WEBVTT 00:00:00.000 --> 00:00:02.339 You know that feeling, right? That moment of 00:00:02.339 --> 00:00:04.780 just pure magic. Oh, absolutely. It's intoxicating. 00:00:04.860 --> 00:00:07.660 You ask a large language model to write some, 00:00:07.660 --> 00:00:10.640 I don't know, complex block of code, and it just 00:00:10.640 --> 00:00:14.179 works perfectly. First try. You feel like you've 00:00:14.179 --> 00:00:16.640 skipped four years of a CS degree and gone straight 00:00:16.640 --> 00:00:19.300 to the finish line. Right. But then... Then you 00:00:19.300 --> 00:00:21.260 try to make one small change. You need to tweak 00:00:21.260 --> 00:00:23.260 a variable or maybe just fit it in your file 00:00:23.260 --> 00:00:25.600 structure, and boom. Everything breaks. Everything 00:00:25.600 --> 00:00:27.800 breaks. You spend the next two days chasing errors, 00:00:28.100 --> 00:00:30.280 breaking more things, until you're just staring 00:00:30.280 --> 00:00:33.899 at this tangled mess you don't understand. You 00:00:33.899 --> 00:00:36.359 quit. That mess, that is the classic result of 00:00:36.359 --> 00:00:38.799 what we call vibe coding. You're just vibing 00:00:38.799 --> 00:00:41.659 with the AI, no structure, no plan, and it's 00:00:41.659 --> 00:00:44.219 the number one reason these early AI projects 00:00:44.219 --> 00:00:47.259 fail. So today, we're taking a deep dive into 00:00:47.259 --> 00:00:50.759 how to escape that exact cycle. Our mission is 00:00:50.759 --> 00:00:52.759 to pull out the concrete systems, the mindset, 00:00:53.119 --> 00:00:55.320 used by places like Google and NVIDIA. Right. 00:00:55.340 --> 00:00:58.060 We want to move you from just, you know, guessing 00:00:58.060 --> 00:01:01.539 with prompts to reliably building real applications 00:01:01.539 --> 00:01:03.640 that last. We're going to transform you from 00:01:03.640 --> 00:01:06.719 a vibe coder into an AI engineer. Yeah. We'll 00:01:06.719 --> 00:01:09.019 cover the history, this idea of software 3 .0, 00:01:09.739 --> 00:01:12.280 the huge difference between the vibe coder and 00:01:12.280 --> 00:01:14.719 the architect, and the tools you actually need. 00:01:15.310 --> 00:01:17.689 And most importantly, this idea of the human 00:01:17.689 --> 00:01:21.409 in the loop. So let's unpack this. To really 00:01:21.409 --> 00:01:23.450 get why we're even talking about code this way, 00:01:23.670 --> 00:01:25.269 I think we need a little historical context. 00:01:25.650 --> 00:01:28.230 The way we build software has gone through, what, 00:01:28.290 --> 00:01:30.989 three big stages? Three fundamental shifts, yeah. 00:01:31.569 --> 00:01:33.530 So software 1 .0, that's the starting point. 00:01:33.670 --> 00:01:37.469 Think way back, like the 1940s onwards. The old 00:01:37.469 --> 00:01:39.760 school. Completely, it was all manual coding. 00:01:40.180 --> 00:01:42.420 Every single line was written by a human. If 00:01:42.420 --> 00:01:44.239 you use a cooking analogy, you were writing the 00:01:44.239 --> 00:01:46.459 recipe by hand. And if you forgot to write, add 00:01:46.459 --> 00:01:49.900 salt. The dish had no salt. Simple as that. Exactly. 00:01:50.019 --> 00:01:52.359 The computer only did what you explicitly typed. 00:01:52.879 --> 00:01:56.560 Then around 2012, things started to change. We 00:01:56.560 --> 00:01:58.540 entered software 2 .0 with machine learning. 00:01:58.680 --> 00:02:00.420 That's when we started hearing about neural networks. 00:02:00.640 --> 00:02:03.420 And a neural network is really just an algorithm 00:02:03.420 --> 00:02:05.840 that learns rules from huge amounts of data. 00:02:06.060 --> 00:02:08.280 It finds the patterns on its own. So we stopped 00:02:08.280 --> 00:02:11.099 writing the rules ourselves. We'd just feed the 00:02:11.099 --> 00:02:13.159 computer, say, a thousand pictures of a cat, 00:02:13.319 --> 00:02:15.099 and the system would figure out what cat means. 