Multiple bank accounts, credit cards and savings products — but no overview. Expenses tracked manually in spreadsheets, budgets existed only in my head, and the monthly reconciliation was always a puzzle from exported CSVs.

Existing apps were either too complex, too privacy-unfriendly, or too reliant on bank integrations that barely work reliably in Germany.

A personal Finance OS: accounts, expenses, budgets, wealth and goals — all in one system. No bank integration, no third-party cloud syncs. Full data control.

Above that: an AI layer — not a chatbot, but a silent CFO that recognises patterns, creates forecasts and assists with decisions.

Claude is the thinking partner for architecture decisions, data models and complex UI logic. Lovable translates the ideas into working React code. Supabase handles backend and authentication.

Dashboard, account management, transaction entry and budget overview are complete. Visuals show sample data only — no real financial data.

Next: recurring entries, annual planning and the AI analysis layer.

Sensitive financial data demands clear decisions from the start: What data is stored? Where? How is it protected? These questions can't be optimised in after the fact.

AI-assisted development accelerates implementation considerably — but content control remains with me. Claude makes suggestions. I decide.

Training plans in one app, habits in a second, body data in a third — and actual progress visible nowhere. Data points were everywhere, just not together.

Existing fitness apps are designed for gamification, not personal insights. I wanted a system that listens to me — not one that drives me.

No generic tracker. A system tailored to my training style, my habits and my goals. Built for one person: me.

Training, sleep, nutrition, habits — all in one view. No social pressure, no leaderboards. Just data and patterns.

Same as Finance OS: Claude helps with data model and logic, Lovable builds the interface. What works, stays. What doesn't convince, gets reworked — often in the same session window.

The tracker runs and I use it daily. Visuals show demo values only — no real health or body data.

Next steps: weekly review, pattern analysis and export function.

The key question isn't "Can I build this?" — but "Does this tracking change my behaviour?" Some data points only create data maintenance, not insights.

Operating a self-built system is also overhead. I'm learning what is truly useful — and what just sounds interesting.

A PDF resume can describe what I can do — but not show how I work. I wanted a presence that demonstrates AI competency, not just names it.

The website itself should be the proof: built with AI, iteratively developed, publicly documented.

Three artefacts that belong together: the website as interactive presence, the PDF as the classic application document, the GitHub repository as transparent development history.

The process is visible. Commits, iterations, decisions — all publicly traceable. That's not a side effect, that's the concept.

No framework, no build system — deliberately. Vanilla HTML, CSS and JavaScript. Claude Code is the conversation partner: ideas get discussed, code gets generated, I review and decide.

The website is live and actively being developed. You're looking at the current state right now — including this page.

New sections, refinements, corrections — all happening publicly in the repository.

Transparency about how you work is a choice. Keeping commits and iterations public means making mistakes visible — and corrections too.

This website doesn't describe AI competency. It's a result of it.