Everything on Earth is connected. A solar flare triggers aurora, tectonic forces drive both earthquakes and eruptions, and ocean warmth fuels the storms that light up the sky. Most dashboards show these in isolation — Earth Pulse shows them together, watching for the moments when independent signals align.
A single-page 3D globe pulling 13 data layers from USGS, NASA, NOAA, Carbon Mapper, and other free public APIs. Every data point on the globe is clickable for details, and a cross-feed correlation engine detects when signals converge.
Most feeds are live. The Carbon Mapper plume layers are satellite detections published on a delay, so they show where large methane and CO₂ releases were observed roughly a month ago — not what is venting right now.
Planet Brain is a browser-local AI narrator powered by WebLLM. A small language model (SmolLM2–360M) loads directly into your GPU via WebGPU — no server, no API keys, no data leaves your device. On desktop, a larger model (Llama–3.2–3B) silently upgrades in the background for deeper analysis.
It reads all 13 data layers in real time, narrates cross-feed patterns every 45 seconds, and lets you ask questions about what's happening on the planet right now. Click a data point and Planet Brain adds contextual AI analysis alongside the raw data.
Built with Three.js and zero server-side dependencies — no framework, no build step. Data is fetched directly from public APIs in your browser, except the Carbon Mapper plumes, which are relayed through this site because that API does not permit direct browser access. The correlation engine watches for patterns: geomagnetic storms, seismic clusters, volcanic-seismic coupling, and solar flare impacts. Planet Brain runs entirely client-side via WebGPU — if your browser doesn't support it, the rule-based correlation engine provides narration instead.