Jev clearly explained (and how to make money with it)

All right, today we have Nick back on the podcast and he's going to be going over Jev for us. So everybody's heard about Jev. If you're spending any amount of time on X, it's like all anybody's talking about over the last three or four days. So Nick, why don't you tell us what is somebody going to take away by the end of this episode in regards to Jev? Yeah, I think the biggest thing is like this is a new paradigm of a model and everyone's wondering, okay, how does this change my current business that I'm running with AI or this tool that I built using AI? Does this affect me? Can this thing make me money? I'm going to talk through all of that practical use cases here that we can use Jev for and whether it's really something that you should be paying a lot of attention to or is it a distraction from something you're already doing? We'll dive into all of that. Awesome. And so before we start diving in, can you just like define Jev for us? Like think about the non-technical entrepreneur business owner who's listening to this, who's like, you know, maybe they've never even heard of Jev. Maybe they don't hang out on X. So they're like, what is Jev? Can you just define it for us? Yeah. So Jev is, I'm kind of jamming over here. So hopefully this is okay for everyone, but Jev is like this new paradigm of a model where it's not a LLM where it's like outputting language for everything that it does. Jev thinks in terms of multiple choice options and it's, you know, choice A, choice B, choice C, choice D. You define what these multiple choice options are and Jev will choose kind of whatever its confidence is for each answer. That's the model. It's choosing its confidence on what the answer is. You ask it the question, you give it the answer choices, it will answer the multiple choice questions. So all that to say, it can't output text. It can only really think in tools in terms of what you provide it as a capability, but it still unlocks a lot of really cool use cases. The two main benefits of Jev being it's super fast and it's super cheap. So we can dive into all of that and kind of what that looks like, but does that make sense already a little bit? That does make sense. So if I'm understanding correctly, cause even, I mean, it took me like a couple of days to really grasp, like what the heck even is this and how does it differ from like a chat GVT or a Claude? So if I'm understanding correctly, it's like Jev where Jev excels is as a decision maker. Like Jev is not going to sit there and help you, you know, plan through your strategy or help you, you know, wireframe your app idea or whatever. But if you already have a very specific task or a very specific set of actions that you need AI to take, you can essentially give Jev those actions or give Jev that assignment. And it's going to go and execute exactly the way that you've laid out. You know, I'm seeing like 200 times faster and like 400 times cheaper. Like, is that accurate way to think about it? Absolutely. That's the way to think of it. And here's a really good example. Like when I first started building computer use agents, so this was back in November, 2025, this was pre open claw pre Claude bot pre any of that. Okay. I had one month of runway left. I was living in a hacker house. This is where I met Spencer, my co-founder now at Orgo. And he was building the early days of Orgo and Orgo is a platform. You can build computer use agents. You can give an agent a computer, et cetera. I was building the first computer use agent for a customer of mine. She was a promotional product distributor. So she had this need where these websites, they had all these products listed on the website. I forget the name of the website, but they had all these products listed on the website. And what I would do is I had a computer use agent go into the website, log in, click around, find each product, download the product information, find the price, the categories, all these labels. And they would do that sequentially. So if there was 60 products that she had in a presentation for a customer, the agent would go one by one by one by one by one to each product, scraping all the information. Then it would store all of that in her Zoho CRM. And she would make like this custom quote slash invoice for her customer, send them the email, attach this like, you know, quote, whatever. And so it was a very defined process, but we were using computer use agents for it because there was still intelligence involved in terms of like, okay, finding the product, finding the information, understanding how to label it, et cetera. So that used to cost, I believe it was around $30 an hour for every hour that that agent was running. It would cost $30 an hour. And sometimes it would take two, three hours for it to get through, you know, 60 products of a presentation today. I don't, you know, that customer, she's good. She's taking care of, she still uses it. I can go actually go back to her and I probably should do this and rebuild that using Jev because it's a semi deterministic workflow. Jev can choose from multiple options. Oh, what is the color of this? It knows the answer instantly. What is the, what is the size? Okay. What