and then you can upvote it almost like it's like Reddit. And then the ones that get the most upvotes, like you put like a bounty on it. And then I can imagine that you do these micropayments using stablecoins or something where the higher the upvotes, the higher the challenges on something, the more someone gets paid out. And then you can see a leaderboard of how much money people made. So I could see the future of organizations as being super gamified.

Yeah, totally. It reminds me of what Sahil Lavingia did at Gumroad for a period of time, where their entire Trello board or whatever was effectively public, and any engineer could just basically take down a bounty and get paid for it. Yeah, I think a similar thing can happen for sure.

Cool. All right, let's get into this.

Basically, what happened is the oracle ran, it scanned Slack, it scanned Notion pages, so recent pages and meeting notes, it scanned my Gmail, so sent and important inbound, it did the internet reader, which is the live X watchlist and any news from like the labs, from Wall Street Journal, McKinsey, Bain, BCG, et cetera. And then it ran. So these are the spikes. So here's an example. One was around Anthropic just filing to go public. So then it shares what the idea is, why it matters, what the hook could be. Another one, and this is what I'm planning on doing for my 30 Days of AI idea today, is talking about context management and how to do good context management and why it's so important. And then just to show you what happens, Again, we have the vault hooked up. So if I just stop sharing this and I share the vault, basically all of these spikes are then shared here, right, in this vault. And so then what happens is any of the creators at 10x who are looking for a good idea, they can come into the vault and find an idea that works for them. So here's one, turn every employee into a creator. It's just like me talking about this idea of distribution as a moat and how you turn people into creators. Anyone can pick up this idea. Let's go back to the next step in the process.

I think when you showed me this stuff, for the LinkedIn post, it's not you writing it. Is it writing in your voice or is it someone from your team taking an AI skill of your voice and writing it or how's that working for you?

Yeah, totally.

You're not doing it.

Yeah. So the whole beauty of this and the reason I think Content Machine solves this issue and this complaint about AI slop is my general view on any kind of AI-enabled process in the future, but especially this one, is I need to be involved in the first mile and the final mile. So the first mile is I need to provide the direction and the idea that I choose. I need to provide the thoughts or the context, and then I need to be involved involved in the QA. Everything else should be handled by AI. So let me just— I'm going to just show you what the next step in the process is, and I'll demo it quickly just so you can see what it looks like. And it'll answer your question about how is the content actually written. So if we go back to the breakdown of the content machine, so we just did the Oracle. What happens now in the Oracle is like, if I select an idea, let's just say I selected the idea around the Anthropic IPO. And I like, I'm like, I don't actually know all that much about the Anthropic IPO. I can run a research step where it builds a research brief before step 2, which is where I get interviewed by a world-class panel. So this is kind of the key of not creating AI slop, because what happens is when I move on to step 2, a panel of world-class interviewers asks me questions and pushes me to be more specific, to steelman arguments, uh, to, um, you know, provide frameworks. And so when I do that, that is acting as the meat for any content that we're actually drafting moving forward. And so just to show you what that looks like, I basically constructed this panel of who I think are 6 of the best interviewers in the world: Tim Ferriss, Joe Rogan, Larry King, Barbara Walters, Michael Barbaro, and Howard Stern. Each has a different style, and within the Content Machine directory of skills, there are— each interviewer here has its own skill. So Tim Ferriss has a skill, Joe Rogan has a skill, Larry King has a skill. And those skills basically define what is the interview style of these interviewers, what do they typically push on, what are the questions they ask. And so I'll show you in a minute, but when I pick the content idea I want to go with and I want to get interviewed, you'll see different members of the panel will come in at different times and they will push on the things that they are most known to push on.

Nice, I like it.

So here, I'll just show you what the next step looks like. So let's share.

This.

