Hub
Craft
Four practical guides for anyone using Live Vibe Coding to build something with AI — how to ask for code well, what “it works” leaves out, the vocabulary underneath the studio, and the accessibility checks a generated build still needs. Each one is sourced to a named authority — MDN, W3C/WAI, OWASP, vendor documentation — and says so.
Nothing here is invented to fill space. Where a claim, a number or a benchmark can’t be verified against a named source, it is left out rather than guessed — see each page’s own numbered Sources section.
Build Scope Estimator
Components rather than pages, a revision allowance, and the cut priced in weeks.
Open the ledger → Technique · Sortable ledgerHow to Prompt for Code
What actually changes the output when you ask AI to write code — give it a contract instead of prose, iterate in small diffs, ask for tests — and the failure modes each technique closes off, hallucinated packages included.
Open the ledger → Interactive · Readiness checklistFrom Prompt to Production
The honest gap between code that generates and code that ships — input validation, error handling, auth, rate limits, tests, a real data layer. Score your own build from your own answers; no false reassurance at 100%.
Run the checklist → Glossary · Sortable & filterableWeb App Anatomy
Frontend, backend, API, state, database, deploy, domain, DNS — plain-English meanings for the words you’ll hit while building, why each one matters, and the misunderstanding each one invites.
Open the glossary → Interactive · Contrast calculatorAccessibility Basics
Semantic HTML, keyboard navigation, contrast, alt text, focus states and reduced motion — cited to WCAG 2.2 and WAI, with a working contrast-ratio calculator using the real formula, not a rule of thumb.
Check your contrast →