00:02:15.780 --> 00:02:17.939 Right. In that cooking analogy, it's like showing 00:02:17.939 --> 00:02:20.300 a chef a thousand photos of a perfect pizza. 00:02:20.659 --> 00:02:22.560 They'd have to figure out the recipe through 00:02:22.560 --> 00:02:25.060 trial and error. And that brings us to now, software 00:02:25.060 --> 00:02:29.039 3 .0. This really kicked off around 2019 -2020 00:02:29.039 --> 00:02:31.560 with large language models. This is where the 00:02:31.560 --> 00:02:34.360 game totally changed, because now we can program 00:02:34.360 --> 00:02:37.080 just using plain English. You don't need to know 00:02:37.080 --> 00:02:38.979 the technical syntax of the kitchen anymore. 00:02:39.439 --> 00:02:41.639 You just sit down, tell the waiter at the LLM 00:02:41.639 --> 00:02:44.120 you want a spicy, thin -crust pizza, and they 00:02:44.120 --> 00:02:46.900 handle the rest. And that simplicity is just... 00:02:46.879 --> 00:02:49.099 It's blowing up the market. I mean, the scale 00:02:49.099 --> 00:02:51.379 is hard to even imagine. You mean beyond just 00:02:51.379 --> 00:02:53.740 regular software? Way beyond. The current VW 00:02:53.740 --> 00:02:56.259 software market is, what, about $230 billion? 00:02:56.719 --> 00:03:00.080 But the entire labor market, paying people to 00:03:00.080 --> 00:03:02.840 do jobs, that's in the trillions. And now AI 00:03:02.840 --> 00:03:05.960 agents can do those jobs, answering emails, analyzing 00:03:05.960 --> 00:03:08.900 data. Exactly. So the market for what software 00:03:08.900 --> 00:03:12.139 can do is about to grow, maybe 50x, even 100x. 00:03:12.520 --> 00:03:14.939 The barrier to entry is just... gone. You don't 00:03:14.939 --> 00:03:17.340 need that formal degree anymore. You just need 00:03:17.340 --> 00:03:20.560 to communicate effectively. So if the market 00:03:20.560 --> 00:03:23.120 is going to expand that much, where should a 00:03:23.120 --> 00:03:26.020 beginner focus their energy first? Focus on learning 00:03:26.020 --> 00:03:28.599 to communicate effectively with the models, not 00:03:28.599 --> 00:03:31.780 on mastering old school programming syntax. That 00:03:31.780 --> 00:03:33.900 feels like the perfect transition into this core 00:03:33.900 --> 00:03:37.620 idea. Vibe coding versus AI engineering. It's 00:03:37.620 --> 00:03:40.439 the whole philosophy. And vibe coding at its 00:03:40.439 --> 00:03:44.139 heart is just It's lazy. You're leaning 100 % 00:03:44.139 --> 00:03:46.180 on the AI to get it right. You're not looking 00:03:46.180 --> 00:03:48.419 at the code, not planning the file structure. 00:03:48.560 --> 00:03:50.500 You probably don't even know which file does 00:03:50.500 --> 00:03:52.539 what. And the biggest consequence of that is 00:03:52.539 --> 00:03:55.039 something called technical debt. Right, the house 00:03:55.039 --> 00:03:57.939 analogy. If you're building a house just on instinct, 00:03:58.639 --> 00:04:00.580 technical debt is like building walls without 00:04:00.580 --> 00:04:02.400 checking if they're straight. It looks okay for 00:04:02.400 --> 00:04:04.710 a minute. For a little bit, yeah. But then you 00:04:04.710 --> 00:04:06.689 try to put the roof on or add a second floor, 00:04:06.969 --> 00:04:09.509 and the whole thing just collapses. That's why 00:04:09.509 --> 00:04:12.210 these vibe -coated projects, they almost always 00:04:12.210 --> 00:04:15.069 die within about 60 days. They just get too messy 00:04:15.069 --> 00:04:18.649 to fix. And look, I gotta be honest here. A bit 00:04:18.649 --> 00:04:21.209 of a vulnerable admission, but I still wrestle 00:04:21.209 --> 00:04:23.769 with prompt drift myself sometimes. You do? Oh 00:04:23.769 --> 00:04:26.649 yeah. It's so tempting, you know, when you're 00:04:26.649 --> 00:04:29.009 tired or you're in a rush, to just let the AI 00:04:29.009 --> 00:04:31.680 handle it all. But that's the trap. That is the 00:04:31.680 --> 00:04:34.040 60 day trap you have to fight every single day. 00:04:34.220 --> 00:04:37.060 So the way out is AI engineering. It's a total 00:04:37.060 --> 00:04:39.680 mindset shift. Completely. You are not the builder 00:04:39.680 --> 00:04:42.120 anymore. The AI is the build. You're the architect. 