900+ hours of Learning Claude Code/Cursor in 10 minutes

All right. So, this video is going to be the

right. So, this video is going to be the anti-vibec coders guide to building with AI. And it's basically going to be a lot

AI. And it's basically going to be a lot of things I wish I knew earlier about using clawed code and cursor. So, hacks

and a few things I've not seen online that have really helped me when I've been using AI to code. Because there's

two ways I think that you can use AI.

You can be like a pigeon that's just had brain surgery and you're just like continue continue and then you're getting all these errors and you have to keep restarting your project or rolling

back or use it in a more considered way and get the AI to work for you because these LLMs basically they're just guessing. The human's job is to make

guessing. The human's job is to make sure those guesses are better. So in

this video I'm going to show you how I do that. And I am a senior software

do that. And I am a senior software engineer. So over the last year, I've

engineer. So over the last year, I've seen my role just totally change to the point where this has become like one of the main skills people have to learn now if you're an engineer. So things like

prompting, context, like memory, and how to basically use AI effectively or if you're watching this and you just want to make a better app and just spend less money on tokens. These are hacks and

things I wish I knew earlier. So let's

get stuck in. Now I make content for developers, programmers, people who work in tech and a lot of people get pissed off when you talk about AI. They feel

that you become stupid by coding with AI and that it's an either or situation. I

think you can use AI, do it quickly, like build faster, but also learn at the same time. So later on in the video, I'm

same time. So later on in the video, I'm going to share how I do that because I thought about this loads, right? But

let's start with the number one mistake I made and I see people making when they're using claude and cursor to code.

And that is not using plan mode. Because

when I started coding, you wrote a lot of the code manually. So you didn't have to plan too much. You just write a few notes and then build out the feature because you had a lot of control and you were doing it step [music] by step. So

basically, you couldn't it up too much.

With AI, it's a lot different because it's [music] like a train and it codes so fast. It can go in the wrong

so fast. It can go in the wrong direction really quickly. And the way you like circumvent that is focusing a lot on getting the AI to plan the project for you. And it really can feel

like magic seeing this LLM just like generating all its code. The reason why that it's better to get the AI to do it with plan mode with things like Ultraink, get it to plan out the project

for you is that you can go back and forth with the AI and also it will just discover edge cases and side effects which you not thought about. So things

which will pop up and you'll be like oh that's actually a good question. So, it

will just save you a load of money the start of every project. Make sure you plan it. Now, the next one is one I

plan it. Now, the next one is one I struggled with for ages and I just put it off for a while because I'm a software engineer and I like typing. But

it's better to just dictate your prompts. So, to speak, use a voice

prompts. So, to speak, use a voice control feature and then just talk your voice prompts rather than typing them out. [music] And the reason why it's

out. [music] And the reason why it's better to do voice prompts because I've gone back and forth with this, but I've just noticed when I do voice prompts, when you're typing, you're kind of limited and is timeconuming. With voice

prompts, they're usually longer. I've

noticed for me, more specific, I add more context and also it's just a bit more human as well. So, this gives the AI better instructions than these like shorter typed prompts. So when you can,

obviously if you're in a cafe, whatever, it's a bit weird, but when you can, just try out doing just dictating your prompts and see how it goes. If you're

building an app in 2026, then you want to set up some automations on a platform like N. And it's best to self-host

like N. And it's best to self-host these. That way, you'll have control of

these. That way, you'll have control of your workflows and your data, and you don't have to pay cloud pricing for something that you can just run yourself. And you can do that with

yourself. And you can do that with hosting a VPS, who are sponsoring this part of the video. N is addictive and if you're a programmer and you're not making automations, you're missing out.

There is a lot of money to be made making these right now and it's fun. Let

me show you. So, I want to send one WhatsApp message like this and the automation does the rest. It all starts with this. The trigger my WhatsApp

with this. The trigger my WhatsApp message. That message is transcribed.

message. That message is transcribed.

Then the AI detects my email intent, identifies a contact and the Gmail thread, generates my reply and sends the email automatically. And then boom, it

email automatically. And then boom, it sends me a WhatsApp confirming it. Then

you got to host it. Go to this URL for hosting a VPS. It's up to four times cheaper. It's fast, unlimited workflows

cheaper. It's fast, unlimited workflows and one click and you're done. They've

even got a sale on now. You'll get 10% off with this code. Take this as your sign. Get involved in this AI automation

sign. Get involved in this AI automation error. Okay. Next is to use MCP servers.

error. Okay. Next is to use MCP servers.

