챗투브트 개발에 참여했던 연구원이 새로운 AI를 공개했지만, 챗투브트와는 전혀 다릅니다. 이름은 제브예요. 그리고 Jev는 텍스트, 이미지 또는 비디오를 생성하는 대신 기본적으로 AI 의사 결정자입니다. 몇 가지 정보와 가능한 결과 목록을 제공하면, 프로그램은 가장 가능성이 높은 결과를 신뢰도 점수와 함께 결정합니다. 이것이 왜 그렇게 중요한 문제인지 설명하자면, 바로 속도와 비용 때문입니다. 누군가 제프에게 1,700개의 이메일을 입력하고 스팸 여부를 판단하고, 중요도를 평가하고, 답장을 보내야 하는지 여부 와 어떤 범주에 속하는지 결정하도록 했습니다. 이는 약 420만 개의 입력 토큰을 사용한 작업이었고, 전체 작업에 18센트가 들었습니다. 그러고 나서 그들은 그것을 장편 영상으로 만들었습니다. 제브는 약 3초 만에 잠재적인 바이럴 영상 클립 17개를 찾아내고 점수를 매겼습니다 . 또 다른 데모에서는 이 기능을 사용하여 브라우저를 제어하고 취리히에서 런던까지 가는 항공편을 7.1초 만에 찾아냈습니다. 그러니까 거의 모든 앱이나 기업 내부에서 얼마나 많은 사소한 결정들이 이루어지는지 생각해 보세요 . 이 단서는 괜찮은 건가요? 이 이메일은 급한가요? 지원 티켓은 어디로 보내야 하나요? 영상의 어떤 부분을 짧은 영상으로 만들면 좋을지, 매번 모든 결정에 대해 긴 LM 파일을 작성하는 대신 어떻게 해야 할지 알려주세요 . Jev는 해당 계층을 매우 빠르고 저렴하게 처리할 수 있는 잠재력을 가지고 있으며, 이는 완전히 새로운 범주의 AI 앱을 탄생시킬 수 있음을 의미합니다. 그러니 테스트해보고 싶으시면 [음악]이라고 댓글을 남겨주세요. 제가 도구를 보내드리겠습니다.

AI will build you pretty much anything in an afternoon now, but it won't stop and ask you if that specific thing is actually needed in the first place. I coach businesses on AI every single day and almost none of them set out to build software. Instead, they wanted a specific task to take less time. But what happens is the app is built and then in a month later they have to keep taking care of it and nobody uses it. And more importantly, there's zero leverage in the business from this. So, I want to show you why AI keeps steering you there, what it costs you later, and the one question I ask before anyone builds anything when they work with me. So, let's get into it. Now, a common question that I ask myself when working with my clients is why does so many of them build apps? And I bullet it down to two things. Incentives and recommendations. Often times when you work with AI, you want to build something that's impressive. So, you can show somebody visually what you did. So, that could be your team, your company, whatever else. But the other side is recommendations. Often times when the AI works with you, it loves to build things. It loves to build apps. So it's going to recommend that you build a dashboard, a prototype, a mobile app or something like that to solve your problem. And often times that's unnecessary. When I say often times, probably 90% of the knowledge work that a lot of people that are non-technical that are working on can build this without applications that are standalone. And it's not just that you can do it, but that you should do it. And I'll explain why I feel that way as we go through this. So the trap here is simple. We have a thing that we want to automate or be partially augmented with AI. The benefit here is to save time or create leverage in the business to do something you could have never done before without people to fill that specific role. When we have the task at hand, most of the time we're going to want to save time. So, we want to save some time. But the issue here is that when you ask the AI to assist in this process, saying, "Here's a thing that I want to do. What should I do?" It's going to offer up an application that you should build most of the time because it's unaware of its own features. And this is kind of weird with AI today. I'm sure it will be fixed in the future, but right now if you ask AI what it should do, it's not going to reference its features that it has such as scheduled tasks, skills, projects, GPT sites, all those things that it can use in its own product, it very rarely recommends, even though it is a simpler solution. And when you build the app, it's easy to build the app initially. You can build a prototype and it looks amazing. The thing that people run into is maintaining the app over time. Now you have another thing you have to take care of. And instead of taking stuff off of your plate, you put something on your plate. And that's counterintuitive to what we're trying to do here. We're trying to take ourselves away from these tasks so we can work on things that are higher leverage for the business. Now, another trap here is that when you're building these applications, it feels like progress. It's something you could have never done before, especially for those people that are non-technical. You're building applications. There's tons of code and the AI is doing a lot of things and making a lot of progress. And also, you can see it. Like I said previously, it's something you can visually see and you can show to other people. That's not the case when you're building skills and projects inside these products because it's just a text file. It's not a fancy dashboard that you can show to somebody. And another thing with AI is it keeps on recommending new features. So when you're building with the AI, it says maybe we can add this or expand that. And all of it sounds like great ideas to you. And also the AI does it so effectively and so quickly you're like why not just add that? But every addition usually migrates you further and further away from the core problem you're trying to solve in the first place. And that's when the original problem just fades in the background. And what happens is at the end of a weekend or a week of you building this thing with the AI, you look at the finished product and you realize that it's both not ready to use and also it's not solving the core problem that you wanted to in the way that you want it to be solved. Quick pause in the video. If you're enjoying this, you're likely going to enjoy two other things. First off, if you want to work with me, blow our series of offerings to see if there's a good fit from the two of us. And the second thing is I have a 30-day AI insight series that's completely free. You can click that link and it'll give you 30 insights in your inbox of how you can apply AI to your business and your work. Now, let's get back in the video. Now, let me give you an example here. This is a fake ex

