WEBVTT 00:00:00.000 --> 00:00:02.220 Think about the subscriptions quietly draining 00:00:02.220 --> 00:00:04.559 your bank account right now. Oh yeah, it adds 00:00:04.559 --> 00:00:08.220 up so fast. $300 a month for tools that still 00:00:08.220 --> 00:00:10.179 don't really fit how you actually work. Yeah. 00:00:10.279 --> 00:00:12.759 What if you could build a perfect private app 00:00:12.759 --> 00:00:15.519 for $10 in a single weekend? I mean, it sounds 00:00:15.519 --> 00:00:18.719 completely impossible, but the crazy thing is 00:00:18.719 --> 00:00:23.399 it's actually real. Welcome to today's deep dive. 00:00:23.920 --> 00:00:27.739 we are unpacking Max Anne's 2026 AI Coding Guide 00:00:27.739 --> 00:00:31.039 to explore what he calls the hyperspecific app 00:00:31.039 --> 00:00:33.920 framework. We'll cover what these apps are, look 00:00:33.920 --> 00:00:36.320 at real -world examples, and explore the exact 00:00:36.320 --> 00:00:39.429 five -step building framework. Plus, we'll discuss 00:00:39.429 --> 00:00:41.649 the massive mindset shift from renting software 00:00:41.649 --> 00:00:44.329 to building it yourself. And honestly, the subscription 00:00:44.329 --> 00:00:46.810 fatigue is just so real right now. I think you 00:00:46.810 --> 00:00:48.789 feel it. We all feel it. It's like death by a 00:00:48.789 --> 00:00:51.369 thousand, $10 monthly charges. Exactly. Let's 00:00:51.369 --> 00:00:53.390 start with that exact problem. You routinely 00:00:53.390 --> 00:00:56.149 pay $300 or more a month for scattered software 00:00:56.149 --> 00:00:58.609 tools. Right. You've got invoices in one app, 00:00:58.750 --> 00:01:01.149 habit tracking in another, calendars over here, 00:01:01.250 --> 00:01:03.539 notes over there. Right. And none of them actually 00:01:03.539 --> 00:01:05.459 talk to each other. You just end up copying and 00:01:05.459 --> 00:01:07.819 pasting data constantly to get your work done. 00:01:07.959 --> 00:01:11.090 They force compromises. You end up adapting your 00:01:11.090 --> 00:01:13.549 personal workflow to fit the generic tool. Yeah. 00:01:13.750 --> 00:01:15.450 When it really should obviously be the other 00:01:15.450 --> 00:01:18.170 way around. But commercial software companies, 00:01:18.349 --> 00:01:20.310 they don't really have a choice here, do they? 00:01:20.430 --> 00:01:22.409 No, they don't. They have to serve millions of 00:01:22.409 --> 00:01:24.290 people just to keep the lights on. They build 00:01:24.290 --> 00:01:26.530 generic features aiming right for the middle 00:01:26.530 --> 00:01:29.109 of the bell curve. Right. If a feature only helps 00:01:29.109 --> 00:01:31.629 1 % of their user base, they simply can't afford 00:01:31.629 --> 00:01:34.049 to develop it. Which brings us to the core concept 00:01:34.049 --> 00:01:38.290 of this guide, the hyper -specific app. Narrow 00:01:38.290 --> 00:01:41.489 focus is its absolute superpower. Exactly. A 00:01:41.489 --> 00:01:44.129 hyper -specific app is a small piece of software 00:01:44.129 --> 00:01:47.769 built with AI to solve one exact problem for 00:01:47.769 --> 00:01:50.870 one user. You. Yeah, it's the exact opposite 00:01:50.870 --> 00:01:54.090 of commercial size. Using generic software is 00:01:54.090 --> 00:01:56.349 kind of like wearing someone else's tailored 00:01:56.349 --> 00:01:58.969 suit. It covers you, sure. Yeah, it technically 00:01:58.969 --> 00:02:02.560 functions as clothes. Right. but it's never quite 00:02:02.560 --> 00:02:04.879 comfortable and it constantly bunches up in the 00:02:04.879 --> 00:02:07.040 wrong places. That is a perfect way to look at 00:02:07.040 --> 00:02:08.979 it. You don't wait for a product roadmap anymore. 