← Back to the graph
NEURA

An operating system for intelligence.

Neura is an infinite, living knowledge graph — part encyclopedia, part AI collaborator, rendered as a 3D universe you fly through rather than a page you scroll. This page walks through what it can do, what it's built with, and what went into building it.

Everything inside the graph

Every feature below is live in the app right now — nothing here is a roadmap item.

Living 3D knowledge graph

A full Three.js universe — nebula, dust, aurora and starfield layers, shader-driven neuron materials, and a custom camera rig — built on react-three-fiber rather than a flat node/edge diagram.

🌐

Wikipedia-powered expansion

Click any idea to pull in the real Wikipedia article graph around it, turning encyclopedic topics into connected, explorable neurons.

AI graph expansion

When Wikipedia has nothing for a topic — a niche idea, a made-up concept, plain jargon — an LLM invents a handful of related concepts and weighted connections on demand.

💬

AI chat panel

A conversational panel that's aware of whichever neuron is currently selected, so questions can stay grounded in the part of the graph you're looking at.

📝

AI topic explanation

One click fills in a thin or empty neuron's description — useful for AI-generated concepts that only shipped with a title.

Expand topic with AI

Generates 3-4 claims about a topic and has a second model rebut each one directly, dropping every claim/rebuttal pair onto the graph as a mirrored, color-coded, linked cluster.

🎙

Live Debate Mode

A full-screen takeover where two AI agents argue a topic live, turn by turn, each one reading the transcript so far and rebutting the other's weakest point.

Command palette

⌘K opens a keyboard-first way to jump to a neuron, pull in a new Wikipedia topic, or ask the AI to expand one — without touching the graph directly.

🔍

Search, context menu & picking

A live search bar, a right-click context menu on any neuron, and a dedicated picking layer so hovering and selecting nodes in 3D space feels precise, not fiddly.

🎥

Around the web & videos

Contextual live web and video results for the selected neuron, pulled from Tavily's search API rather than a static link-out.

By the numbers

Counted straight from the repository — not rounded up for effect.

Time to build
27
React components
73
TypeScript files
~7,200
Lines of code
Solo build — one developer, start to finish.

What it's built with

Next.js 16 (App Router)React 19TypeScriptThree.js@react-three/fiber@react-three/postprocessingTailwind CSS 4Groq (Llama 3.3 70B + Llama 3.1 8B)Wikipedia REST APITavily Search API