And the best way I found to think about MCP servers is think of it like a USB port for AI to connect to the real world. So you can use it to pull in like

world. So you can use it to pull in like live data, APIs, and documentation into your project. And a really good one is

your project. And a really good one is called Context 7. And what this will do is just pull in all up to-ate documentation for you. And if you've like coded professionally, you've worked

on a professional codebase, one of the biggest pain in the is when for example in JavaScript, there's a load of packages and when one updates, it can affect others and particularly stuff

like routing. So MCP servers will give

like routing. So MCP servers will give your project just better context. But

there is some caveats here. You want to be careful about what you're bringing into your project. For example, there's an amazing superbase MCP server, but you're giving that a lot of access to your project and in theory it could

delete and tank your whole database and schema, but it's probably worth the tradeoff. So yeah, just be careful about

tradeoff. So yeah, just be careful about the access you're giving them, but long story short, better context equals better code. Okay, next one. Treat AI

better code. Okay, next one. Treat AI

like a junior developer with ADHD. Let

me explain. No disrespect to anyone with ADHD. [music] Um, but I think it's just

ADHD. [music] Um, but I think it's just like a nice like metaphor and it does make sense. It'll do things really quick

make sense. It'll do things really quick and really good as well and you're just like wo it just generated a load of code a feature and it's just like amazing.

That's exactly what I wanted. Other

times it'll do things really quick. But

you've got to like guide it a lot. You

got to it needs constant guidance, really tight context and also just give it specific tasks. Otherwise, it will guess, but it will guess really confidently and it will take you to like error hell. When you think about it, how

error hell. When you think about it, how LLMs work is their job is to please you by trying to fill in the blanks. And

your job in all this, your job as the human is to try and make those blanks better. And by using MCP servers, you

better. And by using MCP servers, you give it more context. Thus, you try and make the blanks better. Okay. Next, if

you're a junior developer and you're watching this and you're like, I like coding with AI, but I'm worried that I'm losing my coding ability. Or maybe

you're like a senior and you're just in the same boat where you're like, I don't think I'm learning as much. Maybe I'm

losing some of the craft of coding. This

is the mindset I would have. Obviously,

if you're building your own app and your goal is to build an own app, this this isn't important. I'd say the way I treat

isn't important. I'd say the way I treat it is like pair programming and treat it like a collaboration. So pair

programming, if you don't know, I've spent probably thousands of hours of my life pair programming. It's where you buddy up with someone who's usually like senior and just way better than you at the job. You get a problem together, you

the job. You get a problem together, you share a screen, and then you go back and forth. And usually they code a lot. You

forth. And usually they code a lot. You

ask questions and you just watch how they do things. If you really want to learn and grow, then treat coding with AI using cursor and claude like this. So

you're asking questions. You're like,

why did it do that? This is going to take longer, but you're going to learn more. So, it's that balance of being

more. So, it's that balance of being productive, but also learning. This is

with AI. Now, this is, I think, massive.

If you're working for a company, don't sacrifice your learning for their productivity. And another really

productivity. And another really important and helpful concept is this idea of compound engineering. So, like

money, your codebase and your AI should get smarter and smarter with each prompt and each iteration. So a big part of your role as the human being in all this is to just give it as much context to

allow it to compound engineer as a project goes on. So by the end of it, it knows what you want. It knows what you want the project or the app to look like. And a big way you can do this is

like. And a big way you can do this is by providing and using claude MD and cursor rules. So what kind of stuff do

cursor rules. So what kind of stuff do you put in these files? The way I think about it, it's like documentation for your app. So stuff that you keep

your app. So stuff that you keep repeating again and again like themes, how you want the components and stuff to look, if you find yourself repeating it again and again, just put it in these files. Start basic and then grow from

files. Start basic and then grow from there. And whenever your AI is

there. And whenever your AI is generating code, it's going to refer to these this documentation in the context.

You want to be careful though. You you

don't want to put too much in there because it will blow the context window.

Anything that you think will help, anything that you're repeating again and again, put it in these files. So, you're

helping compound engineering by using claw.md and cursor rules. Okay, the last

claw.md and cursor rules. Okay, the last one, last hack I really like. It's for

anyone who is designing an app and they just suck at design. Cuz what you can do, you can actually just steal designs and use AI to implement it and it actually works pretty well, I found. So,

there's a website called Mobin. It's a

good way to like browse different designs. If you see one you like that

designs. If you see one you like that you think is cool, take a screenshot of it. Go back to cla cursor and just say I

it. Go back to cla cursor and just say I want you to update the design of this app using this image. Then drag it in, but give it a constraint. Tell it only

use the color scheme and the font from this design. That's it. Otherwise, it's

this design. That's it. Otherwise, it's

just going to like change loads of stuff. But I did this the other week and

stuff. But I did this the other week and it worked really well. And that is it.

If you enjoyed the video, if you got any value whatsoever, all I ask, it's free for you. Just like the video if you want

for you. Just like the video if you want and subscribe if you enjoy this kind of content. I've also got a newsletter

content. I've also got a newsletter where I'm going to be posting like deeper insights into the app I'm making right now using AI. So, subscribe for that. I'll see you in the next one.

that. I'll see you in the next one.

Happy coding. Ciao.

[music]