0:00 The smartest people on Earth do not use AI like you do. And the difference is not a better model or some sort of fundamentally different tool. It is almost entirely in how they talk to it. Uh so, I've made over $10 million over the course of the last few years with AI and AI tools, and I want to show you seven simple upgrades you can make to the way that you work with AI to make significantly more and significantly less time. The first is talk, don't type. The average person writes at approximately 30 to 50 words per minute with a keyboard, but they speak at 150 to 200 words per minute. Not every one of those words is going to be perfect, but when you do the math, you end up being able to get more out onto AI than you can if you just use your fingers. Don't let your fingers be the bottleneck. Your voice box was developed for far more than that. The simplest way to get this set up is use a free voice dictation tool. You don't need to pay any money to anybody. Just type like free voice dictation tool on Google if you don't believe me. Otherwise, um use a tool called Hex. I've used it myself. You also have the ability to like train it on a bunch of your particular words 0:50 and vocabulary uh to make it a lot better. The next is set up some sort of always-on voice stream during all of your recordings and meetings. So, the world's best companies take advantage of voice dictation software, things like GPT Real-Time 2.1, to basically always have an assistant listening to everything they do and say during a meeting or some other sort of social activity. Um you can think of this as an EA that is always with you taking meeting notes and then turning that into some sort of actionables or project file. The craziest version of this is I built a couple of systems that will take what you say during the meeting and actually go and build the thing that you said you needed to build. So, you know how many to-dos do you generate during a meeting? Oh, Ashley needs to do this. Sam needs to do this. Sarah needs to do this. You guys can actually go and like do the task during the meeting, so she have the results back while you are still inside meeting. I'm not a fan of meetings, but this makes them like a 100x more productive. The simplest way to set this up is use a frontier LLM that is the best LLM that you can, and say, "Hey, I want to build something 1:44 that essentially listens to my mic and maybe screenshots my page and listens to the system audio so that it can hear everything that goes on during a meeting, and then I want you to be able to transcribe the whole thing, and and use it to actually kick off real projects in the systems that I use day to day, whether it's like Notion or Asana or Monday or one of the many ones. This sounds really crazy, but you can actually build the system and have it running extraordinarily cheaply for like a dollar or two a day, and then this dollar or two a day system can legitimately replace a hire that might cost you, you know, 300 or 400 bucks a day if you wanted like a really competent EA. We're actually there right now with that technology, believe me. The next is to visualize things using 3.js, HTML, or some sort of visual renderer like Unreal Engine. Um rather than tell, the best people always show. So, they don't spend a bunch of time in word tokens, okay? A picture is worth a thousand words, and then a video is worth 24 to 30 pictures a second. Um so, that's what they do. They'll create visualizations, usually using tools like 3.js, HTML, Unreal Engine, Blender, and 2:39 so on and so forth. And they actually have the agents do it for them to explain things better. So, an example is maybe your company's getting a new office, or maybe you are moving into a new home. Um don't just look at the floor plan. Actually, like tell AI, "Hey, create me a 3D visualization of this with realistic lighting conditions, uh pulling images off of Google Maps and stuff like that to look out the windows, and then plot me in it so I can walk around and see what it looks like. Then I want you to add real furniture, let's say, with Amazon checkout links such that if I like uh product, then I can actually just click a button and then buy it." You you could have it uh visualize dense, complicated reports. You could have it visualize um analytics that occur on a daily or weekly basis. You could turn it into an animated explainer just like the one that you guys are currently watching. Basically, just use AI to show, not tell. Next is to have AI interview you. So, all top AI users are doing some version of this right now. Uh we've had a variety of different skills that have come out of this like grill me, interview me, roast me, and so on. But basically, if you 3:29 want to get an AI to do what you want to do, it's not enough just to tell it what you want it to do. A lot of the time, you yourself are bottlenecked by a lack of knowledge, and also a lack of explicit verbalization ability of the thing that you want. And so, better than being again bottlenecked, piping the ocean through a straw of your experience, just have AI ask you, "Hey, I want to do X, Y, and Z. Could you just ask me anything and everything that I might need to know in order to give you the best way of doing the task? If you feed this to one of the frontier models, it will do an extraordinary job, no problem. The next is do not ask for opinions, ask for flaws. AI models are trained using reinforcement learning through human feedback or RLHF. That means that oftentimes they're rewarded more for agreeing with you than actually giving you the right answer. So, in consumer applications will typically default to agreeing with you even if your answer is wrong. That means you cannot just ask it, "Hey, what do you think about this? Hey, is this email all right? Hey, what do you think about my, you know, project?" If you do that, it will always default to saying something 4:23 good. And maybe that's not going to be the case in a couple years from now, but it definitely is right now. So, instead of asking for generalized opinions, always ask for what is wrong with some deliverable or output. Um always say, "You know, what is wrong with my idea? Uh poke holes in my thoughts. Uh you know, roast my essay." The whole idea is like it needs to hear from you that you are explicitly asking it to give you bad feedback. And then with that bad feedback in mind, you yourself can be the judge the bad outweighs the good. Next, if you're skeptical about whether AI can do something, literally just ask it to do the thing for you. Do not rely on opinions from people like me or people on the internet. In the time that it takes to even finish this video, you probably have could have gotten a frontier model to do a lot of the things that you were wondering, "Hey, you know, can I do this with a frontier model?" And I'm not saying this cuz I don't want you to watch my YouTube videos, obviously, but because I want you to be very familiar with and understand that things are changing so quickly nowadays. They're accelerating so fast that what I say today may not be necessarily true 5:18 tomorrow. Even if it's the same model, there might be a new way of doing things. Some sort of harness update under the hood. Um you know, literally every day that passes, capabilities updates, and then benchmark scores are going up by like percentage points such that, you know, at the end of the month, uh performance on a benchmark that previously might have been considered impossible might now be 50% or more. So, this is what top CEOs do. They actually go and they actually try the thing themselves. You know, Jensen Huang, for instance, he's not just like listening to people talk to him all day. He actually has Claude Code, Codex, Nvidia's NeMo try, whatever the hell models he's running um set up on his computer and he is typing things in and he is voice transcribing things in. You know, like big CEOs actually have these tools set up on their computers and they are trying these things constantly because they understand that first principles is typically the best way to answer the question, is this thing possible? Best way to answer the question, is this thing possible, is by going out and doing it at least once. Finally, have an open mind. These AI 6:08 tools are going to diversify into many forms. You'll probably have like AI avatars on your screen pretty soon that are chatting with you just like a co-worker might. You also might have totally different ways of doing things. You might have different sorts of hardware, little pendants that you wear and stuff like that. Um try them out, right? Just like the previous one, you're not going to know until you know and the only way to know is to try. Either you're going to try, succeed, or you're going to fail and that's totally okay either way. You will have explored that opportunity and the opportunity cost of not having done so is typically so god damn high these days with tech being what it is uh that it's just a logical decision. Hopefully you guys appreciated this video. If you guys want to learn more about how to use AI like a business person, somebody that is a lot, definitely check out Maker School. It's the first link in the description. It's my 90-day accountability program where I will help you get your very first customer for an AI or automation service. If I don't, I give you all of your money back. Aside from that, please like, comment, and subscribe and I'll 6:58 catch all y'all in the next video. Thanks.