is the quantity? And it could essentially take that $30 an hour task and probably shrink it down to less than, less than 10 seconds is what I would assume. And then also be like free, practically free in terms of cost. So this is kind of that what it unlocks now. Does that mean that you should go today and start building with Jev right away? Well, look at my use case here that I just described. I already had a workflow working. I already had some process built out and it made sense at that point for me to come in and I could rebuild it with Jev. If you don't have that already, don't kind of get distracted with this right now. Build, go out, build a, build an agent. That's really good at specific workflows for a customer. Do that first. But when it comes to like, you know, implementing Jev, it really starts to matter at how much volume are you doing? How important is time for the process? Is it a, is it a timely process already for your agent? Is it an expensive process already? If not for neither of those answers, probably don't pay attention to it right now. It's probably just a distraction, you know? Does that make sense? That makes perfect sense. Yeah. Because I'm thinking through like with my AI concierge clients, for example, like what I'm helping them build and what I'm building for them in a lot of cases are, you know, agents inside of Claude Coworker inside of Claude Code that do very specific things like you said. But these are not tasks that take a ton of time to do. Like for example, a recent client, we automated her podcast production workflow. Now she's doing one podcast a week and you know, the Claude, the, the, the workflow that we automated with Claude Coworker takes 45 minutes a week down to roughly five minutes. But it's not, it's not like it, you know, we went from 20 hours a week and we need it down to 10 seconds, right? Like it's, I think it comes down to knowing when to use which tool for the job. And what you're saying is that for most people, it sounds like from what you're saying, most people, like the tool for the job are your traditional agentic tools, like a Claude Cowork or, you know, Hermes built on top of Orgo or things like that. And it sounds like where Jev excels and maybe the angle that people can take to actually, you know, make money with Jev and where it makes sense to actually build with Jev are when you're dealing with a very high volume of inputs and a very structured workflow. Yeah. Like where my mind goes with this is, you know, think like large insurance company that's processing a ton of documents or a ton of claims. And they're usually the same forms, like it's the same documents. It's just, they're dealing with hundreds of them a day or thousands of them a week. Sounds like that is an opportunity for somebody to, you know, if somebody had a connection with an insurance agency like that, go in and build, you know, build a workflow or build an agent that uses Jev to go in and take that. Like we said, down from maybe, you know, even the AI automated version of that might take 30 minutes a day, but you could take that down to like 10 seconds with Jev. And make it virtually free. Exactly. Like, is that a use case that might make sense? Of perfect use case. And the key there is like Jev is not going to be the model that makes you money per se. It's going to be the model that saves you money and saves you time. And so this is kind of a good example of how, like, if you're building with agents, this is why when I first started with charging like 5k a month for an agent, for a managed agent on OpenClaw or Hermes and you build the computer on Orgo. People were like, 5k a month. How do you charge 5k a month? And I would say, I would give the, I would give the customer unlimited tokens, unlimited agents, unlimited workflow automations. And obviously you scope it, but, you know, so you deliver on a deliverable, you know, weekly cadence or what have you. But people are like, how are you doing unlimited? How are you doing unlimited? The cost is so expensive. But I was like, even if, and this was the case, even if on month one, month two, you're breaking even because they're using it so much. A, that's great. They're using your product or service bunch. And that's great. B, okay, we're breaking even fine. But the cost of the models are going to go down. And this is the perfect example. It's like, if you were at 0% profit margin or 50% profit margin before with some sort of product or service involving AI, you might be able to get that down to like 90, 99%. Like, like this is what that kind of unlock is. And so if you, if you're not already making money with AI, forget about it. Go solve a problem first and do it, you know, how you know how to do it with LLMs. They're kind of easier to understand intuitively. When you get to the point where scale matters, volume matters, pricing matters, time matters. Jev is a good place to start ideating around how you can incorporate it. Got it. Okay. So that makes a lot more sense to me. And like the mental unlock for me was when you said like, Jev is not, you're not going to make money with Jev. You're going to save money with Jev or you're going to save time with Jev. And again, it sounds like the advice is like, go