# Greg Ruthenbeck PhD > Entrepreneur | Developer | Researcher. Specialist in VR surgical training simulations, haptics, and medical image processing. Author: Greg Ruthenbeck. Site: https://greg.ruthenbeck.io Currently AI Developer at Virt-X. 10 projects and 8 articles, written by the author and published as static pages: no backend, no database, no tracking. The most recent article is "A content format for hand-authored records", published 2026-11-09. Projects are research, software and teaching work; articles are notes and arguments. --- ## Projects - [Lexible](https://greg.ruthenbeck.io/projects/lexible/): A Scrabble-style word board game that runs in the browser — with friends, against the computer, or passing one phone around the table. - [MLAD.ai](https://greg.ruthenbeck.io/projects/mlad/): A curated library of working prompts and agentic patterns from professional sources, articles from practice, and a learning system in early access. - [ChatGPT Helper for Finding ATO BIC Codes](https://greg.ruthenbeck.io/projects/chatgpt-plugin-for-ato-bic-codes/): A custom GPT that helps an Australian business work out which Business Industry Code to put on its tax return. - [Enabling Massive Biomechanical Sims](https://greg.ruthenbeck.io/projects/biomechanical-simulation/): A finite-element software pipeline that let biomechanics researchers simulate the proximal femur at a resolution that had previously needed supercomputers. - [Does Haptic Feedback Improve Learning?](https://greg.ruthenbeck.io/projects/does-haptic-feedback-improve-learning/): A visuo-haptic versus visual study of spatial learning in a virtual reality nasendoscopy simulation, presented at Eurohaptics 2016. - [Drones for Teaching](https://greg.ruthenbeck.io/projects/drones-for-teaching/): A teaching innovation grant that put drones, from nano to micro, into first-year STEM classes at Flinders University. - [Haptic Simulation Collection](https://greg.ruthenbeck.io/projects/haptic-simulation-collection/): The work I presented at Eurohaptics 2016 in London: ten years of haptics, from sinus surgery to a gravity game that teaches the difference between mass and weight. - [Measuring Bone Healing](https://greg.ruthenbeck.io/projects/measuring-bone-healing/): A double-blind study of a new bone-regeneration substrate — and the GPU-accelerated image analysis I wrote to measure how well it worked. - [Haptic Illusion 3D Print](https://greg.ruthenbeck.io/projects/haptic-illusion-3d-print/): I 3D modelled and printed Masashi Nakatani’s fishbone tactile illusion so I could feel it for myself. - [Interactive Soft Tissue for Surgical Simulation](https://greg.ruthenbeck.io/projects/phd-thesis/): My PhD thesis: an ablatable soft-tissue simulation framework that makes efficient use of parallel hardware, and the medical training simulators built with it. ## Articles - [A content format for hand-authored records](https://greg.ruthenbeck.io/articles/a-content-format-for-hand-authored-records/): Why the records stopped being element lines: a metadata line, a blank line, a markdown body, and what that costs once code, diagrams and recordings live inside the prose. - [AI SEO in 2026: Structured Data as Identity Layer](https://greg.ruthenbeck.io/articles/ai-seo-in-2026-structured-data-as-identity-layer/): In 2026, structured data stopped being a rich-result optimisation and became an identity layer. What the MLAD prompt corpus and a production SEO spec reveal about the gap between being found and being cited. - [What 5,399 Prompts Reveal About Marketing AI Architecture](https://greg.ruthenbeck.io/articles/what-5399-prompts-reveal-about-marketing-ai-architecture/): Hightouch crossed $100M ARR by grounding AI in brand context before generating anything. A browsable corpus of 5,399 classified prompts shows the same architecture emerging independently across open-source marketing skills, coding workflows, and commercial products. - [Simple Skill, Full API: Replacing an MCP with 218 Lines](https://greg.ruthenbeck.io/articles/simple-skill-full-api-replacing-an-mcp-with-218-lines/): An MCP server silently dropped 9 Ghost API fields. We replaced it with a 218-line Claude Code skill that passes calls straight through the client library. - [Securing Your Vibe-Coded App](https://greg.ruthenbeck.io/articles/securing-your-vibe-coded-app/): Nearly half of AI-generated code contains security flaws. Five checks every vibe-coded app needs: server-side auth, secrets management, access control, XSS prevention, input validation. Platform fixes for MoltBot, Lovable, Codex, and Cursor. - [Skills Are More Context-Efficient Than MCP](https://greg.ruthenbeck.io/articles/skills-are-more-context-efficient-than-mcp/): MCP servers can consume 55,000 tokens of Claude Code's context window before you type. Skills achieve the same automation at 3x lower cost. Token comparison for Playwright and a framework for choosing the right approach. - [A Taxonomy for AI Prompts](https://greg.ruthenbeck.io/articles/understanding-ai-prompts-a-taxonomy-for-the-age-of-coding-assistants/): A classification system for AI coding prompts, built from 1,751 real prompts across 12 sources. Three types and four axes turn ad-hoc prompting into deliberate practice. Browse the full collection in the interactive Prompt Explorer. - [The Proposal-Driven Cycle](https://greg.ruthenbeck.io/articles/the-proposal-driven-cycle/): A lightweight workflow for AI coding that captures intent alongside code. Proposal, Plan, Execute, Record. Version Docs complement git history so you can audit decisions, seed new contexts, and scale documentation to match task complexity. ## Feeds - [RSS](https://greg.ruthenbeck.io/rss.xml): projects and articles, newest first. - [Sitemap](https://greg.ruthenbeck.io/sitemap.xml)