open source project just dropped that when you combine it with claude code literally becomes self-improving AI. This is not engagement farming or hype bait. This is a real repo that was just released by Andre Karpathy who is widely renowned as one of the foremost voices in AI and machine learning research. And basically what
Поиск по транскрипциям
make this as fast as humanly possible, both for myself and for you guys, what I'm going to do is I'm going to open up three separate Claude Code little plug-in widgets here. And in the first I'm going to switch all these to bypass permissions and make this a little bit smaller so you guys
done this what three times now, I'm not going to bore you all with the details, but suffice to say, using Gemini 3. 1 Pro in conjunction with Claude Code for both a mixture of front-end design and then backend configuration, route creation, and then payments, and so on and so forth, is pretty dang efficient
just going to run through the motions that I've already showed you guys multiple times at this point before talking a little bit about the economics of vibe coding and then wrapping up. Man, I am so excited for this app because as somebody that produces a tremendous amount of content nowadays, having an app that can basically allow...time I'm going to voice dump everything that I think this app is going to include. Um and then I'm going to send it over to claude code so that I can have two of these instances operating simultaneously. One that's probably a little bit better at architecture design than the other that you know is better
freaking love Cloud Code skills. I think you do, too. But sometimes they're a little bit unreliable. I would say about 70% of the time I run a skill, I get an intended...output. About 30% of the time, it's a bag of rocks. What I wanted to do in this video is I wanted to show you how to combine Claude Code skills with a new development in the AI space called Auto Research to achieve significantly higher reliability, accuracy, and allow your skills to quite literally improve themselves overnight
this uh one back up here. Let's open that up. Okay, one last look at what we did during this stream. All right. So, we had AI Claude code specifically with the front-end design skill create a landing page for this. And then we used a bunch of different techniques to end up with this an objectively better
think we probably just ran into that. Okay, just ran out of usage again after developing this for a good hour or so. So, I've moved over to Claude Code. Uh I'm still going to try pushing and generating the variants here. We're going to see how that goes. 22 Mississippi, 23 Mississippi, 24 Mississippi, 25 Mississippi
this link, and then go straight over to my anti-gravity, and I'll say use this skill to design a high-end website about interior design. From here, Claude Code is going to fetch the skill repository to understand what it provides. And I'm just going to give it a good 2 minutes or so to actually
here is that you should use AI to, you know, write your campaigns. They think I'm going to show them a big long script or prompt or a claude code skill that effectively shows you how to go from no campaign written to a full campaign in front of you. They probably think that because you know colloquially
with less to get more stuff done in my business. I've put this into practice over the past 3 weeks and built my own AI operating system using claude code which we'll go into in a second. Uh that allows me to have this workspace that not only takes a lot of work off my plate, but also
What you're going to find with this workflow is you'll have a decent amount of downtime. What do you mean by that? Have you ever used clawed code? You gave it a command and it just started working for 20 minutes. Let me ask you, what did you do during those 20 minutes? Be honest. I can take...road map. So instead of you going here and just doom scrolling for an hour while the AI chat's working, and I know what you do, claude code builds for 5 minutes, but you end up doom scrolling for half an hour and you lose a ton of time. Instead, what we're doing is we're being productive
floating around on GitHub as people try to make their own versions of this and in many ways you can make your own version of this with something like claude code with something like lang chain langraph or even things like agent development kit from Google. The thing that made Open Clause so different though was in many ways
secure important victories and climb up the rankings. So, look at that. It autocompleted our notes for us. The AI works. It really is that simple. We had Claude Code build out the AI functionality. And we took our API key from OpenAI and we put it in. I like using OpenAI for the API stuff. It makes it really
here's why this matters more than you think. Every model has blind spots. GPT is amazing at reasoning and coding. Claude is incredible at planning and handling huge tasks. Gemini is a beast at research and working with big data. But none of them are perfect at everything. So when you only use one, you get one perspective
operating on my Mac Mini. I'll go ahead and fire up the terminal. Move to a temp directory. And so, let's go ahead and boot up a claude code instance here. Let's run a ping just to make sure we're all good on the device. Good. And actually, we don't even have the library installed
Как сэкономить 10 миллионов токенов в нейросети Claude Code? Допустим, у вас нейронка пишет код за вас. Она создает за вас приложения, телеграм-ботов, расширения, пишет сайты. И во время своей работы нейронка все время вызывает какие-то команды в терминале. И эти команды в терминале, они огромные. Там очень много повторяющихся символов, каких-то лишних цифр, букв, которые забивают
like this is how I'm basically translating the application into different languages. And what I'll do is make a brand new skill where every time Claude Code adds a brand new string into application, then it adds it into these two files here, and it finds a relevant point to add it into. You can also do that
like UI related components. And also the questions that it has been asking me seems to be more natural than the questions that I've been getting from Claude Code in the plan interactive questions mode. So I think there's a lot of room in this like planning sort of domain for development, and I'm pretty excited
prompt, let's just put in our Claude rules here. These are going to be rules that are included as context to every single prompt that we give to Claude automatically. We don't even have to tag it. So, here are the rules. I'm putting them down below. Feel free to copy and paste them. Pause here...these rules are inspired by things he's done with his AI agents. Very helpful. Okay, so here's the prompt we are going to give Claude code. This is an in-depth one, but this
didn't do that yet. Um and then all my other tasks like I have calls with a couple people today. Um, I'm writing a Claude code safety article. I've got to add some images to it. All this stuff just shows up on my to-do list every day based on what I put in my tasks