27:01Build Great Products · Tutorial
Claude Code /goal Just Dropped and it Can Build Literally Anything
A 27-minute live head-to-head where two AI coding agents each build a full 62-task SaaS app from a spec doc in 32 minutes flat.
May 13thThe argument in one line.
Skipping structured product planning is why most AI-built apps fail — a structured vision-intake questionnaire run through an agent skill before touching Claude Code produces better architecture, a real brand direction, and a phase-by-phase roadmap that eliminates the need for detailed build prompts.
Read if. Skip if.
READ IF YOU ARE…
- You are building apps with Claude Code, Cursor, or Windsurf and keep ending up with something generic or structurally brittle after the second feature.
- You have tried dumping an idea directly into an AI coder and ended up with spaghetti architecture mid-project.
- You want a repeatable planning process you can run on any app idea in under an hour before committing to a build.
- You are a solo or part-time builder who cannot afford to rework a bad initial architecture mid-build.
SKIP IF…
- You already write detailed PRDs before building — this teaches the habit, not the advanced version of it.
- You are looking for a code-quality deep-dive; the actual build is barely touched here.
The full version, fast.
Most AI-built apps fail not because of bad code but because they start with no product context. PLAID fixes this with an eight-section structured interview (founder fit, target user, transformation, one-liner, business model, brand, tech stack) run through Claude Cowork, then auto-generating four markdown files: product vision, PRD, phased roadmap, and go-to-market plan. The host demos this live, planning a YouTube-to-social-posts app from scratch. The finished roadmap is handed directly to Claude Code, which reads the PRD references baked into each phase and builds without needing further prompting.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 01:35
01 · Introduction to agent skills
What agent skills are and why they change app quality; channel intro from Chris.
01:35 – 04:50
02 · What PLAID is
Product-Led AI Development: three-step process (vision intake, document generation, build-ready roadmap) and four output documents explained.
04:50 – 08:08
03 · Installing PLAID in Claude Cowork
Download ZIP from GitHub, upload via Cowork sidebar customize > skills, set working folder.
08:08 – 12:50
04 · Running the intake questionnaire
Activating the skill, entering the product idea (content repurposing app), answering the first sections on founder background and target audience.
12:50 – 17:18
05 · Transformation and value proposition
Defining the before/after transformation, the magic moment, and differentiators vs. tools like Repurpose.io and Cast Magic.
17:18 – 22:23
06 · Business model and goals
Choosing freemium + subscription ($19/$49/mo), setting 90-day and 6-month goals, discussing constraints and go-to-market channel.
22:23 – 26:11
07 · Brand direction and tech stack
Choosing brand personality (creator's secret weapon), visual mood (clean and bright, serif headings), tone (hype creator), then selecting Next.js, Convex, Clerk, Polar, and Claude Code.
26:11 – 31:50
08 · Reviewing generated documents
Walking through all four generated markdown files: product vision, PRD, roadmap (6 phases), and go-to-market plan.
31:50 – 34:38
09 · Handing the roadmap to Claude Code
Opening Claude Code in the same folder, using plan mode, instructing it to read the roadmap and create a plan for phase zero. Claude Code begins building.
34:38 – 35:18
10 · Recap and resources
plaid.build, skills.sh, community link at skool.com/aiapps.
Lines worth screenshotting.
- Planning brand, user persona, and business model before writing code means AI generates architecture that fits the product, not just the feature.
- The magic moment question is the most valuable part of any product intake: forcing yourself to articulate the first wow experience determines MVP scope better than a feature list does.
- Founder market fit — building a tool you personally use and understand — is the single strongest signal that an app idea has a working chance.
- A roadmap with PRD references baked into each phase removes the need to write detailed prompts every build session; you just say build phase one.
- One free use is the right pricing wedge for AI tools: it lets users experience the magic moment before money enters the picture.
- Agent skills are installable context bundles — they do not add new code capabilities, they add structured domain knowledge the AI uses to ask better questions.
- Generic AI slop in app design comes from skipping the brand direction step; even one deliberate choice (serif headings, hype-creator tone) separates your product from the default output.
- Spec-driven development is not about writing more — it is about writing the right things once, in a structured format AI already knows how to read.
- skills.sh gives pre-built skills for specific stacks (Convex, Supabase, etc.) that layer on top of a planning skill during the build phase.
- A PRD that is too long is a real failure mode — concise, AI-readable context beats comprehensive documentation for speed.
Plan before you prompt — structure beats speed.
WHAT TO LEARN
The gap between a generic AI demo and a shippable product is almost always a planning gap, not a code gap.
- Running a structured vision intake before opening a code editor forces you to answer questions — target user, magic moment, pricing model — that will break your build if you answer them mid-development instead.
- The magic moment question is the highest-leverage part of any product spec: defining the first wow experience tells you exactly what the MVP must include and what it should cut.
- Founder market fit is a build advantage, not just a marketing credential — personal experience with a problem means you can evaluate AI output for accuracy instead of accepting whatever it generates.
