I work at the edge where AI becomes daily practice: personal operating systems, sovereign tooling, and coaching for people who need to stay capable while the tools around them accelerate. My public GitHub shows selected artifacts; much of the real work is private by design because it runs on personal data, local machines, and client context.
| Area | Public proof | What it demonstrates |
|---|---|---|
| Personal OS / sovereign AI | sterngold.nl | A public-facing view of a private AI operating system: local-first tools, memory, agents, and identity infrastructure. |
| AI tooling | PromptTranslator, anders-repo-template, notebooklm-skill | Reusable infrastructure for working with AI agents without losing repo discipline, source grounding, or verification habits. |
| Human-AI writing and coaching | WerkAnders, The Symbiotic Mind | The human side of AI adoption: agency, judgment, shame, capability, and collaboration design. |
| Project | What it is |
|---|---|
| Personal OS | A private, local-first operating system for my own life and work. The public site shows the experience; the core repo stays private because it touches personal context and local infrastructure. |
| Anderson | A sovereign AI engine for local RAG, methodology, and personal knowledge workflows. Private/local by design: your hardware, your data, your rules. |
| PromptTranslator | A product case study for an AI prompt interpreter: translates messy human intent into clearer machine-readable instructions. |
| anders-repo-template | A model-agnostic repo baseline so Claude Code, Codex, Cursor, Aider, and humans all read the same conventions. |
| notebooklm-skill | A Claude Code skill that lets an agent query NotebookLM notebooks for source-grounded answers from uploaded documents. |
| The Symbiotic Mind | Writing on AI x HI: not "how to use AI better," but how to design the relationship between human and machine intelligence. |
I prefer local-first systems, explicit repo conventions, verification over vibes, and public artifacts that explain the work without exposing private context.
AI is embedded in my daily build/write/verify loop. The signal is not "more prompts" or "more commits" as a scoreboard; it is that the tools are used deeply enough to produce inspected artifacts, reusable systems, and public writing.
Recent local operating footprint: thousands of contributions, frequent AI-assisted coding sessions, new specs, tests, screenshots, and deployment-ready public surfaces. I treat those numbers as evidence of cadence, not a claim that speed alone equals quality.