AI ConsultingAboutVideosBookFAQContact
DE|EN
Book Free Consultation

What this is about

The videos in this hub are about concrete builds: a sticker app, a website, small offline tools and the path from design to code. These are not perfect product launches. They are working examples.

The pattern is the same every time: keep the scope small, plan first, build second, check after. That is what separates useful AI work from vibe coding that only looks good in a screenshot.

How I would read the workflow

  • 1. Start with a real problemThe small HTML app and StickerLog show why concrete friction beats an abstract app idea.
  • 2. Do not skip designClaude Design is not a replacement for product thinking, but it helps make an idea visible before code is written.
  • 3. Build locally and versionedThe Wix rebuild shows why files, GitHub and Netlify often work better for AI-assisted websites than a closed builder.
  • 4. Plan for a second lookClaude Code vs Codex and StickerLog show why review, debugging and a second opinion are quality control, not bureaucracy.

Relevant videos

FAQ

For small, clearly scoped tools, yes. You still need judgment, testing and a tight scope. The more critical the system becomes, the more technical review matters.

Yes. Claude Code can build real apps, and you can start very small: in the video a missing Qwixx score sheet becomes a single HTML file that works like an app and runs offline in the browser. StickerLog, a web app for sticker collectors, shows a bigger build. The bigger and more public the app, the more a clean plan, testing and security matter, especially once other people use it.

For a simple page, a site builder is often enough. You get more control with an AI coding agent like Claude Code or Codex: you describe what the site should do, the agent writes the files, you check the result in the browser and improve it step by step. The site can then go live through GitHub and Netlify, for example. That is how this website was made, built in two days. You do not need to be a developer, but you need judgment and careful checks: do links and forms work, does everything look right on a phone?

There is no blanket winner, both are strong coding agents. Combining them is more useful: for StickerLog, Claude Code planned and built, while Codex reviewed the plan first as a second opinion and found gaps in the data model, input handling and edge cases before implementation started. In the Voicebox debugging test, the two also followed different clues.

All videos

Want to build cleaner AI workflows?

I help freelancers, consultants, coaches and small teams turn AI tools into real workflows. Practical, inspectable and matched to how you actually work.

Book a free strategy call