- Agent skills work because they package structured domain knowledge, not new code capabilities; what changes is the quality of questions the AI asks, which changes the quality of the context it builds from.
- A phased roadmap with PRD references embedded in each phase converts planning work into a reusable prompt — you stop writing build instructions and start approving plans.
- Brand direction is not a launch-week polish task; it shapes architecture decisions made in phase zero, because a hype-creator SaaS and a minimal professional tool require different component structures.
- A PRD that is too long is a real failure mode: AI coding tools perform better with dense, well-structured context than with exhaustive documentation that buries the important decisions.
- The first free use is the correct pricing wedge for AI tools with per-use API costs — it gets a user to the magic moment without creating unsustainable API burn on free accounts.
Terms worth knowing.
- Agent skill
- A ZIP-packaged bundle of instructions and context files that gives an AI agent a specialized workflow. Installing a skill adds structured domain knowledge the agent can activate on command.
- PLAID (Product-Led AI Development)
- A custom agent skill that guides a founder through an eight-section product vision intake and auto-generates four markdown documents used to plan and build an app with AI coding tools.
- PRD (Product Requirements Document)
- A technical specification covering architecture, stack choices, data models, user stories, and screen designs. Used as persistent context for AI coding tools.
- Magic moment
- The first product experience that makes a new user say wow. Identifying it early determines what the MVP must include and what it can cut.
- Claude Cowork
- A non-technical, task-oriented version of Claude Code available in the Claude desktop app. Same underlying agent capabilities, optimized for workflow and product tasks rather than raw coding.
- Founder market fit
- When the person building an app has direct personal experience with the problem they are solving, reducing the risk of building for a persona that does not exist.
- Convex
- A TypeScript-first serverless backend with real-time data syncing and zero boilerplate, popular for AI-assisted app development for its strong code-generation support.
Things they pointed at.
Lines you could clip.
01:33
“The whole point of building this skill was to translate my exact build process for building apps with AI coding tools into something that was reusable and could be used by anyone.”
24:08
“Design and branding is hugely important as a reason for why people pay for something. And this is something we don't wanna just leave up to AI because it's just gonna give us generic AI slop.”
31:54
“You don't have to write any detailed prompts for Claude Code. All you have to say is build phase one.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
00:00I wanna show you how to use agent skills and Claude Cowork so that you can stop building broken apps and start building apps that actually work. Now agent skills are like super powered AI workflows that you can add to existing AI tools like Claude Cowork and Claude Code to massively improve the capabilities.
00:18And so in this video, I'm gonna be showing you what skills to use, how I use these skills to plan my apps before actually building anything, and how this makes building 10 times faster and easier just by using this small set of agent skills that you can install in Claude Cowork. If you don't know me, my name is Chris.
00:37And for the last fifteen years, I've been designing apps and advising startups on product and design. And if you are building apps either to make money, to save time, or for your business, I have a community helping people do that and have helped hundreds of people build apps with AI. If you're interested, I've put the link in the description down below.
00:53Otherwise, let's get straight into Clawd Cowork and show you how to use these skills to help you build apps that actually work. So if you don't know what skills are for AI tools, basically, skills. If you go to agentskills.io, this gives you an overview of what skills are and how they work.
01:07But in essence, it's a simple open format for giving AI agents new capabilities and expertise. So think of this as training for your AI agents.
01:18And by AI agents, I'm specifically talking about Claude Cowork and Claude Code in this instance. You can install skills with any other AI agent out there including cursor, codex, and a bunch more tools. So now that I've broken down what skills are and how they work, I wanna show you the skills that we're gonna be working with today and mainly this one skill that I have developed called product led AI development.
01:41And product led AI development is essentially an AI skill that guides you through a structured series of questions to generate all of the context that your AI agent will need to build your app successfully so that it actually works.
01:55And the whole point of building this skill was to translate my exact build process for building apps with AI coding tools into something that was reusable and could be used by anyone. And essentially, this skill works by going through three steps. The first step is a vision intake, which basically is an interactive conversation with your AI agent that captures your product idea through eight structured sections.
02:19And having this context at a higher level is what is gonna make the difference between an app that just looks fire coded and is generic and something which actually has a bit of personality and character to it. The second step is document generation. So once you've gone through the intake questions with Plaid, it's then gonna generate all of the all of the documents that you need to actually start building your app.
02:40And these are four different documents. So this is the GitHub repository which is public which you can go onto today and install this agent skill yourself.
02:49I'm gonna leave a link in the description below for the plaid skill as well so that you can start using it. And essentially, it generates four documents for you. A product vision markdown document, which is the strategic foundation, the vision, the mission, the brand, user research, product strategy, and design direction for your app.
03:05The PRD, which is the technical specification and architecture for your app. The product road map, which is a phased build plan with check boxes against each task for sequential execution, and this is what we're gonna use to actually build our app. And then a go to market markdown file which essentially is the launch strategy and growth playbook for your app.
