previous
BasedAI Nexus is a real-time 3D visualization platform I designed and built as lead developer for BasedAI, a decentralized AI infrastructure blockchain. It turns raw on-chain activity into an explorable spatial environment — every node, transaction, and state change rendered as a living 3D scene rather than a wall of charts and terminal output.
The premise was simple: what if you could watch a blockchain work, instead of read it?
Lead developer. I wrote the design thesis, architected and built the frontend, built the Node.js backend for on-chain data aggregation, designed the data pipeline connecting blockchain events to real-time 3D rendering, coordinated 3D asset production with an external artist, and managed deployment across Vercel and Railway.
Traditional blockchain dashboards inherit their aesthetic from databases — grids, lists, columns of numbers. That works for developers querying specific values, but it loses the thing that makes decentralized networks worth caring about: the emergent behavior of hundreds of independent nodes operating in concert.
The BasedAI network runs on Brains — on-chain compute nodes that host AI agents. Each Brain has its own properties: agent count, emission rate, tempo, activation status, staked tokens. I needed a visual language that could encode all of those dimensions at once without overwhelming the viewer.
The solution was spatial. Each Brain is rendered as an icosahedron positioned along a golden ratio Fibonacci spiral. Geometry encodes identity. Color encodes activity — brightness scales with agent count, from dark teal at idle through full cyan at peak load. Active Brains spin. Inactive ones hold still in wireframe. The health of the entire network reads at a glance.
Double-clicking any node triggers a smooth camera transition into the Brain itself, opening a detail panel with its metadata — owner address, ticker, agent count, fractionalization status, linked imagery. Macro to micro, no context switch.
The frontend is Next.js on Vercel. The 3D scene runs on React Three Fiber and Three.js, with @react-three/drei handling camera controls, text rendering, and interaction primitives. State flows through Valtio’s proxy-based reactive store, which proved ideal for a scene where dozens of meshes need to respond to shared state without triggering full re-renders.
The chart system renders price, market cap, and 24-hour volume for multiple tokens directly inside the 3D scene as geometry. Price lines are Three.js Shapes that fill downward, giving the data physical presence in the environment instead of sitting on top of it as an overlay.
The backend is a Node.js service I wrote to ingest live data from the BasedAI chain. It connects through WebSocket RPC, queries smart contracts via Alchemy and ethers.js, and aggregates from CoinGecko, DEXScreener, and Etherscan. Brain registration events come from on-chain logs. Token supply, burned and locked balances, and real-time block height all converge through normalized API routes with caching and error recovery built in.
I worked directly with a 3D artist on the platform geometry, environmental elements, and lighting rigs. I defined the visual specifications, provided reference, and integrated the exported GLB assets into the scene — coordinating closely on scale, polygon count, and material properties to keep performance tight without sacrificing visual fidelity.
The application renders 280+ interactive 3D meshes with spring-based entry animations, real-time hover states, camera transitions, and concurrent data fetching while holding smooth frame rates. Memoized geometries, instanced materials, selective re-rendering through Valtio snapshots, and careful scene-graph management got it there.
Deployment runs through Vercel for the frontend and Railway for the backend, each with environment-based configuration and independent scaling.
BasedAI Nexus is what I think the next generation of data interfaces looks like — not flatter dashboards but richer spatial environments where complex systems explain themselves through form, motion, and interaction.