Okay, so, so I'm gonna go with context management. Let's just go with that. And, and we don't have to go through the whole thing, but let's say— let's go forward with spike 1 about context management, and let's move to the interview panel to talk about this spike. So now what's going to happen is it's going to skip the research step since I said I want to just go straight to the interview panel. And so this panel is going to start asking me questions about context management. And the idea is I'm going to provide all these thoughts. Those thoughts are then going to get saved as a raw Markdown file. And so then when we go to create the first piece of content, it is pulling from everything I've said. So it's never Using to create the content things that I did not say, and I think that's the key. Like when we talk about AI slop, typically AI slop comes from comes from you not actually using your words. Okay, so let's see. Forget the abstractions for a second. Walk me through the most recent time. This week you sat down with an AI agent to do real work. What was the task? What exactly do? Give me the actual.

It actually interviews you.

Oh yeah, it interviews you. And if I don't like a question, I can be like, you know what, I don't love this question. I want it to really focus on the best tactics and best practices for doing smart context management and context hygiene. Can we reorient the interview around that, please? And so what I'll do is I'll get it to ask these questions, and I'll probably spend, I don't know, 15 minutes on it, just yap to texting. And once I do that, Like, that is all I need for a good anchor piece of content. Okay, so now see, it reoriented it. Let's go straight to the tactics. Give me your top 3 context hygiene best practices, the specific repeatable moves. So I'll just be like, okay, first of all, I know a lot of people get worried around compaction. So when the context window gets filled up, whether it's in Claude Code or Codex, people worry that when compaction happens, you lose important information. I think these models have gotten so good, where you should not be worried about compaction and the important essence is actually kept. That's number one. Number 2 is having really standard conventions for how you set up your folder and file system is one of the best ways for good context management because then you can easily point a model to the right folder or the right file and it doesn't need to waste tokens Looking for a diamond in the rough. And then the final piece is, for any kind of net new workflow, you should think about starting a new task and then having kind of like a file that stores the memory of a separate task, but you need to relate back to in this new task. So you start with a fresh context window, and the only thing that the context fills up with is the file that you want it to read.

And you're using Whisperflow for this?

Yep. Just using Whisperflow. And so then what you'll see is I talk through it and it'll think through. The next interviewer won't be Tim Ferriss again. It'll be one of the other interviewers and it'll specifically ask about the thing I answered.

Yeah. I think it's important to call this out too, but before this shows up too, like the fact of the matter is you're not just handing it off. You're actually spending the 15 minutes to co-create the content. Versus it's not just the, hey, just YOLO and go. So that's important.

Yep, exactly. And so, I mean, what we can even do is go to what does the next step in the process look like? And I can even use the 30 Days of AI. Day 1 was yesterday, and I did it about software factories and this concept of a software factory. And so I can show how the next step works.

So let me go to— I've got a question here. Yeah. What was the difference of the output from before this automation to now?

So for example, one post a day to day 5. Do you understand the question?

I think I understand the first one. What was the difference of output from before this automation? Yeah, probably referring to quality. Yeah.

So I mean, the short answer is the quality is markedly better. And the reason is twofold. One is Having this interview structure basically just pulls all of my thoughts onto paper. And so when content is created, it is taking my words. And so generally, people do not believe AI is AI slop when your own words are being used.

Yeah.

The second thing is, and this is helpful for me because I have a lot of content online, is like this voice, the VoiceMD file that I have that's part of, again, my profile as a creator in the content machine, has ingested all of my podcast interviews, all of my X posts, all of my LinkedIn posts. So there's just a lot of rich information there. But the other reason this is so valuable and so much better than what I had before is let me just show one piece to this. And my view is any good, and I think we talked about this, Eric, any good AI system has, like, completes a full loop. And so what that looks like for me is, okay, so basically I was interviewed by the interview panel, just so you guys know what happens next. Then I would go to write a piece of content. And when I go to write a piece of content, it asks me basically, what is the anchor piece of content you want to create? Whether it's a newsletter, a long-form LinkedIn post, a YouTube script. It gets written, and then there's a writers council. So these are 6 of the best writers that I look up to. So Morgan Housel, Tim Urban, Greg Eisenberg, David Perell, Sean Puri, and then a slop detector. And what they do is they look at the initial draft that is written by the content machine, and they score it. And if it scores below a 9 out of 10, there are 2 possible paths. One path is if, as they score it, they say, Alex didn't provide enough information here, here, and here. It sends it back to the interview panel where they ask me for more information about the missing information. If they feel like it is just not written well, so it's all editorial changes, then they do a revision loop until the council scores above a 9 out of 10, and that is when they show me the final piece.