03:26And this is everything that you're gonna need to start building your app in phases to have the confidence that you're gonna build something that actually works and to give you the confidence that you have a brand and marketing plan set up around that to actually build something that is gonna work. The other thing that I'm gonna show you how to use is skills.sh, which is the open agent skills ecosystem from Vercel and this has thousands and thousands of agent skills on here.
03:53But what we're gonna look for is we're gonna look for skills that are related to the PRD that we're gonna be using to build our app. So first, let's get set up with Plaid in Claude Cowork. And if you haven't got Claude Cowork, you need to download the Claude desktop app.
04:06And then once you have the Claude desktop app, you will go to this middle tab here called Cowork, and that will allow you to access Claude Cowork. Now if you don't know what Claude Cowork is or you haven't used it before, essentially Claude Cowork is a nontechnical version of Claude Code.
04:20So it's an AI agent that uses the exact same framework as and capabilities of Claude Code, but but tailored specifically to nontechnical tasks. And you can set Claude Cowork up with a bunch of extensible features like connecting to skills, adding connectors to your favorite tools like Notion or Gmail or Google Docs, and a bunch of plugins that have been built by Anthropic as well for specific knowledge based work.
04:45Now we're not gonna use any of that stuff apart from skills today. And let me show you how to set up the Plaid skill in Claude Cowork. So the first thing that you wanna gonna wanna do is to click on this button here to view the Plaid skill on GitHub.
05:00Once you click that, you're gonna go to the Plaid GitHub repository, and you can click on this code option here and then click download ZIP. Once you have the ZIP file from GitHub, you're then gonna want to open the sidebar here in Claude Cowork and go to customize.
05:14And then once you've gone into customize, we want to go to the skills section here. And then when you go to skills, you can see I've got one skill that I've built here, is a go to market planner to reach your first 10 k MRR. You have to do is click this plus icon here, click upload a skill, and then we're gonna drop our skill zip file into Claude Cowork.
05:35I'm gonna drop this into Claude Cowork now, and that has uploaded the plaid skill and all the files associated with that underneath. You can see here it's called product led AI development, guides founders through defining their product via a structured conversation that generates four documents, use when someone says plan a product, help me build something, define my vision, generate a PRD, etcetera, basically.
05:55And this is the entire description of the Plaid skill. Now you don't necessarily need to read through all of this. I'm just gonna show you how to jump straight into a conversation here with and use Plaids to help you plan your app.
06:07And so now that we have a skill installed, let's click work in a folder because we wanna start working in the folder that we're gonna be using to build our app. So let's select a folder here to work in. So I'm selecting this content machine folder which is the name, just a placeholder name of the app that we're gonna be building.
06:24I'm gonna say always allow to allow that to access all the file the files in this folder and on my computer whenever I give it a folder to work in. And now I'm gonna just gonna say, help me plan an app using the plan skill. Now I've got this set to Opus 4.6 with extended thinking turned on, I would always recommend doing, especially if you have Claude Max Plan.
06:42Just set it to the best model with extended thinking always turned on. That's gonna give you the best results. And let's click let's go.
06:50So now Claude Cowork is gonna activate this skill for us and help us go through planning all of the things we need to build an app that is actually gonna work, have the chance of making money, or is actually gonna help us in our business to save time or make more money or automate any process that we have. The idea for this app that I wanna create today is basically gonna be a content machine.
07:12Like, it's gonna help me create a bunch of social content from my YouTube videos. So I want it to take a transcript from my video and I wanna turn that into a bunch of different types of content from maybe short form video, maybe written maybe let's just do written content first. So I'm gonna turn it into x posts and blog articles and LinkedIn posts, that sort of thing because that stuff is time consuming and I don't wanna necessarily just do that manually all the time.
07:38I kinda wanna automate a lot of that. And this is an idea of something that you could do in your business as well if you're thinking about what process could I automate with AI. So Plaid is gonna basically it's it's launched the Plaid integration integration skill here.
07:49So you can see it's running the skill. The user wants to plan an app using Plaid. Let me check the current state of the workspace, and then it's gonna read the Plaid intake guide and vision schema documentation.
07:59So it's reading the files within this skill, and now it's gonna guide us through this questionnaire that we set that I've set up inside of Plaid. So what do you want to build? I'm here to help you go from idea to full product spec, strategy, technical blueprint, build plan, and go to market playbook.
08:14Tell me what's on your mind, whether it's a fully formed concept or just a spark of an idea. So I'm gonna say here, I'm building an app to help content creators to turn their content, their YouTube videos, uh, long form content into social posts for multiple other channels.
08:34I only want it to help them output written content and I want that content to be generated following the best practice for each of those individual platforms and to allow the user to add examples of successful posts in their niche that it can be generated by as well.
08:53So that's our idea. We're gonna send it off here. And the whole point behind Plaid is not only to help you create a detailed spec to actually build your application, but it's also to help you think more deeply about the idea that you're building as well.