00:04:42.540 --> 00:04:44.660 You hold the map. You check the work. You understand 00:04:44.660 --> 00:04:47.699 the foundational pieces. That control is how 00:04:47.699 --> 00:04:50.360 you build things that last for years and actually 00:04:50.360 --> 00:04:52.680 make money. So beyond that architectural planning, 00:04:52.980 --> 00:04:55.139 what's the single most important tool to help 00:04:55.139 --> 00:04:57.620 a beginner stay in that architect role? You need 00:04:57.620 --> 00:05:00.800 a tool that gives the AI full context over your 00:05:00.800 --> 00:05:03.319 whole project instantly. Which brings us to the 00:05:03.319 --> 00:05:05.980 toolkit. Exactly, the tools that enforce that 00:05:05.980 --> 00:05:08.300 structure. For level one, where everyone should 00:05:08.300 --> 00:05:10.660 start, the tool a lot of engineers are recommending 00:05:10.660 --> 00:05:14.360 right now is cursor. It looks just like VS Code, 00:05:14.399 --> 00:05:16.519 which most people know, but it has the AI baked 00:05:16.519 --> 00:05:20.100 right in. It's all about creating project -wide 00:05:20.100 --> 00:05:23.100 context. Meaning it knows about all your files 00:05:23.100 --> 00:05:25.899 all the time. Yes. And there are three features 00:05:25.899 --> 00:05:28.029 in there you just have to master. The first is 00:05:28.029 --> 00:05:30.370 called cursor tab. We call it the mind reader. 00:05:30.649 --> 00:05:32.889 The mind reader, I like that. As you start typing 00:05:32.889 --> 00:05:35.730 maybe a variable or a function, you'll see this 00:05:35.730 --> 00:05:39.389 faint gray text appear suggesting the rest of 00:05:39.389 --> 00:05:41.410 the code. You just hit tab and it fills it in. 00:05:41.470 --> 00:05:43.509 It feels like it's reading your mind. Okay, that 00:05:43.509 --> 00:05:45.509 handles the boilerplate. What's next? The second 00:05:45.509 --> 00:05:48.240 is command K. This is your editor. This is for 00:05:48.240 --> 00:05:50.980 changing existing cart. You highlight a small 00:05:50.980 --> 00:05:52.920 chunk of code, just a few lines. So you're being 00:05:52.920 --> 00:05:55.720 very specific. Extremely. You highlight it, hit 00:05:55.720 --> 00:05:58.420 command K, and say, add error handling here, 00:05:58.480 --> 00:06:01.420 or make this cleaner. It only changes that one 00:06:01.420 --> 00:06:04.019 specific part, which limits the blast radius 00:06:04.019 --> 00:06:06.519 of any mistake. And the third feature, the most 00:06:06.519 --> 00:06:10.000 powerful one, is command I, the composer. This 00:06:10.000 --> 00:06:12.759 is the big one. It opens a sidebar where you 00:06:12.759 --> 00:06:15.019 can talk to the AI about the entire project. 00:06:15.079 --> 00:06:17.040 It's read everything. So you can actually say 00:06:17.040 --> 00:06:19.980 something like, create a new file for user login. 00:06:20.439 --> 00:06:22.500 And it will create the file, write the code, 00:06:22.639 --> 00:06:25.519 and this is the crucial part. Link it to your 00:06:25.519 --> 00:06:29.000 existing navigation files automatically. It manages 00:06:29.000 --> 00:06:31.800 multiple files at once. Whoa, wait a second. 