Domino Uroy @CommandingResou

@MuhammadUs12678 AI

Web designers charge $10k for this. I just did it in 20 minutes. Seedance 2.0 dropped this week and I had to test it immediately. Here's the workflow I used — image generation → looping background video → full deployed website, start to finish: Generate a visual in Seedance 2.0 Write a video prompt around it → get a cinematic looping background Feed that into Claude Code with a site brief Deploy to GitHub + Vercel The site looks like someone paid a designer real money. It took me one lunch break. No design experience. No production budget. Just two AI tools that actually talk to each other well. Want Full breakdown along with the prompts comment "Ai" below 👇

@MuhammadUs12678 Ran into the same problem

@MuhammadUs12678 😂🤣😅

@MuhammadUs12678 🤔

@MuhammadUs12678 🤔

Custom code feels more "real." Until you're 3 hours deep in Amadeus API auth debugging JSON at midnight. There's a smarter way to build this. Travel APIs are a different beast. Amadeus. SkyScanner. They both return these deeply nested JSON structures that look simple in the docs and are absolutely not simple when you're trying to map them into a readable chat response. Custom code means you're handling that mapping yourself. Every edge case. Every malformed response. Every auth token expiry at the worst possible moment. n8n means you can see exactly what went wrong and fix it in 30 seconds. That visual debugger alone is worth it for travel APIs specifically. Here's what I'd actually build: Pinecone vector store for the RAG layer. Feed it the agency's brochures, destination guides, package details. Everything the chatbot needs to answer questions about their specific offerings. Tavily Search node for real-time deal hunting. Live web search baked directly into the workflow. Client asks about flights to Bali next month Tavily goes and finds what's actually available right now. n8n AI Suite to wire it all together. One dashboard. API auth handled. Error logging visible. The whole thing debuggable without touching a single line of code. 90% n8n. Maybe 10% custom code for edge cases. The goal isn't to write impressive code. The goal is to ship something the travel agency can actually use. Those are very different targets.

Most agency owners are building products when they should be selling workflows. The difference is worth $3,000 per client. Before a year I fell into the trap. I have non-technical clients. Construction owners. Office managers. People who run real businesses but wouldn't know what an API is if you explained it twice. So I think I'll build them a white-label product. Custom frontend. Branded interface. Something that feels like software.4 Months of building. Then I try to sell it. And they look at me and say "so it's like ChatGPT?" Yes. Exactly like ChatGPT. Except I spent four months building it. But the thing that makes it for non-technical clients are: They don't want a login. They don't want a dashboard. They don't want to learn new software. They want their problem solved. A construction manager on a site doesn't want to open an app and type a prompt. He wants to text "is the cement delivery confirmed?" and get an answer in 10 seconds. That's a WhatsApp bot connected to Claude via n8n or Make.com. Took me two days to build. You charge $2,000 to $5,000 as a Digital Transformation fee. Then a monthly retainer to maintain it. No white-label product. No months of development. No trying to compete with Anthropic on UI. Just a workflow that solves a real problem for someone who will pay well to have it solved. Your value isn't the software. It never was . It's knowing which workflow to build and building it fast.

@MuhammadUs12678 I can relate to that

@MuhammadUs12678 At least now you know you can relax a bit

A founder described his AWS stack to me. $2,800/month. "Not sure where it's going." I ran it through a tool I built yesterday. Found $800/month in waste in about 8 seconds. The mistakes weren't complicated. RDS Multi-AZ on a staging environment — $140/month for redundancy nobody needs. NAT Gateway routing all S3 traffic — $0.045/GB when VPC endpoints are literally free. 12 Lambda functions with provisioned concurrency. Most got under 100 requests a day. Nobody had looked at the bill in 6 months. I'd been messing around with the Claude API the night before. Wanted to see if I could call it directly from a plain HTML file no backend, no framework, no deployment, nothing. Turns out you can. So I kept going. The idea was simple: Describe your stack in plain text. Claude scans it, returns structured JSON. The UI parses it into a full audit report. Every issue ranked by severity. Dollar estimates per problem. Actual fix commands not "consider optimizing." Real AWS CLI. A prioritized action plan sorted by ROI. One bug that got me: Claude was returning 2,800+ characters of JSON and I had max_tokens set to 1,000. Response was getting cut mid-string. JSON parser was dying. Took me longer than I'd like to admit to figure out what was happening. One line fix. max_tokens: 4000 The part that stuck with me: The expensive mistakes are almost never technical. It's always the boring stuff. Staging databases with production-level redundancy. Logging everything at 30-day retention by default. Services nobody turned off. The bill creeps up $50 at a time until it's $800/month in waste and nobody knows why. Built it in one afternoon. It's a single HTML file. I called it CostRadar. Demo video below. If you want the file, reply or DM happy to share it. And if you paste your stack in the replies, I'll run it through. Curious what it finds.

@MuhammadUs12678 Cool

The moment a client says "I have a lot of data" Stop. Breathe. Double your price. Then read this. Someone came to me needing a Telegram chatbot that could answer questions across 50GB of data. They had no idea what that meant technically. I did. 50GB is not a chatbot project. It's an enterprise search engine project with a chatbot on top. Here's what most people get wrong when they build RAG systems at this scale. Chunking. If your chunks are too small the LLM loses context and gives vague answers. Too large and you waste tokens and money on every single query. Getting this right for 50GB of mixed data is a project on its own. Then you need a Reranker Cohere or BGE work well sitting between your retrieval step and your LLM. Without it your chatbot pulls the 10 most similar chunks but not necessarily the 10 most relevant ones. Reranker fixes that. Then there's the embedding model. Not every model handles every data type well. You need to test before you commit to one at this scale. Now for the pricing conversation. Don't charge for hours on this. Charge for infrastructure and accuracy. $5,000 to $10,000 setup fee minimum. You're building a custom ETL pipeline and managing a massive vector database. That's not a weekend project. Add $500 per month for maintenance. That 50GB doesn't index itself. It needs hosting, monitoring and periodic re-indexing as the data changes. You're not building a chatbot. You're building their entire company's knowledge search engine. Price it like that.

@MuhammadUs12678 You hit the nail

The biggest lie in tech "You can't build a real business from your phone." Watch me. Most people think you need a MacBook, a dual monitor setup and a proper desk before you can start building AI automations. You don't. Here's what you actually need. First stop using your tools in a standard mobile browser. It's painful. Instead install the PWA versions of n8n and Make.com. They behave like native apps, load faster and handle the layouts properly on a small screen. For coding and script tweaking Claude Code and GitHub Codespaces both work surprisingly well in a mobile browser. Better than you'd expect. You can build, edit and deploy without touching a laptop. Now the secret weapon. A Bluetooth foldable keyboard. It fits in your pocket. Unfolds in seconds. Suddenly your Android is a workstation. Long logic-building sessions, client messages, workflow documentation all of it becomes actually comfortable. Your phone + PWA tools + Bluetooth keyboard. That's the whole setup. People waiting for the perfect desk setup before they start are just procrastinating with extra steps. The best time to start was when you had the idea. The second best time is right now with whatever you have.

@MuhammadUs12678 Sure do

@MuhammadUs12678 What went wrong?

@MuhammadUs12678 True talk

The exact stack I'd use to build this: 🔵 Image Generation → Fal.ai (Flux) 🔵 Structure Control → ControlNet 🔵 Technique → Image-to-Image 🔵 Frontend → React 🔵 Secret layer → Prompt Engineering Save this. You'll need it.

@MuhammadUs12678 What reasons 🤔