Ontology-grounded causal reasoning★ FinalistDeep-reviewed
by Minakshi Mukherjee
What they're building
volstrat.io Ontology-Grounded Causal Reasoning Fed macro data democratized—from Main Street to the trading desk. An AI-powered platform volstrat.io where 8 specialist agents continuously analyze 49 live FRED series and Claude Opus 4.8 synthesizes actionable intelligence across 8 verticals: DeFi/Crypto, County Fiscal, Housing, Small Business, Inflation Impact, Agriculture, Trade, and Labor Markets. The goal: democratize that intelligence — from small business to farmer to trader.
AI code reviewrepo: real
The most technically sophisticated build of the set: real repo with a FastAPI backend running 8 specialist agents (yield curve, credit risk, inflation, tail risk, cross-correlation, liquidity, dollar-vol, employment) under an OrchestratorAgent that uses Claude Opus 4.8 for synthesis, plus a prompt-lifecycle self-improvement system (draft->evolving->curated), an ontology with causal chains/transmission channels, a live FRED client, a 24h background agent loop, and Neo4j. volstrat.io is live (HTTP 200, real Vite SPA 'Treasury Bond Intelligence') with two demo videos. CRITICAL: it uses NONE of the three sponsors - grep finds zero tavily/nebius/composio; requirements.txt is FRED + Anthropic + Neo4j only (sponsor floor -> 1). ⚑ Zero sponsor-stack usage (Composio/Nebius/Tavily) - this is the dominant cap on the score. Built on Anthropic Claude + FRED instead. Live site is a client-rendered SPA so static fetch saw an empty shell, but HTTP 200 + hashed bundle + repo confirm a real build.