00:02:09.199 --> 00:02:12.259 You just create a tool that matches your exact 00:02:12.259 --> 00:02:15.159 daily rhythm. But looking at this shift, I have 00:02:15.159 --> 00:02:18.539 to ask, is this movement really just about saving 00:02:18.539 --> 00:02:21.900 money or is it fundamentally about removing workflow 00:02:21.900 --> 00:02:25.009 friction? Oh, it is entirely about the friction. 00:02:25.229 --> 00:02:28.129 The money is a nice bonus, sure. But when a tool 00:02:28.129 --> 00:02:30.530 actually works the way your brain works, you 00:02:30.530 --> 00:02:32.909 get your time and your sanity back. So it's about 00:02:32.909 --> 00:02:35.370 perfect fit, eliminating the friction of generic 00:02:35.370 --> 00:02:38.729 software. Beat. Spot on. Now that we know what 00:02:38.729 --> 00:02:41.509 a hyper -specific app is, what do they actually 00:02:41.509 --> 00:02:44.469 look like in practice? The guide highlights four 00:02:44.469 --> 00:02:46.969 examples that really show the art of the possible. 00:02:47.129 --> 00:02:49.210 Yeah, let's start with the cross -border accounting 00:02:49.210 --> 00:02:51.930 app. Okay. This one addresses a huge pain point. 00:02:52.050 --> 00:02:54.229 Different currencies, conflicting tax rules, 00:02:54.550 --> 00:02:56.349 generic accounting software just breaks down 00:02:56.349 --> 00:02:58.849 completely here. Wait, if you're feeding sensitive 00:02:58.849 --> 00:03:02.169 financial data into an AI, isn't that a massive 00:03:02.169 --> 00:03:04.550 privacy risk? Well, that's the genius of this 00:03:04.550 --> 00:03:07.590 specific build. They run the app locally using 00:03:07.590 --> 00:03:10.770 open source AI models, so the data never actually 00:03:10.770 --> 00:03:13.810 goes to a cloud provider. Ah, I see. So it stays 00:03:13.810 --> 00:03:16.210 entirely private on their machine. Exactly. That's 00:03:16.210 --> 00:03:18.729 huge for business owners. Yeah. So what if the 00:03:18.729 --> 00:03:21.430 bottleneck isn't numbers, but visual branding? 00:03:22.050 --> 00:03:24.990 That brings us to the second example, a brand 00:03:24.990 --> 00:03:27.590 -specific slide generator. Yeah. Presentation 00:03:27.590 --> 00:03:30.110 tools are everywhere, but getting them to match 00:03:30.110 --> 00:03:33.610 your exact visual style takes hours. Right. Tweaking 00:03:33.610 --> 00:03:35.849 margins and fixing fonts for every workshop. 00:03:36.030 --> 00:03:38.960 Right. this custom app takes raw text content 00:03:38.960 --> 00:03:41.979 and instantly turns it into finished slides that 00:03:41.979 --> 00:03:44.780 perfectly match a unique specific style saving 00:03:44.780 --> 00:03:47.259 literally hours of busy work it's like having 00:03:47.259 --> 00:03:49.379 a graphic designer living in your computer yeah 00:03:49.379 --> 00:03:52.500 and the third example is custom email and calendar 00:03:52.500 --> 00:03:55.639 agents simple tools that filter draft replies 00:03:55.639 --> 00:03:58.680 and manage schedules your exact way saving hours 00:03:58.680 --> 00:04:02.120 every single week but the fourth example This 00:04:02.120 --> 00:04:04.979 is where things get truly wild. The manga generator. 