I love that. That's great because here's the thing, I've been using these eval panels, and but the thing is, It's auto-scoring its own AI-created content without my voice. The major thing is your voice is included and it's going back and forth with you. I think most people get it wrong. I've gotten it wrong because I just rely on the AI Discord AI.

100%. Then what happens is the final piece of content's created. Then step 7 gets to the Gary Vee content pyramid, which is we got to the final anchor piece of content. Now it's going to create derivative content. So short-form video script, X article, LinkedIn article, 1 to 2 short posts, 1 to 2 short posts, and a playbook for 10x's website. And for each one of these formats, there's a prior of what, what that format looks like that I actually created. So again, similar to how there's a voice.md file, there's a format.md file for each of these formats based on what I've created in the past. The final piece to this Is, and just so you know, like I haven't finished building this part in, but the next, the final piece here is I'm building in the distribution piece where every piece of content that's created gets UTM'd. I can schedule it from here. It gets attached to HubSpot, and we can do attribution straight from the content machine. But the final piece that's really important is where is it?

Ba ba ba ba.

One second. Where is it? There is here lessons. Okay, so basically what you see here is content lessons dot md. Let me see if there's another place that okay, this is it: the learning loop. So basically what happens is after I approve a piece of con or after I get the final piece of content from like the editorial committee who's reviewed the draft, did the revision loop to get it to a nine out of ten. I look at the content and I make my final edits. I feed the final version post my edits back to the machine. And what it does is it looks at the diff between what the editorial committee gave me and what I provided it. It extracts lessons from the feedback I provided it, and it saves those lessons into a contentlessons.md file. So in the future, we loop back up when it is reviewing my content. It has this contentlessons.md file, so it is looking for any of its previous failures from my perspective of where the writing wasn't sharp, and it's making sure it's not making those mistakes again.

Yeah, dude, this is great because one of my friends, we did an AI mastermind a couple of weeks ago, and then he has a skill that helps him find real estate deals and evaluates real estate deals. I'm like, how long did it take you to make this skill? He's like, dude, hours and hours and hours. I'll just ask you straight up, how long did it take you to make this skill? Because I think a lot of people are like, oh, I can probably do this in a couple minutes or so. But how long did it take for you?

I've probably spent minimum 50 hours on this. I've spent minimum 50 hours. And the reason, again, I think it's worth the 50 hours is now I really only focus my time in 3 places: picking the idea, being interviewed about the idea, and providing feedback on the final draft. And my view is everything else gets taken care of for me. Oh, sorry. And the final— the 4th thing that I spend my time on is improving the system. And then because of that, not only am I spending way less time on content, like the thing I posted about the software factory yesterday, previously that would have taken, I don't know, that would have taken like probably 20 to 30 hours to write. It took me like 3 to 4 hours. And the way I view it now is like, I'm also kind of like, as AI helps to write content, I almost think of myself as like a disc jockey, like a DJ of remixing a lot of stuff, Versus just creating a lot of net new stuff. The final piece is, is if I truly want all 40 team members here to become content creators, they're going to need a system like this because they can only spend, let's call it, 2 hours a week max putting content online.

Yeah. And so, by the way, so let's call it like an 80 to 90% reduction, but it still takes a good amount of time, 3 to 4 hours. By the way, it used to take 3 to 4 hours maybe for like a, like a, you know, 1,500-word, 2,000-word blog post, but Even then, if you really wanted to go deep, you would spend the time that you're talking about. How much reach do you think you're getting right now across the board with your content per month right now?