build with AI the traditional way, like build agents the traditional way. And when you find those agentic workflows, you find those clients that have workflows at scale that are costing a lot of money in terms of time or tokens. That's when you could sub in a model like Jev to, again, not necessarily make them more money, but save them a massive amount of time and therefore a massive amount of money in the form of token costs. So that's the play for, for using Jev to make money. Exactly. And like, we can kind of show through here, like you can even tell by like this first demo video I made on Jev, you can see here on Twitter. I'm like showing how you can use it for. Can you zoom in a little bit? Yeah. Let's zoom in. You see this better? Yeah. Yeah. It's a lot better. So this first demo video here that I'm showing up Jev, I'm actually showing it playing Minecraft. So it's kind of, I didn't plan it for this to be a non-productive, non-money-making thing, but it kind of shows the nature of Jev is like, of course, the first thing I do with it is not a non-money-making thing. Kind of, kind of shows you the, um, the vibe of the model. It's not going to make you money. Right. A cool use case might involve it playing Minecraft, right? But it's still really cool. So what you see here is I set Jev up, uh, to play Minecraft and I gave it a bunch of, well, essentially what I did was I pointed GPT-6 Astra at, uh, at understanding Jev and how to use it. And it gave Jev a bunch of, um, tools and, you know, multiple choice options that it could use to play the game Minecraft. And then we could define like a goal, like build a house in Minecraft, and then Jev could go accomplish that goal using its multiple choice tools. You can see here, it's kind of crazy of how it uses tools. It's very fast, very like spasm, spasmic. Um, and that's fine. And that's what it's good at. So it's a huge unlock for computer use. We've added it to our playground in Orgo. We're going to really build a whole harness for computer use around Jev so that it works really fast. And that's a great unlock, but, um, yeah, it's not like it's like going to be able to do something that an LLM can't necessarily do. It'll just do it faster and cheaper. Um, and you can see there, that was like six pennies, six pennies to be playing Minecraft with computer use. This is insane. So, uh, another demo here, this is someone building on Orgo. They have 16 computers here and they built a look now, uh, detector using Jev. They have a bunch of different websites open and they have Jev reading all of them in real time, essentially. And telling the user alerting them, Hey, this new computer use model just dropped. Fun fact that one uses Jev. And you can see it alerts him, uh, of four variables here. These are the multiple choice options that this creator has defined. And Jev is saying, Oh, this, this little thing right here, this model announcement matches three of four criteria. You should look now. Here's my confidence. And, and then it shows it as the main screen. So like, and then it's keep scanning, keep scanning. Okay. Okay. It's reading everything, reading everything. And then what alert you should look here because this matches two of your four criteria. You know, this, this is kind of makes sense. It does. So yeah, I was going to even dive in deeper on that one for my own understanding and hopefully for the audience too. So like, but this look now example. So this is like you said, these are 16 different independent cloud computers all running on Orgo's platform. And by the way, guys, for those listening to this, we have a promo for anybody that wants to check out Orgo. So Nick is the co-founder of Orgo it's Orgo.ai slash Corey, I believe. Yeah. And so we'll put, we'll put the link to that in the description. That'll get you, I believe it's your first three days of Orgo for free. And then 20% off your first three months. If you want to test out Jev or you want to test out, you know, Hermes or a, you know, open claw deployment on an Orgo virtual computer, that's what we use in our business. But what we're seeing here is somebody has basically spun up 16 different cloud computers in Orgo. Inside of each of those, each of those computers is running Jev. And it's, it looks like each one of those computers is what they're doing. It's kind of just like watching breaking news, or it's like watching X to see, okay, what are all these new tech releases or new model releases that are coming out each day? And because the creator of this gave Jev very specific criteria, it's like, Hey, these are the four criteria you're going to look for when you judge a model release and any release that has at least three of these criteria, you're going to alert me to say, look now. So, so because of that, because the, I guess criteria is like very strict and very defined, all Jev has to say is like, yes, no, that's something that like is a perfect use case for Jev. Whereas if we had something like, you know, a GPT six or a fable, or even like an opus or a GPT five, six soul, if we had it do that same task constantly, it would cost a freaking fortune. Yeah. It would cost a fortune because it's constantly having to scan through news. It's constantly having to, you