Headline Oracle

• 

  Landing page

• 

  Analysis page

Inspiration

In a world where news spreads fast but understanding lags behind, I wanted to build something that doesn't just report the headlines, but actually explains them, learns from history, and predicts what might happen next.

The idea struck me while thinking about Nigeria — how leadership has changed hands over the years, yet we often forget the patterns, promises, and consequences tied to those leaders. What if, when a political figure runs for office, the news didn’t just announce it — but brought up their past policies, historical context, and a forecast of what’s likely to happen if they win? "What if we had a system that remembered, reasoned, and helped us choose better?"

So I created Headline Oracle — a real-time, AI-powered news copilot that reads headlines, analyzes them with context and history, and helps people make sense of what’s unfolding — not just for today, but for what it might mean tomorrow.

What it does

Headline Oracle is a real-time, AI-powered platform that turns breaking news headlines into actionable insights.
It doesn't just show what’s happening — it helps users understand why it matters and what might happen next. For every headline, Headline Oracle:

• Fetches real-time reporting from trusted sources (CNN, BBC )
• Uses Perplexity’s Sonar Reasoning and Deep Research APIs to analyze background, actors, and implications
• Searches for historical parallels and outcomes of similar events
• Predicts 2–3 likely scenarios with confidence levels

🛠️ How I built it

• Framework: Next.js (App Router) for building both backend and frontend
• Styling: TailwindCSS/ShadCN for responsive UI components
• ORM/DB: Prisma + MySQL for managing headline data and predictions
• Data Fetching: TanStack Query for smooth, reactive frontend queries
• AI/Reasoning: Perplexity’s Sonar API
  • Used to generate context, historical insights, and predictions
  • Responses are parsed and formatted into structured “prediction cards”
• Headline Source: RSS feeds from trusted news outlets

⚠️ Challenges I ran into

• Parsing LLM output: Turning free-form text into structured prediction cards required custom logic to extract outcomes, confidence levels, and reasoning.
• Prompt tuning: I had to iterate on prompt design to get balanced, relevant, and non-repetitive insights from Sonar.

🏆 Accomplishments that we're proud of

• Built a fully functional fullstack app from scratch in less than 3 weeks
• Successfully chained real-time search → reasoning → structured predictions
• Created a clean, responsive UI that communicates complex geopolitical and social dynamics clearly
• Turned Sonar into a powerful research assistant for the news — not just a search engine

📚 What I learned

• How to prompt Sonar effectively to simulate a multi-step thought process
• The value of combining live information + historical memory to make real predictions
• How to build for both user clarity and technical robustness under tight deadlines
• That building products which enhance understanding — not just deliver content — is what makes AI truly useful

🚀 What's next for Headline Oracle

• Add a “Was it right?” tracker — compare predictions to real-world outcomes
• Launch a journalist/analyst dashboard mode for deep research workflows
• Explore integrations with Notion, Slack, and newsletter APIs for daily briefings
• Custom search for headline

Built With

• axios
• lucide-icons
• mysql
• next.js
• perplexity-api-(sonar-+-deep-research)
• prisma
• react
• react-markdown
• rss-feed
• shadcn-ui
• tailwind-css
• tanstack-query
• typescript
• vercel
• zod