Oddly News

oddly.news is an audio-first platform that transforms prediction market data into accessible, on-demand intelligence briefings. Think "Bloomberg Radio meets AI research assistant" - users can listen to market analysis anytime, anywhere, without needing to parse complex financial data themselves.

Core Functionality: The platform operates through specialized AI agents, each focused on different market sectors (politics, crypto, economy, etc.).

When a user requests a briefing, the system:

• Pulls real-time data from prediction markets (Polymarket, Kalshi) to get current odds, trading volumes, and market movements
• Searches for context using Perplexity AI to understand the factors driving market sentiment
• Generates analysis via OpenAI to synthesize market positions with explanatory context
• Creates audio briefings using ElevenLabs text-to-speech, delivering 60-90 second professional radio-style reports

Technical Architecture:

• Backend: Node.js/Express API with Supabase for data storage and audio file hosting
• AI Pipeline: Integrated APIs for market data, news context, script generation, and voice synthesis
• Frontend: React-based dashboard with music player-style interface for audio playback
• Smart Caching: 30-minute briefing cache to avoid unnecessary regeneration -ENS Integration: Each agent has its own L2 subdomain (politics.oddly.news, crypto.oddly.news)

Unique Value Proposition: Unlike traditional financial news that reports after events happen, prediction markets aggregate information from traders with money at stake, often signaling trends before mainstream coverage. oddly.news makes this intelligence accessible through audio format, perfect for busy professionals who want research-grade market analysis during commutes or workouts. Monetization Strategy:

The platform essentially democratizes access to prediction market intelligence, transforming raw betting odds into actionable insights delivered in an accessible audio format.

oddly.news: Topic agents that turn prediction market intelligence into 60–90s audio briefings.

Core loop: Polymarket data → OpenAI insights + search queries → Perplexity context → OpenAI radio script → ElevenLabs MP3 → Supabase Storage + DB.

ENS Subdomain Management: Each AI agent gets its own ENS subdomain (politics.oddly.news, crypto.oddly.news) stored in the database, creating a decentralized identity system for different market expertise areas.

Backend Foundation:

Node.js/Express API server handling routing and business logic Supabase for PostgreSQL database and file storage (stores agent configs, briefing history, audio files) Environment Management via dotenv for API keys and configuration

Working: Each agent turns Polymarket data (by tag) into a 60–90s audio briefing by flowing events through OpenAI (to extract insights and generate “why” search queries), Perplexity (to fetch concise current context), OpenAI again (to draft a radio‑ready script), and ElevenLabs (to synthesize an MP3), then storing script/audio in Supabase (Postgres + Storage) and serving via a simple API with a 30‑minute cache. We provision agents with deterministic HD wallets (bip39/hdkey) and register <topic>.oddly.eth subdomains on Base using ethers.js and the L2 Registrar, giving each agent a verifiable on‑chain identity. Partner tech accelerated everything: OpenAI for synthesis, Perplexity for retrieval, ElevenLabs for natural TTS, Supabase for managed DB + public files, and ENS on Base for cheap, fast identity