00:06:31.920 --> 00:06:35.220 So you're saying you could scale that across 00:06:35.220 --> 00:06:38.949 a massive... data pipeline, or use it to manage, 00:06:38.949 --> 00:06:41.329 I don't know, a billion queries without having 00:06:41.329 --> 00:06:44.250 to link anything by hand. That's the power. It's 00:06:44.250 --> 00:06:45.910 truly transformative. It makes the architect 00:06:45.910 --> 00:06:48.250 so much more productive. Okay, quick practical 00:06:48.250 --> 00:06:51.430 tip. For the model choice, you mentioned starting 00:06:51.430 --> 00:06:54.610 with Claude 3 .5 Sonnet. Why that one specifically? 00:06:55.189 --> 00:06:57.769 For beginners doing coding tasks, my advice is 00:06:57.769 --> 00:07:00.160 to start with Sonnet. The general feeling in 00:07:00.160 --> 00:07:01.879 the engineering community right now is that it 00:07:01.879 --> 00:07:04.399 follows complex multi -step instructions really, 00:07:04.399 --> 00:07:06.360 really well. It just reduces friction. And once 00:07:06.360 --> 00:07:08.259 you're good with Cursor, you move to level two, 00:07:08.259 --> 00:07:11.220 which is coding agents. Yeah, we use the intern 00:07:11.220 --> 00:07:13.800 analogy for this. Cursor is a tool you hold. 00:07:14.180 --> 00:07:16.699 An agent is like an intern you hire. You give 00:07:16.699 --> 00:07:19.300 it a task and walk away. Exactly. You say, build 00:07:19.300 --> 00:07:21.439 me a login page. And you come back in 20 minutes. 00:07:21.500 --> 00:07:24.800 It runs. It finds its own errors. It fixes them. 00:07:25.079 --> 00:07:27.939 And it tries to complete the entire job by itself. 00:07:28.360 --> 00:07:31.139 What are some good examples to start with? Replet 00:07:31.139 --> 00:07:33.480 Agent is a good one. It's in the browser, very 00:07:33.480 --> 00:07:36.100 friendly for beginners. Windsurf is another. 00:07:36.720 --> 00:07:39.180 Terminal agents are a bit more advanced, so I'd 00:07:39.180 --> 00:07:41.379 say stick to the Bryzer -based ones first. So 00:07:41.379 --> 00:07:43.899 with all these tools that handle the actual typing, 00:07:44.519 --> 00:07:46.779 what concepts are absolutely necessary for the 00:07:46.779 --> 00:07:48.639 architect to understand before they even start? 00:07:48.860 --> 00:07:51.600 Forget the syntax. You need to focus intensely 00:07:51.600 --> 00:07:54.579 on understanding the core software concepts first. 00:07:54.889 --> 00:07:56.529 And that really is the secret sauce, isn't it? 00:07:56.649 --> 00:07:58.889 It's everything. You don't need to memorize where 00:07:58.889 --> 00:08:01.370 the semicolon goes anymore. The AI does that. 00:08:01.670 --> 00:08:03.670 But you have to learn the high -level concepts 00:08:03.670 --> 00:08:06.430 so you can guide the AI's decisions logically. 00:08:06.689 --> 00:08:09.509 OK, so let's list them. The five essential concepts 00:08:09.509 --> 00:08:12.089 to master. You should ask your AI to explain 00:08:12.089 --> 00:08:15.110 these to you. Number one, front -end versus back 00:08:15.110 --> 00:08:18.129 -end. Two, what a database is. It's basically 00:08:18.129 --> 00:08:22.089 a digital filing cabinet. Three, the ATI. The 00:08:22.089 --> 00:08:24.089 waiter analogy is perfect for this. It's the 00:08:24.089 --> 00:08:26.649 go -between for two different systems. Four. 00:08:27.009 --> 00:08:28.970 Environment variables. These are your secret 00:08:28.970 --> 00:08:31.189 keys and passwords. And this one's critical. 00:08:31.230 --> 00:08:33.669 Oh, it's so critical. If you don't understand 00:08:33.669 --> 00:08:37.009 this concept, you risk the AI accidentally pushing 00:08:37.009 --> 00:08:39.909 your secret keys, your digital wallet, to a public 00:08:39.909 --> 00:08:41.789 place like GitHub. Which happens more than you 00:08:41.789 --> 00:08:44.629 think. It really does. And the fifth concept 00:08:44.629 --> 00:08:46.789 is deployment, which is just the process of putting 00:08:46.789 --> 00:08:49.049 your app live on the internet. So how do you 00:08:49.049 --> 00:08:52.230 learn these? You use the AI to teach you. Open 00:08:52.230 --> 00:08:54.850 Cursor's chat, that's command L, and type in 00:08:54.850 --> 00:08:57.250 a very specific prompt. Something like, I'm a 00:08:57.250 --> 00:09:00.389 beginner. Explain what an API is using a simple 00:09:00.389 --> 00:09:03.169 real world analogy. Do not use technical jargon. 