00:04:05.379 --> 00:04:07.740 Oh man, this one is incredible. It's built for 00:04:07.740 --> 00:04:10.340 someone with zero drawing skills. You know, they 00:04:10.340 --> 00:04:12.860 have a story mapped out in their head, but absolutely 00:04:12.860 --> 00:04:15.120 cannot illustrate. Right, which used to be a 00:04:15.120 --> 00:04:17.819 complete dead end for creators. Exactly. It takes 00:04:17.819 --> 00:04:20.300 their story beats, along with some rough visual 00:04:20.300 --> 00:04:22.720 references, and it actually generates the art. 00:04:22.779 --> 00:04:25.160 It maintains character consistency across panels, 00:04:25.300 --> 00:04:28.040 and then binds it into a finished manga. Whoa! 00:04:28.680 --> 00:04:31.180 I mean, imagine scaling your creativity like 00:04:31.180 --> 00:04:33.420 that, generating a whole manga without knowing 00:04:33.420 --> 00:04:35.839 how to draw. It's fascinating, but it begs a 00:04:35.839 --> 00:04:39.740 bigger question. Do these AI tools replace human 00:04:39.740 --> 00:04:42.579 creativity or just change where we apply it? 00:04:42.720 --> 00:04:45.060 They absolutely just change where we apply it. 00:04:45.100 --> 00:04:47.819 The tool handles the technical execution, but 00:04:47.819 --> 00:04:49.959 you provide the inspiration and the emotional 00:04:49.959 --> 00:04:52.139 resonance. Right. They automate the technical 00:04:52.139 --> 00:04:54.800 bottlenecks, leaving the creative vision entirely 00:04:54.800 --> 00:04:57.759 to you. beat. That is the real magic of it. These 00:04:57.759 --> 00:05:00.040 examples sound incredible. But let's look at 00:05:00.040 --> 00:05:03.160 the four whys and the hidden catch. I mean, why 00:05:03.160 --> 00:05:04.920 should a non -engineer bother building these? 00:05:05.079 --> 00:05:07.839 Well, the first why is customization. The software 00:05:07.839 --> 00:05:10.180 shapes entirely around your personal workflow 00:05:10.180 --> 00:05:12.779 rather than forcing you to adapt. The second 00:05:12.779 --> 00:05:15.360 is cost. And the math here is pretty stark. You 00:05:15.360 --> 00:05:18.079 pay $10 to $20 in AI compute just once. Yeah, 00:05:18.120 --> 00:05:20.639 compare that to $50 to $200 every single month 00:05:20.639 --> 00:05:23.319 for generic sauce. Right. The third is privacy. 00:05:23.920 --> 00:05:26.139 Like we saw with the accounting app, local handling 00:05:26.139 --> 00:05:28.860 keeps sensitive data safe. It stays on your machine. 00:05:29.079 --> 00:05:32.259 Which is huge. And the fourth why is accessibility. 00:05:33.040 --> 00:05:35.480 You know, the AI does the heavy technical lifting. 00:05:35.980 --> 00:05:39.379 It writes the code and you just direct it. But 00:05:39.379 --> 00:05:41.519 we need to be realistic with you here. There 00:05:41.519 --> 00:05:44.939 is a catch to all of this. It's not literal magic. 00:05:45.180 --> 00:05:47.660 It takes real time and effort. Yeah, if you're 00:05:47.660 --> 00:05:50.500 tech savvy, maybe it takes a few hours. If you're 00:05:50.500 --> 00:05:53.850 a total beginner. Expect a few days, maybe up 00:05:53.850 --> 00:05:56.149 to a full week of tinkering. You are trading 00:05:56.149 --> 00:05:58.769 money for upfront effort. I'll be honest, I still 00:05:58.769 --> 00:06:00.829 wrestle with prompt drift myself when building 00:06:00.829 --> 00:06:03.009 from scratch. Oh yeah, prompt drift is rough. 00:06:03.329 --> 00:06:05.870 It really is. It's when you get 80 % done and 00:06:05.870 --> 00:06:08.389 the AI suddenly forgets the original rules. It 00:06:08.389 --> 00:06:10.370 just breaks code that worked five minutes ago. 00:06:10.529 --> 00:06:12.370 You have to actively guide it back on track. 00:06:12.470 --> 00:06:14.810 Plus there are small costs like an AI tool subscription 00:06:14.810 --> 00:06:17.410 or a few bucks for cloud hosting. Are we just 00:06:17.410 --> 00:06:19.829 trading a financial cost for a massive time sink? 00:06:20.170 --> 00:06:22.389 Not at all. If you look at the long game, that 00:06:22.389 --> 00:06:25.110 upfront investment of a weekend pays dividends 00:06:25.110 --> 00:06:28.329 for years. It saves you hours every month forever. 