09:07So Plaid is gonna ask us questions as we go through to help to basically understand a bit more about us, why we're building this, and to help improve the idea across a couple of different aspects, including customer persona and a bunch of other stuff here. So let's continue going through Plaid.
09:22So this is a really compelling idea. Content repurposing is one of the things every creator knows they should do, but nobody does well. Let me walk you through some questions to help us flesh out into complete product vision.
09:32I've already got a good sense of the core concept, so I'll build what you've on what you've shared. For each question, I'll suggest some options based on what you've told me. So it's also gonna give you options and the pros and cons for each option as you build as well.
09:43So what's your name and professional background? What's your area of expertise? Are you a creator yourself, a developer, or marketer?
09:49I'm a content creator myself and have a channel with 44,000 subscribers and post two to four long form videos every week.
09:59And repurposing content takes up a massive amount of my time. And my name is Chris. Let's send that off and we'll continue going through Plaid.
10:07So it's already identified. We have founder market fit. This means that my expertise and what I'm already doing aligns with the type of app that I wanna create.
10:17So my audience is gonna be interested in and I already have experience of this problem to be able to solve it. So it's now asking, let's dig into the purpose a bit more.
10:27You told me the what. Now I want to nail down the who and the why. Who are you specifically building this for?
10:32You're clearly building it for yourself first, but thinking about the broader market, which of these feels right? So creators like you, mid sized YouTubers who post regularly and are drowning and repurposing work, they know they should be on Twitter, LinkedIn, threads, but they can't justify the time, Broader content creators or creator businesses.
10:48So I'm gonna say, um, let's go with number one. So mid sized YouTubers who post regular regularly, that's a tight, well defined audience that you understand deeply, which is great. So now let's talk about the problem and the transformation.
11:01You've already described the pain, but want to sharpen it. What does the before and after look like for your user?
11:06So now we're digging into what is the transformation that we're gonna offer for our product. So this would be in like a value proposition that would be on our landing page, for example. So it's reclaim reclaiming time spent repurposing content.
11:22So stop spending hours repurposing content for all channels would be a great value proposition. Reach unlock. So reach a wider reach a bigger audience and grow your channel faster using this app or quality without effort.
11:36So get more results basically from the content that you're posting. So I think that the one that we're gonna go what we're gonna go for here is let's go for time saved repurposing and grow faster as the transformations.
11:54Quick break in the video just to show you a little bit more about my community for people building apps with AI. And if you wanna build an app with AI that either makes money, saves you time, or unlocks your business, then this is the community for you. Inside the community, we've got over 250 people sharing their progress, sharing their apps as they build, and helping each other go through the same journey of learning how to build apps that actually work.
12:14You'll get access to my full twenty eight day app builder's blueprint course including foundations, how to find the right idea, how to design your app, how to build with leading AI tools, and how to launch and grow your application, as as well as a bunch of other stuff as well. You'll have the chance to win three months of Claude Max and get access to over $3,000,000 worth of AI deals so that you can start building at almost zero cost.
12:36What's more, we have weekly calls and monthly demo days and VIP workshops annual members, and you'll get a free intro call with me plus a free one to one if you're in the community for over three months. So if you're interested, just click the link down below in the description. Otherwise, let's get back into the video.
12:50So it's saying here that the value proposition is a combination. So save hours every week and actually show up on platforms that you've been neglecting. That's a powerful one two punch.
12:59So the transformation is before repurposing takes hours per video or just doesn't happen. And after, every video becomes platform ready social content in minutes, and I'm growing on channels I never had time for, which is great.
13:10Now let's talk about why you're the right person to build this. With 44,000 subs and two to four videos a week, you've clearly figured out the content game on YouTube. What makes you uniquely positioned here is that you know what good repurposing looks like because you've done it manually.
13:21And this, again, is the value of going through a system like product led AI development because this is gonna highlight all the things that actually make an app work and make money. And it's all in the details of things like, are you the right person to build this? Can you prove that you've done this before?
13:37And does that give trust to the app that you're building to enable people to want to sign up and pay for it? And this is really, really important. And this is why I built product led AI development as a skill to help people actually build apps that are gonna work, make money, and that will have a much much higher likelihood of succeeding.
13:56So let's say, yes, this works. So now we're gonna move on to the product. So let's give it a shape.
14:01What's the name? If you've got one in mind, throw it out. If not, here are three directions to riff off.
14:06Repurpose, proliferate, or clip cast. Do any of those land or do you have something?
14:11Let's just go with let's go with content engine for now. Here we've got our one liner and how it works. Here's a draft one liner.
14:18Content engine turns your YouTube videos into platform ready social posts in minutes. Perfect. Does that work or would you tweak it?
14:23And for how it works, walk me through how you imagine the core flow. So again, this is why using this skill is so great or this is why I kind of build this skill is to like help you to think about things that you're not thinking about when you just put an idea straight into Lovable or any other like AI coding tool, even Claude code.
