VibE17: MeetupOS & Community Memory
Personal AI Project
Vanilla ES Modules and Vite deliver sub-100ms First Contentful Paint with zero framework dependency churn.
Architectural boundary isolating private contact channels from consented public profiles and anonymous cohort tags.
Sub-second structured JSON state extraction and high-affinity peer matchmaking with conversational pub icebreakers.
Legacy Ticketing (Eventbrite / Luma)
Rigid 10-field web forms, mandatory account creation, intrusive tracking pixels, and impersonal confirmation emails. Organizers are marooned in static spreadsheets with zero context on who is attending or why they should connect, leading to 50% drop-off rates and awkward pub small talk.
VibE17 Conversational MeetupOS
Conversational Progressive Disclosure: Morris extracts contact preferences and craft curiosities through effortless dialogue and dynamic response chips.
Live Firestore Kanban: Organizers manage a reactive 4-stage pipeline with one-click broadcast dispatch and human-in-the-loop overrides.
Engineered Serendipity: Gemini 3.8 Flash analyzes guest curiosities to generate bespoke peer matchmaking cards before anyone arrives.
“Community networking breaks down when it is left entirely to corporate forms and noisy rooms. VibE17 treats community building as high-craft software engineering—pairing conversational warmth with strict mathematical privacy.”
What AI Skills This Required
- Conversational Dialogue State Tracking: Engineered AI concierge Morris using Gemini 3.8 Flash (
@google/genaiSDK) with structured JSON schemas to track dialogue state, extract preferred contact channels, and capture craft curiosities in under 45 seconds. - Reasoning & High-Affinity Matchmaking: Built a structured prompt pipeline that analyzes confirmed guests' active projects and curiosities, surfacing 3–6 high-value peer introductions with reasoned rationales and witty pub icebreakers.
- Prompt Defense & Anti-Scraping Guards: Multi-window IP rate limiting, heuristic prompt injection screening, disposable email blocking, and randomized typing cadence to deter automated bots.
- Full-Stack Serverless Architecture: Node.js 20 Cloud Functions v2 in London (
europe-west2), live Firestore snapshot listeners driving a zero-reload admin Kanban, and Resend transactional emails with.icscalendar generation.
How Considered is the Idea: 3-Tier Privacy & Community Architecture
Grassroots community trust requires absolute privacy discipline. VibE17 enforces a strict 3-tier architectural separation between private data and public social proof:
| Privacy Tier | Data Elements | Scope & Security Boundary |
|---|---|---|
| Tier 1: Strictly Private | Email address, WhatsApp/phone, LinkedIn | Stored strictly in encrypted Firestore fields. Never exposed in client bundles, public APIs, peer blurbs, or prompt contexts. |
| Tier 2: Consented Public | Name, pronouns, portfolio/GitHub link | Exposed in pre-event peer previews only if the attendee explicitly toggles one-tap opt-in consent. |
| Tier 3: Anonymous Aggregable | Craft category, technical skills, active curiosity tags | Aggregated into real-time cohort teasers ("3 brand designers, 2 WebGPU makers, 1 LLM researcher") so guests can gauge the evening's vibe. |
Performance & Production Results
- Lighthouse Score: 99+ across Performance, Accessibility, and Best Practices.
- Client Footprint: ~25KB total client JavaScript (compressed vanilla ES modules).
- First Contentful Paint: < 100ms on mobile 4G connections.
- Operating Cost: 100% serverless execution fitting entirely within monthly free tiers ($0/month idle cost).
- Event Show-Up Rate: 85%+ attendance rate across 100+ creative salon guests.