Yeah, I mean, I'll just use the example of just yesterday. Yesterday I posted about, again, 30 days of AI, and I can talk about why I'm doing that. Regardless, I would have posted about AI every day for the next 30 days. But when you name something and you make it a contained series, People love following along the journey. So I posted about 30 Days of AI yesterday and I got 10,000 followers in the last 24 hours.

Wow.

And that is just because people, people love story arcs. People love following journeys, especially journeys where there's something in it for them. So just in the last day from that post and my day one post about 30 Days of AI, that is just 1.5 million impressions in the last 24 hours. And so this goes back to like, why do I spend all my time on content? My view is 10x is competing. One way to think about it is we are competing against some of the biggest management consulting firms in the world. McKinsey, Bain, BCG, Accenture, Deloitte, keep the list going. We do not have 50 to 150 years of history that will just automatically make a CEO reach out to us. We don't have— I'm not a partner at McKinsey who has 20 years of history with the CEO of a Fortune 50 business. The only way we are going to build trust and legitimacy with these firms is either through partnerships, like the partnership we have with Anthropic, referrals from our existing customers, or content that shows that we are an authority on doing AI transformation. And so the reason I spend all my time on this is I believe the 1.5 million impressions that I, uh, drove yesterday, that's less important to me than knowing I am sure we are going to get at least one $100 million a year plus business that comes outta the post I made yesterday. That is a real prospect that we can end up doing business with.

Yeah. And the thing is, I think we, you can, you choose if you wanna reveal this or not, but I think it's, we talked about what X is good for, what LinkedIn is good for, what other channels are good for, right? In terms of driving pipeline. And I have, by the way, I found what you said to be true about X being really good at bringing in engineers. So that's been good. But I think, okay, let's say an X post gets 1.5 million. Are you cross-posting it as well to other channels to drive pipeline?

Yeah. So my whole thing is 30 Days of AI. I want to exist on all platforms. So I posted on X and LinkedIn yesterday. My plan is, so starting today, I'm going to post on short form also. So I'm using the repurposing engine on the content machine where basically it took my Software Factory post, which was like 2,000 words, and it was like, okay, this is the part of the post that you should focus on with short form. And I thought it picked the exact right part for a short form video. So I'm going to record those today. The other thing I'm going to do is I'm going to turn it into an email newsletter where it's a 30-day email newsletter that people can follow along and get an email every day with my insight from the day. And then what we'll do is we'll drive a lot of emails from that. 10X already has a newsletter called UltraThink that people naturally gravitate to once the email kind of explodes and the series is done.

How long do you think you're spending on content per day right now? 4, 5, 6, 7 hours?

Yeah, I would say 5 hours minimum.

Okay. Minimum. Wow. Okay. And because you're talking, it's like, I feel like I like the DJ analogy because I do feel like I'm making a YouTube long form and all of a sudden I'm recording a webinar with you and all of a sudden I'm chopping things up for Shorts as well and repurposing. That's what I feel like I'm doing.

Yeah. And also, The reason I also view myself as a DJ right now is like when I wrote this software factory piece yesterday, I do not claim to be a lifelong engineer who has built his own software factory. And so I am also approaching this entire thing as kind of like a— just a student of the industry, almost like Talking to the smartest people I know to get their perspectives on things, and then collating it in a way that makes sense to people. So like, how did I approach the software factory thing yesterday? One is I used the research feature like function of the content machine. Then I went to all of my engineers at TenX, asked them, "What do you think about a software factory?" And then I asked, and then I also had the Oracle find all tweets related where like people like Chamath or Ryan Carson had mentioned a software factory. And so it's like I was almost like pulling all of these things into a patchwork that I thought told a cohesive story to the non-technical person about what a software factory is and why you should care from the smartest voices in the space.

Yeah, this is awesome. And by the way, because you come from such a strong newsletter background, like how does the newsletter play in right now? Like are you guys running a bunch of paid ads for acquisition? Like what's the deal?