14:41This is gonna get you thinking about what are the user journeys? What are the steps I'm going through here to actually build something that really, really works? So do we want them to be at a pet to paste and generate the YouTube URL or to feed and batch?
14:54So you connect your YouTube channel when a new video drops. Content engine automatically drafts social posts for your platforms or guided repurposing. You paste a video URL and content engine pulls out the key ideas and quotes for moments, then you pick which angles to turn into post.
15:06It generates drafts using a niche examples as a style guide. Now I wanna have this as almost as automated as possible. So I'm just gonna do the feed and batch option here because I think that's gonna be like the least effort for someone to actually get something usable out of this.
15:20The one liner works and let's use the feed and batch how it works approach. So now it's gonna turn this user journey into key capabilities for your app. So let's talk about how this basically, the user journey for this.
15:33So and the and the way that the tech is gonna be set up. So we need YouTube channel integration. We need multi platform post generation.
15:40We need platform specific best practices. We need a niche example library, and we need a review and edit queue to basically review them and then edit anything as we go. So I think that's perfect.
15:51But we also might wanna add in a way of scheduling here. I think for now, I'm actually just gonna leave this. So it's not about scheduling.
15:57It's just about generating the content. So let's say that's perfect and move on to the next section. So now it's gonna ask us what platform is this.
16:05It says I'm assuming web app, but I wanna confirm. Which of these best describes how it stands out from existing tools like repurpose dot I o, Opus clip, or Cast Magic? So is it about example driven quality?
16:16Which I think it is. Is it written content focus or is it both? I think it's both here.
16:20And for the magic moment, that first experience, which is super important when you're building apps and you wanna actually make money from them, you have to build in this magic moment. You have to build in a moment where people use your app and they go, wow.
16:32That just works. That's amazing. It this is doing something which I could never have done myself, either this fast or this good.
16:40So it says you connect your channel, your latest video gets picked up, and five minutes later, you're looking at a queue of ready post content ideas. That's really good. You had a few example of posts that crushed it in your niche.
16:49Hit generate, and the helper is so good you post it without editing. Or you check your dashboard Monday morning and realize content engine has already drafted a full week of social content for you from the last two videos. I think it's gotta be number one here because we wanna have someone get a result as soon as they get almost as soon as they get into the app.
17:06So let's go with number one for the magic moment. So let's go with both for the differentiators here and go with number one for the magic moment. Now that we've kind of now we've kind of outlined our products, we're moving into business antenna and business model.
17:20So what's the revenue model for this app? And this is, again, is really important when you build to know what you're gonna build going in and how you wanna structure this pricing strategy. And this is gonna guide you through choosing the right pricing strategy for your app as well.
17:33So you need to do freemium, a free tier with limited videos per month. Paid plan unlocks unlimited videos, more platforms, and niche library. Get creators hooked on the core value before asking for money or a subscription, maybe 19 a month for solo creators, 49 for more power creators, or usage based pay per video processed.
17:52I think usage based is like a nice idea, but I think it's over complicated for an MVP. And I like the idea of freemium of getting a some videos for free. But also, I don't wanna use I don't wanna have a free account with a lot of API usage for the AI generation.
18:08What I'm gonna go with is say, like, get your first video free and then pay it and then a subscription after that. So let's go with get your first video for free and then have a subscription model which is $19 a month for solo creators and has a limit of a certain number of videos per week or per month, and then a higher priced $49 a month for power creators with unlimited usage.
18:34So, yeah, we're letting them experience the magic moment for free with their first video and then tiered subscriptions then match how creators actually think about their output. So what's your ninety day goal here? So here are three frames.
18:45Do we wanna get a 100 creators signed up and 20 paying? 500 plus MRR and a strong signal on which platforms users generate for most, or 50 paying creators, 50 1,500 MRR, and three to five testimonials from creators who've seen growth, or launch from product on?
19:01I think we can go for let's go for this first one here. I think that's probably a good balance.
19:08Let's go for number one. And one thing to know is when you're going through this, you can because we're just talking to AI, you can also suggest your own directions here as well. You don't have to go with the options that Plaid is kind of generating for us here.
19:22And then let's choose a six month vision. So I'm gonna go for let's go let's be bold. Let's go with 500 paying creators, 12 k m r r partnerships with two to three YouTube educated channels, and content engineers to go to repurposing tool in the mid sized creator communities.
19:36Let's go with that. Let's go with number one. So what are your constraints here?
19:40Time, budget, or technical skills? What are you working with? So I'm gonna say I am a focused builder willing to scale back content out, but temporary to get twenty plus hours a week, or maybe creator.
19:51Content stays a priority, but you want to invest as little time as possible as building. You'd rather spend money than time. Open to hiring contractors who using no code where it makes sense.
19:58I think this, like, this might be a question that I'm taking that I wanna take out of here, but I'm actually just gonna go with a part time builder. And how are you planning to get your first users? I'm definitely gonna plan to do, like, my own audience, soft launch to my audience, build in public, make a video about the problem and how you're solving it.
