II
Software understanding
Active
VibeKB
Repository understanding
Turns an unfamiliar software repository into a living understanding site that documents what the software currently does, how components connect, where functionality lives, what is verified, and what remains uncertain.
The current focus: a system for understanding software before the next change.
- The core problem
- AI-assisted development makes software faster to create but harder to understand later.
- The system response
- VibeKB creates a repository-owned, source-grounded model before the next developer or agent changes anything.
Understanding flow
- Repository Source of truth
- Functionality model What the software does
- Warnings & files Evidence with reasons
- Explainable Diagrams Nodes and edges
- Understanding site Static, shareable
Key properties
- Functionality-first model
- Nodes explain what components are
- Edges explain why they connect
- Verified versus inferred relationships
- Files always include reasons
- No graph database required
- Static output suitable for GitHub Pages
Applied to
Specimen A
Arcana
VibeKB documenting a real PHP / MySQL / Gemini production application.
- Authentication and email verification paths
- Credits, plan gating, and Stripe assumptions
- Uploads, image generation, and queue workers
- Gallery behavior, watermarks, and render storage
- Deployment assumptions, warnings, and uncertainty
Specimen B
Stoppr
VibeKB applied to a separate mobile application and its surrounding architecture.
- Implemented versus partial functionality
- Paywall and subscription flows
- Superwall integration and placeholder configuration
- OAuth assumptions
- Widget and app-group concerns
- Verified and inferred architecture boundaries