know, in, uh, infer like, Hey, you know, is this relevant? Is it not like it's generating all these lines of text to, to show its reasoning and to show its thinking process. So, you know, a day of that workflow with, I would, with GPT six Astro, for example, it would probably cost hundreds of dollars, but with Jev, it's costing like pennies, five cents, six cents, something like that. If that, if that, yeah. Got it. Okay. So, yeah. So you could run, I mean, that, that leads me to think to like intelligence reports for, again, I'm always thinking like what's the make money option. Obviously again, with Jeb, it's, it's going to be more of a save money than make money. But I think this unlocks use cases where for like competitive intelligence at scale, like I think it's now reasonable from a, both a timing and pricing perspective to have a model like Jev, you know, depending on what industry you're in, you might have hundreds or even thousands of competitors. I think like e-commerce, right? If you're in the skincare niche, there are thousands of not tens of thousands of skincare brands that you probably compete with. Oh, wow. Well, something like a, something like a Jev could go and monitor like literally every single brand in your category for job postings or, you know, price drops or stock outs, like all this data that would have cost literally a fortune to, for that same monitoring with a, you know, a frontier model. So again, that's just like what came to my mind real quick. Dude, you, you, you got it. Whatever the talent or the, you know, the gift is to know how to turn this into something that can make you money. You have it because that is a genius use case. Like to watch your competitor websites for some like e-commerce or like, let's say like there's limited supply. I don't know. I'm thinking of like, I'm thinking of shoe bots, you know, do you remember the bots? Yeah. That's I come from that background. Like, in fact, now that I'm thinking about it, a lot of my friends in the e-com space, they do online arbitrage where they're buying off of websites. Right. And they're flipping on Amazon for a profit. Yeah. Jev could be constantly scraping, you know, like Kohl's.com, Dick's Sporting Goods, all these websites and looking for those opportunities. And it would cost pennies. Whereas, you know, right now it's just, it's really expensive to run those types of scans. Oh my gosh. That's why did, yeah, that's such a good use case. Okay. So, so there you go. So it's like, um, that's a perfect use case, you know, and I know I said at the beginning and I still stand by it. If you're not making money, if you don't have a system, if you don't have some sort of process already, don't pay attention to Jev, it's a distraction. But if you do example, Corey gives here of like, you know, you obviously, if you're buying and reselling on, on, on, you know, eBay, et cetera, you have a process in place already. That might be a great use case for Jev where it used to be way more expensive. Um, you can try it out in Orgo. So we did add it to our playground. Uh, it's in preview mode right now because it's your, your test doing when you're using it in Orgo, you're just testing the model. So what we're going to do is we're going to create a nice harness around it. So you can just talk to the chat bot, just like you would normally talk to it. And it would do computer use using Jev a lot faster. The way that would work is it would use GPT six as the brain. And then Jev is the executioner. That's kind of how the best balance would be for, for using something like this. If you're interested in that building with it and computer use, that's kind of how to think of it. I think here you can see, I was like asking it to go through a few moves of chess to show you how fast it is. I'll say, uh, please continue, uh, do this for a while. Let's see. Um, and you can see it's like very fast with its, uh, execution. Like it's just moves very fast, but you can see it's kind of dumb. It only did one move and it didn't really fast. We only did one. So once again, you have to define the criteria. You have to be like essentially build its brain out for it to act, let it act. It'll do it instantly. If I say continue again, it'll do this execution instantly. Um, you see there. Yeah. But it requires kind of a process in place. This isn't like it's going to real time reason and play through this chess board unless I give it the kind of the variables and multiple choice questions, uh, to answer for it to do that. So I hope that makes sense. It's definitely a, it's definitely a good set of hands for, for the existing models is what I would say to, uh, to act really fast. Yeah. It makes perfect sense to me again, now that we've kind of talked through it. So my, my final question for you, Nick would be like, is this something that do you see like an open AI or an anthropic kind of creating their own Jev like model and integrating it into their existing ecosystem? Because in my mind, that would be a no brainer. It's like, Hey, you know, when we need actual intelligence, when we need to think through things, when we need to reason, well, we've already got the frontier models models for that. But if we need just like we were saying earlier, like