20:17And your credibility as a creator who actually uses the tool is your biggest go to market asset. So let's definitely go with number one here. And what's great about using this is it has the history of our entire conversation.
20:28So it already knows who I am, what my background is, and it can apply that to all of the questions that it's asking. And they're ordered in a specific way to use the previous information in the questionnaire to then help suggest options as we develop, as we flesh out this idea over time going through this process.
20:46So what's the brand personality here? And now we're starting to get into the design aspect of this. So design and branding is hugely important as a reason for why people pay for something.
20:56And this is something that we don't wanna just leave up to AI because it's just gonna give us generic AI slop. We wanna be really focused and clear about the design and the brand direction for this app. So do we wanna be the efficient creator's sidekick or the smart content strategist or the creator's secret weapon?
21:12Energetic empowering makes you feel like you're leveling up, has a bit of personality and swagger without being cheesy, celebrates your wins. I I think we wanna be the creator's secret weapon. Number three sounds perfect.
21:24Now what's the visual mood? Three directions, dark and bold, clean and bright, or warm and punchy. I wanna go with clean and bright, I think is the right one here.
21:33Let's go with option two, but make it feel more like a premium writing experience. Use like a delicate serif font for the headings and a mods and a modern sans serif font for the copy.
21:47So I'm tweaking the option a little bit here because I kind of know the direction that I want this app to go from a design point of view. And the tone of voice here, either hype creator, energetic, uses creator lingo naturally, celebrates wins loudly, your last latest video just drops 12 posts ready across four platforms.
22:03Feel like a friend who's pumped for your success, confident coach, or sharp and minimal. I think we're gonna go for hype creator. Let's go for number one.
22:11So what do we want it to not feel like? We don't want it to feel like a generic AI tool. We don't wanna feel like let it feel like corporate or sterile or all of the above.
22:19Let's just make it not feel like any of the above. So now Vlad is gonna get into the tech stack.
22:26So since this is a web app and you're part time builder leaning on AI coding tools, let me walk you through the key decisions one at a time. So what are we gonna do from a front end point of view? What are you building the interface here with?
22:37Here's a comparison. So Next. Js, Remix, or Svelte, I think we're gonna go with Next.
22:41Js. Already know that here. So full featured web apps.
22:44AI coding is excellent. Massive ecosystem, versatile one click deployment, and why for you best AI tool support means faster building with limited dev time.
22:52I recommend Next. Js, so let's go with Next. Js.
22:56Let's go with Next. Js. So now Plaid is gonna give us tech recommendations for building our app based on what we're trying to achieve as well.
23:03Now our back end, where does your data and logic live? So we've got three options here, convex, superbase, and firebase. I'm gonna go with convex.
23:11I already know this. I love convex. I love using convex.
23:13It's great. And it's got the y for you as well. So for convex, it says TypeScript end to end zero boilerplate.
23:19It's basically low learning curve and it's excellent for AI coding support. So we're gonna go with Convex here. So now for authentication, we can use a couple of different options.
23:30Either we can use Clerk, auth dot j s, or ConvexAuth. There's some other options out there as well, but these are the ones that's recommended to this.
23:37So if you wanted to use, like, BetterAuth, for example, you could use that. It says, do we wanna do I'm gonna go with Clerk because, again, I love Clerk. I think it's super easy to set up and it's and it works really, really well.
23:47It's just, like, very friendly, like, as a tool to use. You could use convex auth, I generally found that, like, a little bit more difficult to set up. Let's go with Clerk for authentication.
23:58And then how do we wanna do payments here? So we could go Polar. We could go Stripe.
24:01We could go Lemon Squeezy. Now it says merchant of record here. No.
24:05You handle tax. That's incorrect. So Polar is actually a merchant of record.
24:10So this is not pulling through correctly. And also Stripe has a managed payments feature which basically allows them to act as a merchant of record as well. So it says setup is low, built for subscriptions, moderate, very flexible, but more config.
24:22So I'm gonna go with Polar because I've used Polar before and I think it's great for software subscriptions. So let's go with Polar for payments. So we've got our full stack here, Next.
24:32J s, Convex, Clerk, and Polar. And the last thing that's gonna ask us is what tool are we using to build this. Now I'm gonna be using ClaudeCode.
24:39It is gonna recommend ClaudeCode cursor. It recommends Windsurf, but I think we should put anti gravity in here as an option. I'm gonna be building with Claude code anyway, so let's say Claude code.
24:48And this is all important because all of these decisions are gonna inform the documents that can generate it at the end of PLAT. The doc and those documents being the PRD, the product requirement document, and the roadmap that's gonna guide Claude code through building this entire application in phases.
25:02So now we've done the full intake questionnaire. This is gonna save our vision for our product as and is going to generate all of the other documents that we need, our PRD, our go to market document, and our product road map as well.