No, we're not running any paid ads. Again, it's not a huge newsletter yet, so UltraThink is a— we have 12,000 subscribers. It's around a 45% open rate, and it's weekly. We are going to up that at some point to— I think we'll probably get to 3 times a week. Right now, JJ, our creator, who's like the man that does it all, is also writing our newsletter, and he has his own set of skills to write UltraThink. But my view is this, is like, I think all these other channels we have, so 30 Days of AI, if we end up doing webinars, when I bring back my show and I'm— you know how you're going live right now? The thing I'm going to do is I'm going to go live with my show, but I'm going to still make them events so that we capture leads for events before going live. That's going to all funnel into UltraThink. And then the other thing that we plan to do is I'm going to actually sell ads against UltraThink and any of the media we do so that any content hire we make is basically paid for by our ads. And so all Yeah.

I love it. That's awesome. Once you figure that out, we can do another one of these. And by the way, everyone watching right now, you guys should take— once this is on YouTube or whatever channel, you should just take the transcript of this and then ask it to Skillify. Just ask it to interview and then figure out how you can Skillify this and make your own.

Totally. And one thing I would say is like this whole process of turning my content process into kind of like an AI-native workflow, It is not a point-in-time thing. It is like a constant job, right? Even since you and I spoke, Eric, I added in the ability for someone to, like, for creators to have like their own file in this structure. I added in the internet reader thing where it's curating the internet and not just our internal stuff. And so I'm constantly thinking about what are ways that the content machine can be leveled up. Like, one thing I'm thinking about, and I don't know if it's going to be possible because the tech isn't there yet, is like if I Just stepped up to the camera to record a short-form video script that the content machine recorded for me, could I have a remotion skill built into it that does a short-form video edit using the content machine? I'm constantly thinking about any friction or work that exists. How do I take that to zero so I'm only doing the things I have to do?

Yeah, it's constant upgrades. I don't know if you played— this is nerdy stuff, but I used to play Warcraft and you could keep upgrading your town hall.

Same thing. Same thing.

Then they're like a castle, right? So that's great. Just a few more questions just before we wrap up here. So, I mean, okay, maybe going back to the reach thing for a second. So yesterday, 1.5 million on just that X post. How many impressions do you think you're getting per month right now? Are we talking like 20, 30, 50 million?

No, it's not that much yet. Let me look here. I'm just gonna look at Creator Studio as we're talking. Last 4 weeks, I'm at 2.5 million. And then last 3 months, I'm at 8 million. And that's on Twitter. Also, to the point you were saying before, it has been my experience that X is like where you get street cred and where you attract engineers and where you learn.

Yeah.

LinkedIn is where you drive business. Like, all— basically all of 10X's revenue, um, has come through LinkedIn. All of our AI strategists, which are like our technical consultants, have come through LinkedIn. A lot of our engineers have either come through, um, X or through Juicebox, which is like an AI sourcer. M. Yeah. And then the only business opportunity that's come through X is our partnership with Anthropic originally originated with someone. Yeah. So that is huge, but that is the one thing. So yeah, it's been interesting to see which platforms provide value.

So let's say X is like $2, $3 million a month, something like that in terms of impressions. What do you think it is for LinkedIn?

Let me see.

And as you pull it up, the reason why I'm asking for everyone is because to get these types of compounding results, one, it takes a long time, but also the amount of time Alex is putting into this, 5+ hours a day, that's what it takes. Honestly, it's probably like 7+ hours or so, but I just want to give you a sense of what volume you can achieve.

Yeah. So, So LinkedIn last 24 hours was 120,000. LinkedIn, and I have 210,000 followers on LinkedIn. By the way, followers just doesn't matter anymore as it relates to social. Can I— I'm having trouble like—

28 days?

Yeah, that's what I'm trying to do. Let's see. My guess, I can't get it right now, but my guess is last month, it's probably 1.5 million. What I'll also say is, and this is something I constantly think about, is I generally want my impression number to increase, but I also, when I think about who 10X's customer is, it's a very specific person. person. It is basically a CEO, COO, CIO, or CTO at a $100 million a year business or up. So a lot of times I'm thinking about how different is the content I create for that person versus, say, like, just the typical knowledge worker who's trying to navigate AI. And I think there's a Venn diagram.

