Build AI apps · from Vienna
You do not need to be a developer to build your first real app.
I am Bernhard. I build AI products that run in the real world every day. Here I show you the method behind them, step by step: let the agent build, verify the result, repeat.
Inside the AI Builder Guide you get BuilderBob: a prepared AI co-pilot that interviews you and builds with you. It guides; you click, paste, and verify. That way it stays your app, and you understand what is happening.
Already working with Cursor or Claude Code? Go to the builder resources
Vienna, Austria · 10+ years in IT · 100+ AI prototypes · 1st place BITGN Hackathon
Build in public
I do not just talk about agentic building. I build with it.
Not a marketing promise, but the method I use every day on real products that are live now. Not the fastest prototype, but one that holds up. That is the method I teach here.
BuchhaltGenie
LiveAI bookkeeping for sole proprietors in Austria. Photograph a receipt, AI posts it, FinanzOnline export included.
Next.js · Supabase · Stripe
WalkAITalkie
Mac App StorePush-to-talk dictation for Mac. Speak, release, get clean text. Fully on-device and 100% private.
Swift · WhisperKit · MLX
EventDrop
LiveScan a QR code and share event photos. All guests, one album, no app, no login. GDPR-compliant.
Next.js · Made in Austria
- 1st place, BitGN Hackathon Vienna. Solo, 3 hours on site
- 100+ AI prototypes · own agentic tools in daily use
You have the idea. The tools are ready. So why does it still break?
You opened Lovable or Bolt, described your idea in plain language, and ten minutes later a polished demo ran in the browser. Then you clicked publish, and something broke. Credentials were exposed. The logic did not hold. Three weeks and ten prompts later, everything kept looping.
That is not your fault. AI tools are very good at making something look like it works. They are much worse at telling you what will fall apart in production. Nobody shows you whether it holds.
There is a more reliable way to get there, one that still works when the next generation of tools arrives.
The path is not the fastest prototype. It is a build you can verify. Those quiet 80 percent, such as checks, clean secrets, and GDPR basics, make the difference. In the AI Builder Guide you do not do that alone: an AI co-pilot interviews you and walks through every round with you until the next step is clear.
The method loop
Five stations, one loop: collect context, plan, let the agent build, verify, then run the next round.
Verification is the step most people skip. It is the difference between a demo and an app you can actually judge.
Already know how to code?
Then your gap is the move from works on my machine -> works reliably: verification as a habit, clean secrets, reproducible deploys, and checked GDPR basics. That is exactly what the AI Builder Guide is built for. Go to the AI Builder Guide.
What to expect
What is inside the free handout
Part 0: Orientation
What agentic building is, and why checking the results is the decisive skill.
Part 1: Your toolkit
The tool table (Lovable, Bolt, Claude Code, and more), with an honest look at pricing.
The method loop on a real example
A landing page built from start to finish, with no terminal and no setup.
Where to go next
The next step, the Discord community invitation, and a preview of the deeper material.
For people who want the mechanics
The method loop has five steps: collect context -> plan -> let the agent build -> verify -> repeat. Planning is your mini spec, "let it build" is the task you give the AI, and verification is the step most people skip. You will not.
Verification in its simplest form means a check the AI can run itself, with evidence instead of promises. In the handout you do this on a real form: does a test submission really go through and arrive?
Inside the AI Builder Guide
The guide is not a PDF to file away. You get BuilderBob, a prepared AI co-pilot. It interviews you, writes the next prompts, and verifies with you. You still click, paste, and submit yourself. That way it stays your app, and you keep control.
Download the guide and the BuilderBob prompt.
Put both into a new Claude project.
Type "go" and let BuilderBob interview you.
Build together, with a verification step after every round.
The same co-pilot meets you at your level: browser-only with no terminal, or command line with full depth. And with the build log, you can come back days later exactly where you left off.
The handout is the free entry point and is enough for your first clean build. The full depth, including security, costs, deploys, and GDPR, lives in the AI Builder Guide (€29).
Get the free handout
A compact guide that helps you set up your first AI project cleanly, including one complete method loop.
- Delivered straight to your inbox, with no account and no download obstacle course.
- Double opt-in, GDPR-compliant, self-hosted fonts and no Google CDN.
- No tracking pixel, no analytics, no spam.
- Unsubscribe any time with one click.
Straight to your inbox
More in the privacy policy.
Available now · platform access
The AI Builder Guide: from zero to an app you can trust.
The guide takes you from the first prompt to a simple app you can inspect, understand, and publish. Step by step, with a method behind it, not just a recipe.
- The method. A complete walkthrough from the first plan to the first version you can verify.
- 10 use-case cards. Reusable cards for typical ideas that you can adapt to your next project.
- Reliability depth. Verification, security, costs, deploys, and a GDPR checklist. The part quick demos usually leave out.
Written for both ends: approachable enough for a first start, deep enough that you do not get stuck after the polished demo.
Price
Price: 29 euros
one-time purchase · no subscription · all 22 lessons in the browser right away
incl. VAT · voluntary 14-day money-back guarantee. One email is enough.
If the depth does not deliver what it promises here, you get your money back, no debate and no explanation required.
Not sure yet? Start with the free handout ↑
Coming soon
Multi-Agent Orchestration in Practice, the next course (€49)
From a single prompt to a coordinated agent pipeline. No date promised, but you can already take a look.
I do not sell anything here that I do not use every day myself.
I build real products with AI agents: BuchhaltGenie, WalkAITalkie and EventDrop are live. Add more than one hundred prototypes and my own agentic tools that I use daily. 10+ years in IT, now deeply focused on AI, plus AI literacy under EU AI Act Article 4 (ki-fit.at). Here I show the method behind it in public. Not a marketing promise, but the path I actually use.
A security analysis of quickly built apps shows how large the gap is: many critical vulnerabilities. Source: escape.tech . That is why a reusable method matters more than a fast demo.