25:16And the product road map is the really key one here because we're gonna use that with Claude code to start building our app. And using this method, using this skill is gonna mean that we don't actually have to write any any we don't have to write any detailed prompts for Claude code. All we're gonna be able to say is build phase one or create a plan to build phase one and go through it step by step to build each phase as we go.
25:39So it's saying now, let's your vision is captured and validated, ready to generate your product documents. This will create four files in the docs directory. So in this folder on our computer, it's gonna create these files inside of a docs file.
25:52And then when we open Claude code or any other coding tool, we can just say reference the product road map from the docs folder and create a plan for phase one. That's gonna allow us to build it. So let's just say, yes, let's go ahead and generate the documents.
26:05And now Claude Cowork using Plaid is gonna generate all of these four documents for us. So now Cowork has generated these documents using the Plaid skill here, the product vision, the PRD, the road map, and the go to market. Let's actually have a look at them to see what they look like.
26:20And let's have a look. And these are huge documents, by the way, as well. So the product vision here.
26:25So the vision statement, every content creator's ideas reach their full audience, not trapped on a single platform, but alive everywhere their people are. And we can go back and forth here with Cowork's, kind of edit this a little bit if we're not happy with it. The founder's why, the core values, ship it, then polish it.
26:41The output is the product. Learn from the real world, not from rules. Respect the creator's time.
26:45So this is expanding on everything that we've given Cowork here to fill out a document for our product vision, which basically matches like a business kind of strategy document. So strategic pillars, speed to value wins, platform native or nothing, the creator's voice comes first, and solo founder scale.
27:03Success looks like, and that's the information that we put in here. It's done user research here as well, so it's created a primary persona for us and a secondary customer and secondary customer personas here as well, and a list of jobs to be done by these people as well.
27:17So this is gonna give you a ton of information and insight as to why this idea, why the app idea that you're building might work. And actually, all of the customers that you need to be the types of customers that you need to be targeting, their pain points, current alternatives in the market, key assumptions that we're gonna validate here, a user journey map for our product as well.
27:35So from awareness to consideration, first use magic moment and habit formation, and then advocacy. The product strategy that we're embedding into our products that we're building here, market differentiation, and how we're gonna design the magic moment in the app.
27:50The MVP definition, so this outlines our rough definition of our MVP. So with the features that we've got, YouTube channel integration, transcript processing, etcetera, and anything that's explicitly out of scope here as well.
28:03And then the feature priorities that we've got. So the must haves, the should haves, and the could haves. So we're gonna focus on building everything in the must haves here.
28:11Core user flows and a bunch of other stuff, and it's super interesting to read through these documents that AI generates to give you an idea of your entire product strategy. We've got our PRD, which is our product requirement document.
28:25This is an AI friendly PRD. Again, this is super long. I actually wanna make these more concise, and these shouldn't be this long because I think they're too long at the moment.
28:34So I'm gonna tweak the plaid skill here to make this a little bit more concise. These are all markdown files, by the way, so this is perfect for feeding them to AI for context to actually build it. But, basically, again, you get a huge amount of stuff here.
28:46The product summary, the objective, market differentiation, magic moment, the technical architecture with our chosen stack as well. So it's breaking down how this is going to work.
28:56Um, the stack integration guide, so the order that we need to build this application in to make it work, the key integration passion patterns that we need to make sure that we're covering, how we're gonna set up our environment variables, the common things that break using this technology so that we understand exactly what we're looking for, the repository structure for this application, infrastructure and development, security considerations, and Claude code or whatever tool that you're using is gonna be able to take all of this into consideration when building this app.
29:25Here's the data model. I actually think we don't need to be super as specific about this stuff because Claude code is so good at figuring it out anyway. And then the relationships between between different channels, API specification here as well, and just a ton of stuff that's gonna help AI build our app.
29:43A bunch of user stories, which is maybe more of like a human first kind of thing. So, like, acceptance criteria for each user story. And once we kinda go through that, we have detailed breakdowns of every screen here as well within the app, all the screens that we need to create.
29:58The design system based on our kind of chosen design direction is not gonna be perfect, but it will be better than a lot of the AI stuff that you do get and a bunch of component definitions here as well as the tailwind configuration and also implementation. So there's so much stuff here that basically details everything we need to do to build this app in this PRD based on the text tag that we have chosen to build it with.
30:21And then the most important document here is the product's road map. So this is a slightly smaller document, but basically, it's all of the phases that we need to build this app. So we have phase zero foundation and setup.
30:33This is going to initialize the NextJS project, configure tailwind, base set up the convex back end files, basically, integrate Clerk authentication. It's gonna set up all of the foundations of our technology. Then phase one is the YouTube integration and transcript processing.
30:49Phase two is the AI post generation and review queue. Phase three is the niche example library and free video flow. And then phase four is payments and subscription gating.
30:58Page and phase five is polish and launch prep. And then we do have phase six, which is post launch iteration as well. So a bunch of other stuff that we can do here.
