I Was Tracking Wars With 47 Tabs — So I Built This in 72 Hours

15,000 sessions. 50+ countries. ₹0 marketing. 20 days.

The Problem: 47 Browser Tabs to Track One Conflict

Three weeks ago, I was trying to follow the Russia-Ukraine conflict.

My browser:

• Reuters (main news)


• Al Jazeera (different perspective)


• BBC (verification)


• AP News (wire service)


• Twitter/X (real-time updates)


• GDELT (event data)


• FlightRadar24 (airspace)


• TradingView (oil prices)


• ...and 39 more tabs

This was insane.

Every time I wanted an update, I'd spend 30 minutes refreshing tabs, cross-referencing sources, trying to separate signal from noise.

I thought: There has to be a better way.

Spoiler: There wasn't. So I built it.

The Solution: Warfront

The idea: One dashboard. All conflicts. Real-time. AI-powered.

The timeline: 72 hours (Feb 28 night → March 3 morning)

The location: My hostel room at IIIT Naya Raipur

The result: warfront.live

What It Does

1. Real-Time News Aggregation

Pulls from 12+ sources every 5 minutes:

• Reuters, Al Jazeera, AP News, BBC


• The Guardian, CNN, TASS, Xinhua


• Regional sources for local conflicts

2. AI-Powered Analysis

Smart LLM router with 5-provider fallback:

Cerebras → Mistral → Fireworks → Groq → Gemini

If one provider is down, automatically switches to the next. 99.9% uptime.

The system actually uses multiple providers beyond these five,

ensuring maximum reliability.

Analyzes:

• Threat levels (Low/Medium/High/Critical)


• Key developments


• Impact assessment


• Trends and patterns

3. Live Conflict Map

Interactive map showing 50+ active conflict zones:

• Russia-Ukraine


• Israel-Gaza


• Sudan civil war


• Myanmar conflict


• Yemen


• ...and 45+ more

4. Market Impact Tracking

• Oil prices (conflicts → energy markets)


• Defense stocks (Lockheed, Raytheon, etc.)


• Currency fluctuations


• Safe-haven assets (gold)

5. Airspace Monitoring

Track restricted airspace, military movements, unusual flight patterns

6. Multi-Language Support

26 languages including:

• English, Arabic, Russian, Ukrainian, Hebrew


• Hindi, Mandarin, French, Spanish


• Local languages for regional conflicts

The Tech Stack

Frontend

• 
  Vanilla JavaScript (~2,400 lines)
  
  
  
  
  • Why not React? Needed it FAST. No build process, no dependencies.
  
  
  
  


• 
  Leaflet.js for the map


• 
  Deployed on: Netlify

Backend

• Node.js + Express


• 
  MongoDB for data storage


• 
  13 Automated Cron Jobs:
  
  
  
  
  • News scraping (every 5 min)
  
  
  • GDELT events (every 15 min)
  
  
  • Market data (every 30 min)
  
  
  • Airspace updates (hourly)
  
  
  • Database cleanup (daily)
  
  
  
  

AI Layer

Multi-LLM orchestration with intelligent fallback:

const providers = [

  { name: 'Cerebras', speed: 'fastest' },

  { name: 'Mistral', speed: 'fast' },

  { name: 'Fireworks', speed: 'medium' },

  { name: 'Groq', speed: 'fast' },

  { name: 'Gemini', speed: 'reliable' }

];



async function getAIAnalysis(conflictData) {

  for (const provider of providers) {

    try {

      const response = await callProvider(provider, conflictData);

      if (response.success) return response;

    } catch (err) {

      console.log(`${provider.name} failed, trying next...`);

      continue;

    }

  }

  throw new Error('All providers failed');

}

This ensures the AI analysis ALWAYS works, even if a provider goes down.