literally just an execution layer or just like a workhorse to go bang out the actual work or, or, you know, use the computer, uh, to use the computer use example. Like that in my mind would, would allow a Jev like model inside of that ecosystem to be really useful. Like, is this something they're working on? Do you know? I think they're definitely, uh, paying attention to it. And, um, the architecture is like fundamentally different from, you know, the LLMs, but it's, it's actually, yeah, it would make sense for them to have their own kind of like version of this. I saw immediately it got open sourced. Somebody built a version like Jev. I saw that. Uh, and it like open source to right away. So I think, yeah, it's like we're in the intelligence explosion. There's going to be different versions of, uh, there's going to be another Jev. That's going to look entirely different from Jev and other LLMs. And it'll be a new sort of intelligence. And these things are just going to keep coming out. Um, and I think, yeah, the, the, the big labs, their goal is to kind of package all of it together in a product that just works. Um, so for, for everyone else, you know, build on top of it, if it makes sense. If not, don't get FOMO. It'll come to you and it'll be integrated into your existing products. Um, and then it'll, that'll help you use it. But, um, yeah, I think they're, they would definitely be working on something like this to integrate into their existing models. So. That'd be my thought as well. Well, and so again, just to kind of summarize here and to conclude. So the 80 20 would be, you know, if you're not already building with traditional AI, then Jeb's probably a distraction for you right now. It is really good at making decisions at scale, you know, quickly and cheaply, but it's not going to reason. It's not going to do the thinking for you. And then, you know, aside from that, it's more of a money saving model versus a money making model aside from some of the use cases that we mentioned earlier. So from like an 80 20 perspective, is that, does that pretty much summarize it? Is there anything I left out just to kind of summarize for anybody that stuck around? I think that's pretty, uh, spot on. And yeah, if once again, if you have something that involves like multiple, you know, workflows, for instance, I even have something here. Um, this is a video. Can you see this? Yep. I can even make it a little bigger. Well, let me make this bigger. Okay. So I have this here. This is a video that Dewey, Dewey is my Hermes agent. He's still the most impressive agent I've used. I've tried GrokBot. I've tried Muse. I've tried everything. So I told Dewey, I said, Dewey, go put together some use cases for Corey's audience here. We want to use Jev practically to, you know, make money and save money. And Jev put together here some, I'm sorry. Dewey put together here some Jev use cases showing, look, I can take all these downloaded PDFs and I can sort them, organize them, save them to their, um, dedicated folders pretty much instantly. So this is something that Dewey put together. He went and studied the Jev model, integrated the API and built this out and screen recorded it on Orgo. So it's kind of cool. Uh, who needs a reply from your team? So Jev is what is Jev going to do? It's going to triage all of your different channels and tickets and emails tell you who needs a reply. Go research this website and, uh, and, and, you know, scrape it, et cetera. Jev is going to go to the website, understand. Okay. Click through all the, all the, the tabs, you know, say, save a note on it. Maybe kind of like that. So this is just, uh, some more examples here that you can, you can always ask your, your agent to help you understand it. Ask it if you can help it, um, uh, you know, implement it into your current systems. And yeah, that's also a very valid way to, to build with Jev. I think that's a great point too. Just going back to first principles is like, anytime you get stuck or if you need further explanation, just ask your existing agent or ask your existing LLMs to explain it to you and to put it in plain English. And a lot of times that'll give you everything you need to know and also help you determine like, when is it the right tool for the job versus your more traditional tools? So I think that's great feedback, Nick. Great having you on as always, if people want to connect with you or if they want to, you know, find out more about you and especially if they want to check out Orgo, where do you want to send people? Yeah, you can find me on X. Uh, I make a lot of posts about AI agents, you know, how to sell managed agents to businesses as well. Uh, my handle is at Nick Vasilescu. I didn't get my full name. Someone stole it. So I have the last two letters missing Vasilescu. You can also find me on YouTube. Uh, it's the same handle and yeah, I will continue to put out content on how to build agents for businesses, how to make money with agents and any kind of new thing that comes out. I try to cover it pretty fast. So check it out. Awesome. Well, Nick, thanks so much for coming back and for everybody watching this, hope you enjoyed and learned a little bit about Jeb and we will be back soon. Thanks guys. Thanks.