31:07And all that we need to do, now that we have all of these files, we have a go to market plan as well here, which is less important to kinda start out with, but basically gives you an idea of, like, how you would launch this. So, like, announcing the build, build in public part one, part two, recruit beta testers, beta prep, a launch week plan.
31:25So what we're gonna do on the launch week. This again is, like, maybe a little bit too in detail. We wanna I might tweak the way that some of this content is created for this skill, But but it gives you a rough idea of how you would take this app to market as well.
31:37So from this, what do we do next? So what you would do now is we're going to open up Claude code. We're gonna open up Claude code in the same folder.
31:46So I'm just gonna go to code here. And we're just gonna choose the same folder here. So let's choose a different folder.
31:51And I've got the content machine folder open. I'm gonna put it onto we're gonna put it onto plan mode first. And all you have to say here is read the product road map m d file in the docs folder and create a plan for phase zero.
32:04And then we just send that off. Trust the workspace. And this is gonna start Claude code building our products.
32:11And it's all done from the Claude desktop app here. You don't have to use Claude code in the terminal. You just can use Claude code right here working in this folder on your machine to get this app up and running.
32:21And you also have the preview which you can open for the app on the right hand side here as you're building as well. And because we're building a web app, this is gonna show our app as we're building it, which is perfect. And now Claude Code is gonna go through our documents, our roadmap document, and create a plan to build the first phase of this app.
32:40And all you have to do is just build each of the phases one after the other, and then you have your full app and has a much higher chance of working because you got all the detailed documentation. And it is a much better app idea because you've gone through that entire questionnaire with Plaid. Another thing that you can do here, if you do want to improve the build quality of the app as well and use other skills to help build this is in Claude Code, you wanna go to customize again and you can go to skills and we can add other skills in here to help Claude Code build our app.
33:12So if we go back to skills.sh, we can search for convex, for example.
33:17And I can go to this convex skill here. So this is a generic convex skill. We can either use this and ask Claude Code to install it using this this link here.
33:29Or we can go to the repository here. We can just click code and then go download zip and we can add that into Claude code just by adding a new scale the same as we did before with Claude code here. So let's go back to the Claude Code session that we had.
33:43And that is how to use agent skills including Plaid and including Convex and a bunch of other skills from skills.sh to actually build working apps and to massively improve your app ideas before you build them as well. You can see here from the plan that Claude Code has created is contact engine is a new product that turns YouTube videos into platform ready social posts.
34:03It's created a plan to build all of these tasks from phase zero here. And one of the great things that's built into the plan skill is actually the references back to the PRD. So in every part of the roadmap, it's gonna reference parts of the PRD to help it build it the right way as well.
34:18And we're gonna go through and build all of these phases here from these reference files that it's got at the bottom and we can just click approve clause plan and start coding to start building this app. So that is how you can go from building broken apps to building apps that actually work and have a massively improved chance of of actually making an impact from being successful either in your business to automate processes or as an actual product that you're gonna launch to paying customers.
34:43If you do wanna get the Plaid skill, all you have to do is go to plaid. Build and then either go to the GitHub repository and download it or use the install command in the terminal. And that will set you up with the skill.
34:53And then all you have to do is say help me plan a project using the plan skill. And if you wanna browse any of the other skills as well, can go to skills.sh and find any skills that is relevant to your tech stack that you're building with.
35:04And if you are building apps with AI or you wanna start building apps with AI that actually work, then you can check out my community over at score.com/aiapps. If you liked the video, don't forget to like and subscribe. Thank you for watching and I will see you next time.
The Hook
The bait, then the rug-pull.
Every AI-built app starts with the same mistake: dumping an idea into a coder before the product has a shape. This tutorial shows what happens when you run a structured vision intake first — and why the difference between a generic demo and a shippable product often comes down to a 30-minute questionnaire.
Named ideas worth stealing.
01:35list
PLAID Vision Intake (8 sections)
- Founder background and market fit
- Target user
- Before/after transformation
- Product name and one-liner
- Magic moment
- Business model and goals
- Brand direction and design
- Tech stack
Eight-section structured interview that captures everything an AI coder needs to build a coherent product. Outputs become the working context for every subsequent Claude Code session.
Steal forAny new app idea before opening a code editor
26:11list
Six-phase build roadmap
- Phase 0: Foundation and setup
- Phase 1: Core integration
- Phase 2: AI generation and review queue
- Phase 3: Example library and free flow
- Phase 4: Payments and subscription gating
- Phase 5: Polish and launch prep
The PLAID-generated roadmap structure for a typical SaaS web app. Each phase has checklist tasks with PRD references so Claude Code can build autonomously.
Steal forAny subscription SaaS app built with AI coding tools
How they asked for the click.
VERBAL ASK
34:50link
“If you are building apps with AI or you wanna start building apps with AI that actually work, then you can check out my community over at skool.com/aiapps.”
Clean verbal CTA at end, consistent with a mid-video community ad break at ~12:00 that pitches the same community more formally with pricing and benefits.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