Infrastructure

• 
  Hosting: DigitalOcean (migrated from Railway)


• 
  CDN: Cloudflare


• 
  SSL: Let's Encrypt (via Netlify)


• 
  Monitoring: Custom logging + alerts

The 72-Hour Build Timeline

Hour 0-8: Planning & Setup

• Drew architecture on paper


• Set up MongoDB Atlas


• Created Express server skeleton


• Basic HTML structure

Hour 8-24: Core Features

• News scraping pipeline


• GDELT API integration


• Map implementation (Leaflet)


• Basic UI

Hour 24-48: AI Integration

• LLM provider research


• API integrations (5 providers)


• Fallback logic


• Prompt engineering for conflict analysis

Hour 48-72: Polish & Deploy

• Multi-language support


• Market data integration


• Airspace tracking


• SEO optimization


• Deploy to Railway (later migrated to DigitalOcean)

Hour 72: Launch

• Posted on LinkedIn


• Shared with friends


• Crossed fingers 🤞

What Happened Next

Week 1: The Surprise

• 
  1,000 sessions in first 3 days


• Someone in Ukraine started using it daily


• A user in UAE opened it 150+ times


• ChatGPT started recommending it organically

Week 2: Going Viral

• 5,000 sessions


• Ranked #1 on Google for "warfront"


• 500+ ChatGPT referrals


• Users from 30+ countries

Week 3: Validation

• 15,000+ sessions


• 50+ countries


• 1,300+ ChatGPT referrals


• Won Boardroom Battle at E-Summit 2026


• ₹10,000 prize + validation from judges

The Stats (20 Days Since Launch)

📊 15,000+ sessions

🌍 50+ countries

🤖 1,300+ ChatGPT referrals (organic!)

🔍 #1 Google ranking for "warfront"

⚡ 99.9% uptime

💰 ₹0 spent on marketing

Most meaningful stat:

Someone in an actual war zone (Ukraine) uses this daily.

148 visits in 20 days. Another user in UAE: 92 visits.

Both still actively using it.

That validation is worth more than any metric

Challenges & Solutions

Challenge 1: LLM API Costs

Problem: Running AI analysis on every conflict update = expensive

Solution:

• Smart caching (don't re-analyze identical news)


• Use free tiers strategically


• Fallback to cheaper providers when possible

Challenge 2: Real-Time Data at Scale

Problem: 13 cron jobs hitting APIs every 5-15 minutes

Solution:

• Staggered job timing (avoid all hitting at once)


• Efficient database indexing


• Cloudflare CDN for static assets

Challenge 3: Misinformation

Problem: Conflict news is full of propaganda and fake news

Solution:

• Cross-reference 12+ sources


• AI flags conflicting reports


• Prioritize established news agencies


• User can see original sources

Challenge 4: Server Costs

Problem: Railway getting expensive as traffic grew

Solution:

• Migrated to DigitalOcean


• Used GitHub Student Pack credits


• Optimized database queries


• Implemented caching

What I Learned

1. Ship Fast, Iterate Faster

72 hours from idea to launch. Perfect is the enemy of done.

2. Users Don't Care About Your Tech Stack

They care if it solves their problem. My vanilla JS frontend works perfectly.

3. Real Users > Imaginary Users

That Ukraine user taught me more than any focus group could.

4. SEO Matters

Ranked #1 in 2 weeks because I:

• Used semantic HTML


• Server-side rendering for key pages


• Fast load times


• Quality content

The ranking happened faster than I expected - within the

first two weeks of launch.

5. AI Is a Tool, Not Magic

The LLM fallback system is more important than which model I use.

What's Next

Short-term (Q2 2026):

• Premium memberships ($10/month)


• Custom alerts


• Historical data access


• API for developers

Mid-term (Q3 2026):

• Fine-tuned Llama 3.3 for conflict prediction(Datasets are currently generating)


• Mobile apps (iOS, Android)


• Government pilots in India

Long-term (2027+):

• Sovereign AI models for governments


• B2B API for news organizations


• International expansion

Still refining the exact roadmap based on user feedback

and traction.

Try It Yourself

👉 warfront.live

Free. Ad-free. Always will be.

If you find it useful, consider supporting via Buy Me a Coffee.

Questions?

Drop them below! Happy to answer anything about:

• The technical architecture


• LLM integration strategies


• Scaling challenges


• Conflict data sources

Or just say hi on:

• 
  Twitter: @shriabhinav09
  


• 
  LinkedIn: Abhinav Shrivastava
  


• 
  Email: [email protected]
  

Final Thought

Three weeks ago, I was frustrated with 47 browser tabs.

Today, someone in Ukraine and UAE uses what I built to stay informed during a war.

That's why we build.

Built with: JavaScript, Node.js, MongoDB, 5(maybe more) LLM providers, sleepless nights and a lot of chai ☕

and ya forget to tell that thanks to github and anthropic you know why thank them🙃.

#buildinpublic #ai #webdev #startup