Vibe Coding · The Tier 3 Approach
AI-assisted software development ("vibe coding", Karpathy 2025) lets a practising clinician build clinical infrastructure without programming training. Combined with static client-side architecture, the result is privacy by architecture: no server, no database, structural impossibility of PHI exposure. DermTools (dermai.co.il), the reference implementation, includes 390+ clinical templates across 20+ specialty hubs, a five-dimensional differential-diagnosis wizard, drug-interaction and pregnancy-safety alerts, lab-monitoring schedules, patch-test reference, and patient education in seven languages.
Background & Methods
"Vibe coding" (Andrej Karpathy, 2 Feb 2025; Collins Word of the Year 2025) is AI-assisted software development through natural-language prompts. As of May 2026, seven peer-reviewed publications address vibe coding in medicine.
None addresses clinical documentation in any specialty. None addresses dermatology. Narrative review through May 2026 across PubMed, Google Scholar, and preprint servers.
Build Workflow · How DermTools Was Made
🩺
Clinician
Identifies clinical need · domain expert, not coder
→
💬
Natural-Language Prompt
Describes feature in plain English · no syntax
→
⚡
AI Generates Code
Claude Opus · React + static HTML
→
🔒
Static Deploy
No server, no DB · runs in browser
Net effect: domain expertise becomes clinical software; PHI never leaves the device because there is no server to send it to.
The Seven-Paper Vibe-Coding Literature · May 2026
| Author / Year | Domain |
| Chow & Ng (2025) | Clinician-as-creator paradigm |
| Pesce & Cheungpasitporn (2025) | Nephrology education |
| Huh (2025) | Taxonomy of vibe-coding tools |
| Moore & Tatonetti (2025) | Biomedical software paradigm |
| Abgrall et al. (2025) Lancet | Vibe coding to vibe caring · safety |
| Ng & Chow (2025) | Health-professions education |
| Meyer (2026) | Omics data analysis |
Conclusions
Vibe coding plus static client-side architecture lets a single dermatologist build clinical infrastructure that is current, safety-rule-aware, and architecturally privacy-preserving.
The shift is categorical: from clinician-as-AI-consumer to clinician-as-architect of clinical infrastructure. Replicable by any practising clinician with domain expertise; no programming background and no institutional permission required.
Added value: a clinician-built, continuously updated reference suite reflecting the linguistic and clinical reality of Israeli dermatology, at zero patient-data exposure.