00:06:28.649 --> 00:06:32.009 You invest upfront effort once to escape paying 00:06:32.009 --> 00:06:35.889 and compromising every single month. Beat. That 00:06:35.889 --> 00:06:39.550 really is the bottom line, sponsor. So we understand 00:06:39.550 --> 00:06:41.990 the incredible trade -offs now, but how do we 00:06:41.990 --> 00:06:44.430 actually execute this over a weekend? I mean, 00:06:44.430 --> 00:06:46.509 I'm staring at a blank screen. The guide has 00:06:46.509 --> 00:06:49.269 a strict five -step framework. Step one is simply 00:06:49.269 --> 00:06:52.189 find the workflow. And it needs to fall into 00:06:52.189 --> 00:06:55.149 one of three buckets. First, a workflow you actively 00:06:55.149 --> 00:06:58.230 hate doing, like bookkeeping. Or second, something 00:06:58.230 --> 00:07:01.410 you neglect, like health tracking. Or third, 00:07:01.670 --> 00:07:04.069 something previously impossible, like the manga 00:07:04.069 --> 00:07:06.209 example where you just lacked the raw skill. 00:07:06.389 --> 00:07:08.870 Step two is critical. Map the workflow. Yeah, 00:07:08.910 --> 00:07:10.930 you cannot automate what you don't understand. 00:07:10.990 --> 00:07:12.930 You have to map out how the process actually 00:07:12.930 --> 00:07:16.029 works today. Step three is decide what to handle. 00:07:16.490 --> 00:07:18.730 This means you don't automate everything. Just 00:07:18.730 --> 00:07:21.209 find the specific bottleneck. Right. Like in 00:07:21.209 --> 00:07:23.370 the manga process, there were nine steps in total, 00:07:23.430 --> 00:07:26.410 but only four visual bottlenecks. So you keep 00:07:26.410 --> 00:07:29.029 the creative or enjoyable parts completely manual. 00:07:29.230 --> 00:07:32.290 Exactly. Which brings us to step four. Build 00:07:32.290 --> 00:07:35.589 with AI. You always start this with a PRD. Let's 00:07:35.589 --> 00:07:38.430 define that. A PRD is a product requirements 00:07:38.430 --> 00:07:41.610 document, put simply. A simple text description 00:07:41.610 --> 00:07:44.800 of what your software must do. Perfect. Once 00:07:44.800 --> 00:07:47.000 you have that, you pick your tools. Cloud Code 00:07:47.000 --> 00:07:49.980 is great for terminal or complex projects. Warp 00:07:49.980 --> 00:07:52.279 is better for collaborative or agentic environments. 00:07:52.639 --> 00:07:55.560 And Bolt is fantastic for browser -based builds 00:07:55.560 --> 00:07:58.560 if you are a total beginner. While you build, 00:07:58.720 --> 00:08:01.959 remember the mnemonic, tiny ferrets carry dangerous 00:08:01.959 --> 00:08:05.019 sound. I genuinely love that. Thinking, frameworks, 00:08:05.240 --> 00:08:08.259 checkpoints, debugging. and context it's silly 00:08:08.259 --> 00:08:10.360 but it works perfectly you know save constantly 00:08:10.360 --> 00:08:13.240 use frameworks and give the ai context step five 00:08:13.240 --> 00:08:16.240 is host it sensibly where does the app actually 00:08:16.240 --> 00:08:18.899 live cloud hosting is like renting a hotel room 00:08:18.899 --> 00:08:21.639 it's flexible and scales easily a vps or virtual 00:08:21.639 --> 00:08:24.220 private server is like renting an apartment a 00:08:24.220 --> 00:08:26.899 fixed fee with much more control and local hosting 00:08:26.899 --> 00:08:29.439 like a mac mini on your desk is owning a home 00:08:29.439 --> 00:08:32.299 full privacy and control but you maintain it 00:08:32.299 --> 00:08:35.460 yourself yeah the advice is to start simple and 00:08:35.460 --> 00:08:37.649 upgrade later Looking at this whole process, 00:08:37.750 --> 00:08:40.549 it's very methodical. Why is manually mapping 00:08:40.549 --> 00:08:43.690 the workflow so critical before you write a single 00:08:43.690 --> 00:08:46.110 prompt? Because if you skip mapping, you just 00:08:46.110 --> 00:08:48.610 end up building a very fast tool for the completely 00:08:48.610 --> 00:08:51.970 wrong job. You can't effectively automate a process 00:08:51.970 --> 00:08:54.330 if you don't actually understand how it works. 