Yeah.

Like, when we're talking about how to build, like, a directory of skills, like the Content Machine, I think that sits in the middle of the Venn diagram. But there may be things how to think about governance as it relates to AI that I should probably talk about that probably won't get as many impressions, but it isn't relevant to the second audience I just mentioned.

Yeah, it'll get the right people ultimately.

Yeah, exactly. Cool.

And so because you see a lot of these businesses right now, what do you think is the primary thing? Where are most people getting stuck with AI?

What are the patterns that you're Yeah, I think, um, so I think there's a few things. One is like everyone right now is— the focus is really starting to turn to ROI. Like every company that we talk to is like, we're spending a shit ton on this, our revenue is not necessarily going up, we have to like start really going through it with a fine-tooth comb more. And so this is why I think it's interesting. I think that every company in their AI journey should probably start with something that can very clearly map to either revenue generation or cost savings, and very clearly, because my view is there's going to be a lot of AI work you do in a company where it is not super obvious what the ROI is. Just like for a long time, me creating content online, was not super clear what the ROI was. To me, there are a few of these things that very clearly can tie to ROI. One is getting your engineering org to be closer to a software factory, where basically the amount of PRs that are merged and didn't have to go through human review, or at least humans reviewing it, it was the only step and they weren't writing the code. The way to measure that is you just look at what are the average number of PRs merged by engineers before and after? What's the average cost of an engineer? How many fewer engineers do you have to hire now because of it? Very clear. Customer support, also very clear, like number of tickets that were closed by your customer support agent and what is the number fewer of customer support people that you need in your business? Also very clear. Sales gets a little wonkier because like you go into a sales org, typically sellers spend 50% of their time not on kind of like revenue-generating activities like talking to customers. If you can show that you're getting more of their time spent on revenue-generating activities, that's a plus. But unless like their actual revenue that they are driving is going up, it's harder to show. But basically my view is, is you have to focus on ROI-driving activities basically to win more budget for things that are not necessarily going to show ROI over a 3 to 6 month period.

I'll give you the last question. I think we can go on for a very long time, but we'll do another one of these. But how are you measuring productivity right now? Are you looking at— obviously, people like to talk about token maxing, but I think you got to pair it with another metric such as PRs, for example. So how are you looking at it? Is it— you're talking about tokens consumed and also something else paired with it? How are you What do you think about it?

Yeah, I mean, for us at 10X, it's like super easy with our engineers. I can't remember if I told you this, but like we pay our engineers like salespeople. So like our engineers at 10X all have like a minimum guarantee and they are paid based on story points they complete. So basically the way we measure it is how is an engineer's number of story points they complete for complete and are accepted by clients going up over time? And in theory, almost like there's inflation of the US dollar, There should be natural inflation of story points because if you think about it, if AI keeps getting better, engineers should be able to do more just by the AI getting better than the alpha that they get is them improving their own systems and their harnesses for working with it. So that's how we measure our engineers is what is their growth of story points complete over time.

Yeah. I'll throw in a fun question here for you. So do you think this is true or false? Alex is using AI and Eric is not using AI, and Alex continues to compound for 12 months or so, Eric will be unable to catch up. True or false?

With anything or specifically with content?

Let's go on content.

I think it is true if the output that I'm putting out is actually better than a pre-AI world. If the output's not better, then it is false. And actually, I don't think Alex will be able to catch up with Eric.

Yeah. Okay. So here's— that's actually good because then you can apply the law of compounding with content also to business and anything else really.

Yeah.

I think it's true. Okay. So that's a good place to end. So Alex, what is the best place for people to find you online, learn more about 10X?

Yeah. On X, Business Barista. LinkedIn, I'm Alex Lieberman. And 10x.co.co is our website, and always down to chat about it. I'm definitely not the expert. I don't know everything, but I'm just trying to learn as much as possible on this journey.

Cool. Great. And guys, I'll be on Alex's next week, so you can join us over there. But thanks so much for joining, Alex, and thank you all for tuning in.