00:09:03.230 --> 00:09:05.370 You have to be the manager, not the intern. Control 00:09:05.370 --> 00:09:08.090 the decisions. A bad prompt is asking, what should 00:09:08.090 --> 00:09:11.169 I build? It's too passive. Right. A good prompt 00:09:11.169 --> 00:09:13.750 is specific. I want to build a to -do list app. 00:09:14.129 --> 00:09:16.350 Give me three tech stack options and explain 00:09:16.350 --> 00:09:18.750 the pros and cons of each for a beginner. Then 00:09:18.750 --> 00:09:21.250 you read them and you choose the path. Let's 00:09:21.250 --> 00:09:23.230 talk about the common mistakes, the tuition fee 00:09:23.230 --> 00:09:26.190 you pay when you're learning this. Yes. Mistake 00:09:26.190 --> 00:09:29.509 number one. Accepting code blindly. You never 00:09:29.509 --> 00:09:31.690 just click apply. You have to read the comments, 00:09:31.870 --> 00:09:33.870 check which files it's going to affect, and then 00:09:33.870 --> 00:09:36.669 commit. Number two is changing too much at once. 00:09:36.950 --> 00:09:39.679 Oh, this is a huge one. Don't try to build the 00:09:39.679 --> 00:09:42.500 whole app in one prompt. Go install steps. Build 00:09:42.500 --> 00:09:45.480 the login form. Check it. Now, build the database 00:09:45.480 --> 00:09:47.860 connection. Check it. It's like stacking Lego 00:09:47.860 --> 00:09:51.139 blocks one at a time. And the third mistake is 00:09:51.139 --> 00:09:54.159 giving up when it gets hard. The first hour is 00:09:54.159 --> 00:09:56.259 always magic. But then you hit that first real 00:09:56.259 --> 00:09:59.019 bug, and it takes 30 minutes of just grinding 00:09:59.019 --> 00:10:01.419 to fix it. That 30 minutes? That's the tuition 00:10:01.419 --> 00:10:03.659 fee. You learn more in that struggle than you 00:10:03.659 --> 00:10:06.250 do from any... piece of perfectly generated code. 00:10:06.690 --> 00:10:08.309 You just have to push through it. So how can 00:10:08.309 --> 00:10:10.730 an engineer ensure their code doesn't secretly 00:10:10.730 --> 00:10:13.769 revert to that vibe code messiness over time? 00:10:13.970 --> 00:10:17.049 Schedule regular AI reviews to find unused code 00:10:17.049 --> 00:10:19.549 and refactor for cleanliness. OK, let's put this 00:10:19.549 --> 00:10:21.350 all together. Let's walk through building a simple 00:10:21.350 --> 00:10:23.669 daily expense tracker step by step the way an 00:10:23.669 --> 00:10:26.090 architect would. Perfect. So step one is the 00:10:26.090 --> 00:10:28.690 setup. You create a PRD, a product requirements 00:10:28.690 --> 00:10:31.330 document. Which is really just a simple text 00:10:31.330 --> 00:10:34.460 file, plan .md. And the prompt you give the AI 00:10:34.460 --> 00:10:38.519 is key. You say, goal, an app to add expenses 00:10:38.519 --> 00:10:41.740 and see a chart, tech stack, Python and Streamlet, 00:10:41.980 --> 00:10:46.100 dot data, a simple CSV file, write a step -by 00:10:46.100 --> 00:10:48.669 -step plan, and you add this sentence. Do not 00:10:48.669 --> 00:10:50.909 write any code yet. You have to insist on the 00:10:50.909 --> 00:10:53.610 plan first. Always. Step two is architecture. 00:10:54.129 --> 00:10:56.070 You review that plan and then you ask the AI 00:10:56.070 --> 00:10:59.129 to create a detailed readme .md file that lists 00:10:59.129 --> 00:11:01.070 the folder structure and explains what every 00:11:01.070 --> 00:11:02.850 single file is for. So you know the map before 00:11:02.850 --> 00:11:04.970 you start. You're not working blind. Exactly. 00:11:05.590 --> 00:11:07.710 Then steps three and four are phased coding. 