So, a brand new kind of AI model called Gev has been going completely viral all over the internet this week. And this is the launch post right here. This is already more than 35 million views from one of the co-creators of ChatGPT, and people have been testing it like crazy and sharing what they've built with it. Just like this one right here. So, this one, it cuts a cloud code session from nearly a million tokens to 86,000 in just 1 second. And then this one right here, it finds a flight in literally 7 seconds for under a cent. And then this one right here, it's playing 50 games of Subway Surfers at once for literally less than a penny. Now, we have never seen an AI like this. It cannot write a single word. It only makes decisions. And they claim that it is up to 200 times faster and 400 times cheaper than any other LLMs like Claude, Fable, and GPT-6 Astra. So, what is it and where would you actually use it? Let's get into that. So, Gev, it comes from a company called Type Safe AI. They're out of San Francisco, 2 years in stealth, and this is the first thing that they have actually shipped. And the way that it actually works, it's that simple. Now, you give it three different things. The information first, so maybe a contract, a customer message, maybe it's an advertisement, whatever it is. And then a question about it like, "Does this renew automatically?" And the only answers that it is allowed to pick from. Now, it reads from all of that and it hands you back one pick plus a number for how sure it actually is about that. Now, there's only three kinds of questions that you can actually ask it. A yes or no, "Does this message ask for a refund?" It comes back as a probability. So, instead of a yes, you get 94% yes. And then from there, you just pick one from a list like, "Which team should handle this?" Whether that's billing or technical, maybe the sales team. And up to 255 options come back. And you get a probability for everyone plus a confidence number for the actual pick. And beyond that, you score it on a scale that you write yourself. "How frustrated is this customer?" From calm to annoyed to furious. There's actually up to 10 different levels and the answer can land between two of them. That's the entire product. The reason that this is so fast, it's very simple. A normal LLM, like if we take Opus, GPT, any one of them, it writes its answers one word at a time, and every single word has to wait for the one before it. Now, Jev, it never writes anything. It's just like handing somebody a multiple-choice sheet instead of asking for an essay. Now, every answer is already going to be on the page, so it just marks the box, and it marks every box on the sheet at the same time, which effectively means that you can put a dozen questions on the same piece of text, and they're all going to come back together. Now, they actually tested about 13 questions in one call, and it came out over 11 times cheaper and almost 10 times faster than just going one by one, and that's where the speed comes from and the price, because you only pay for what it actually reads. Now, a lot of people, they have been looking at this and saying like, "Okay, that's just a classifier." And yeah, it is a classifier fundamentally, but the difference is you do not have to train this. The old way, you would just feed it 10,000 labeled examples first. With Jev, you just type your options in plain English, and it works on the first try. So, okay, now to actually get Jev in whatever harness that you are using, whether that's Hermes, Claude Code, or Codex, it takes literally just a minute. First, you can sign up through typesafe.ai, make your key. You will have to take a little bit to actually get registered, because you have to go through kind of a tedious process, or you can just go with OpenRouter, which is what I'm going to be using. Now, we simply just have to paste in this command line, or you can just ask AI or Fable or Astra or whatever you're using to help you install this. We just have to provide your OpenRouter key and just tell it what you're trying to install, which is going to be Jev. Anyways, we will have the command available inside of that guide in our free school community along with a plethora of other things from this video and our other videos as well. So, get that there. All right, now diving into the numbers a little bit further, their homepage, it claims 193 times faster and 440 times cheaper from tests that they ran themselves. Now, the price is 4.2 cents per million tokens going in, and every answer coming back is completely free. So, let's say you've got 10,000 decisions to be making, each one reading about 750 words. When that's a support ticket, a lead form, or a review, in total, that's just going to be 42 cents. Now, covering the speed, in our test, One Decision takes Jev about 4/10 of a second. Tera, this takes only 10 seconds. Luna, this is 13 seconds. Soul is 23. So, it's the same job with the same questions.