NBFire

New Burning Fire — Dance Crew Management Platform

What It Is

NBFire is a mobile-first PWA web platform for managing a dance crew. It replaces the fragmented communication across WhatsApp, spreadsheets, and shared calendars, centralizing everything in one place: member management, event calendar, shared payments, proposals with voting, and chat integrated with Telegram.

The frontend is an Angular 21 SPA with Tailwind CSS, the backend a .NET 9 API with Clean Architecture, real-time chat via SignalR with bidirectional Telegram bridge, Google OAuth authentication, all deployed on a Hetzner VPS with Nginx + PM2.

Architecture

graph TB
    subgraph Client["Client — Mobile-First PWA"]
        FE["Angular 21 + Tailwind\nSignals + Standalone"]
    end

    subgraph Server["Hetzner VPS"]
        NGINX["Nginx\nReverse Proxy + SSL"]

        subgraph Backend[".NET 9 — Clean Architecture"]
            API["REST API\nControllers"]
            HUB["SignalR Hub\nReal-time Chat"]
            SVC["Domain Services\nBusiness Logic"]
            EF["EF Core 9\nData Access"]
        end

        DB[("PostgreSQL 16")]
    end

    subgraph External["External Services"]
        GOOGLE["Google OAuth 2.0"]
        TELEGRAM["Telegram Bot API"]
        GEMINI["Gemini API\nTranslations"]
    end

    FE -->|"HTTPS"| NGINX
    NGINX -->|"/api/*"| API
    NGINX -->|"/hub/*"| HUB
    NGINX -->|"/*"| STATIC["Static Files"]

    API --> SVC
    HUB --> SVC
    SVC --> EF
    EF --> DB

    SVC --> TELEGRAM
    SVC --> GEMINI
    API --> GOOGLE
  

Development Time Estimate

How long would a super senior .NET/Angular developer take to build this from scratch, writing all code by hand?

Breakdown by Phase

Where The Time Goes

The chat system (Phase 5) alone accounts for ~23% of the project. Telegram Bot API is the riskiest variable: documentation isn't always clear, API limits, webhook debugging. The first production deploy always reveals issues invisible in local development.

Actual Development Timeline (with AI Coding Assistant)

The project was built by a junior designer directing an AI coding assistant as the AI developer. The AI wrote the code, the human decided what to build, tested on iPhone, and approved. Data extracted from Git history: 344 commits, from December 21, 2025 to February 1, 2026.

Day-by-Day Breakdown

Estimate vs Reality

Per-Phase Comparison

Key Observations

1. Boilerplate and repetitive code = 5-6x faster. AI excels at generating structured, pattern-based code: entities, DTOs, CRUD services, controllers, tests. The core build (Jan 5, 18:48→19:46) proves it: 7 phases in 58 minutes.
2. External API integration = only 1.9x faster. Telegram Bot API and SignalR require real-time debugging, testing with real data, and edge case handling that AI can't shortcut. This phase alone consumed ~33% of total development time.
3. The Jan 5-8 sprint was superhuman. 4 days, 138 commits, 70% of the project built. This volume of correct, tested code is physically impossible without AI assistance.
4. Human work was direction, not execution. Time was spent on: architectural decisions, prompting, manual testing on iPhone, visual debugging, and approval. Actual code-writing keystrokes were minimal.
5. POCs saved real time. Validating Telegram before building (Dec 28-30) prevented false starts and architectural rethinking during the January sprint.
6. Development was not continuous. 42 calendar days but only 16 with activity. The pace was “whenever there was time” — weekends, evenings, holidays. A full-time professional would have finished in ~2.5 consecutive weeks.