00:11:07.960 --> 00:11:10.600 You start with the skeleton, use command i to 00:11:10.600 --> 00:11:12.460 create the first few files like requirements 00:11:12.460 --> 00:11:15.740 .txt and a basic hello world app .py y. You run 00:11:15.740 --> 00:11:18.120 it, make sure it works, get that small win. Right. 00:11:18.159 --> 00:11:20.399 Then you phase in the logic, you add the input 00:11:20.399 --> 00:11:22.600 form for the expense, the amount, the category, 00:11:22.860 --> 00:11:25.159 and tell it to save that data to a CSV file. 00:11:25.240 --> 00:11:27.039 And then comes step five, which is inevitable, 00:11:27.159 --> 00:11:29.440 the review and fix. You're going to hit an error, 00:11:29.639 --> 00:11:31.720 maybe a file not founder. This is the learning 00:11:31.720 --> 00:11:34.519 moment. You don't panic. You use the debug with 00:11:34.519 --> 00:11:38.269 AI button or paste the error in. and ask why 00:11:38.269 --> 00:11:40.409 it happened. And that's where you learn a core 00:11:40.409 --> 00:11:42.649 concept, like realizing you forgot to create 00:11:42.649 --> 00:11:45.210 the CSV file in the first place. Finally, step 00:11:45.210 --> 00:11:49.129 six is the Polish. Once the core works, you prompt 00:11:49.129 --> 00:11:52.090 the AI to add a nice pie chart that shows your 00:11:52.090 --> 00:11:55.549 spending by category. It's iterative, it's controlled. 00:11:55.950 --> 00:11:58.549 So if someone follows this structure and builds 00:11:58.549 --> 00:12:00.970 this little app, what's the fastest way for them 00:12:00.970 --> 00:12:03.830 to build real world confidence? Don't just build 00:12:03.830 --> 00:12:06.460 it locally. You have to deploy the application 00:12:06.460 --> 00:12:09.000 live onto the internet. That's key. That really 00:12:09.000 --> 00:12:11.080 is the core concept we wanted to get across today. 00:12:11.460 --> 00:12:12.960 The difference between people who make money 00:12:12.960 --> 00:12:15.139 with AI and those who are just playing with it 00:12:15.139 --> 00:12:17.919 is it's all about structure. Be persistent. Be 00:12:17.919 --> 00:12:20.059 logical. Own your code. You don't need to be 00:12:20.059 --> 00:12:22.100 a math genius anymore. You just need to be the 00:12:22.100 --> 00:12:24.679 architect. And there is a wide open seat at the 00:12:24.679 --> 00:12:27.019 table for you. The software world is going to 00:12:27.019 --> 00:12:29.899 grow massively next year. So here's a personal 00:12:29.899 --> 00:12:32.399 challenge, a learning roadmap for the next four 00:12:32.399 --> 00:12:35.070 weeks. Let's hear it. Week one. Just install 00:12:35.070 --> 00:12:38.309 Cursor. Get comfortable with Command -K. Build 00:12:38.309 --> 00:12:41.009 a Hello World app. That's it. Simple start. Week 00:12:41.009 --> 00:12:44.269 two. Build a calculator or to -do list. And focus 00:12:44.269 --> 00:12:47.629 specifically on using Command -I to work across 00:12:47.629 --> 00:12:50.529 multiple files. Get a feel for that project context. 00:12:50.970 --> 00:12:53.250 Okay, what about week three? Week three is all 00:12:53.250 --> 00:12:55.929 about deployment. Ask the AI how to put your 00:12:55.929 --> 00:12:58.529 little Streamlet app live on the internet. Learn 00:12:58.529 --> 00:13:01.450 that process from end to end. And the final week? 00:13:01.769 --> 00:13:04.970 Week four. Tackle the hard stuff. Ask the AI 00:13:04.970 --> 00:13:07.730 to teach you about APIs and then build a simple 00:13:07.730 --> 00:13:09.629 app that uses live data maybe from a weather 00:13:09.629 --> 00:13:11.809 service. So don't be a vibe coder who relies 00:13:11.809 --> 00:13:14.889 on magic. Be an AI engineer who relies on structure. 00:13:15.250 --> 00:13:17.649 Exactly. What simple structured project will 00:13:17.649 --> 00:13:18.529 you build this weekend?