00:08:54.529 --> 00:08:57.169 Beat. Precisely. You aren't creating efficiency, 00:08:57.490 --> 00:09:00.610 you just create automated chaos. With the technical 00:09:00.610 --> 00:09:03.110 steps laid out, the final hurdle is changing 00:09:03.110 --> 00:09:05.519 our relationship with software. it's a massive 00:09:05.519 --> 00:09:08.179 mindset shift for a decade we've been conditioned 00:09:08.179 --> 00:09:11.440 to shop for tools right we browse product hunt 00:09:11.440 --> 00:09:14.700 take endless trials and adapt our work to a product 00:09:15.230 --> 00:09:17.629 We need to shift from shopping to building. You 00:09:17.629 --> 00:09:19.870 start asking what exact workflow you need today, 00:09:20.009 --> 00:09:22.289 and you just build it. To pick your first project, 00:09:22.490 --> 00:09:25.289 it must already annoy you. It needs clear steps. 00:09:25.429 --> 00:09:28.350 And above all, it must follow the rule of better 00:09:28.350 --> 00:09:30.509 is enough. Yeah, better is enough. It doesn't 00:09:30.509 --> 00:09:32.610 need a beautiful user interface. It just needs 00:09:32.610 --> 00:09:36.029 to improve your process by 20%. That first messy 00:09:36.029 --> 00:09:39.870 win is vital because progress compounds. The 00:09:39.870 --> 00:09:42.629 first app takes a weekend. And the next one takes 00:09:42.629 --> 00:09:45.149 hours instead of a weekend. You're building a 00:09:45.149 --> 00:09:47.789 new mental muscle. Why is aiming for better is 00:09:47.789 --> 00:09:50.669 enough the secret to surviving the first project? 00:09:50.929 --> 00:09:53.309 Because if you aim for pixel -perfect commercial 00:09:53.309 --> 00:09:55.549 quality, you will get completely overwhelmed 00:09:55.549 --> 00:09:58.669 by the details and quit immediately. Perfection 00:09:58.669 --> 00:10:01.789 causes paralysis. A simple working tool proves 00:10:01.789 --> 00:10:04.450 the concept and builds your momentum. Two sec 00:10:04.450 --> 00:10:06.470 silence. It really does. Let's pull all these 00:10:06.470 --> 00:10:08.309 threads together. The era of software creation 00:10:08.309 --> 00:10:10.929 being an exclusive walled garden for engineers 00:10:10.929 --> 00:10:14.090 is truly over. It's totally gone. We are rapidly 00:10:14.090 --> 00:10:16.730 moving away from renting generic digital spaces. 00:10:17.129 --> 00:10:19.909 We are finally building our own custom, hyper 00:10:19.909 --> 00:10:23.110 -specific tools for mere pennies. It's an incredible 00:10:23.110 --> 00:10:25.629 shift in power back to you. Look at your computer 00:10:25.629 --> 00:10:28.149 desktop right now. If you could build a tiny, 00:10:28.230 --> 00:10:30.899 invisible app tonight. that only does one hyper 00:10:30.899 --> 00:10:33.419 -specific task to save you 10 minutes every single 00:10:33.419 --> 00:10:35.860 morning, what would it be? Think about it. Map 00:10:35.860 --> 00:10:38.279 out your most annoying daily friction point today. 00:10:38.559 --> 00:10:40.980 Just grab a pen. You have the framework now. 00:10:41.379 --> 00:10:44.159 The tools are waiting. Stop renting your workflow. 00:10:44.279 --> 00:10:46.960 Start building it. Thanks for joining us on this 00:10:46.960 --> 00:10:48.299 deep dive of UETR music.