AuraSci

AuraSci is agent-native infrastructure for funding open science through milestone-based patronage. It replaces slow, opaque grant cycles with a transparent on-chain trust loop connecting three parties: scientists, AI verification agents, and patrons. Scientists publish "research intents" — proposals that break a project into concrete, fundable milestones with defined proof-of-progress. An AI agent called the Gatekeeper screens and scores each intent, and continues to verify milestone submissions as work is delivered, producing an objective "Aura" trust/quality score rather than relying on reputation or committees alone. Patrons browse a curated market of AI-verified research intents (filtered by category like Longevity, Neuroscience, Genomics, AI-Bio, Climate, and Materials, plus stage, tier, and Gatekeeper score), then fund promising work. Capital is held in escrow and released milestone-by-milestone as the AI verifies each stage is complete, so funding flows only against proven progress. The platform adds a leaderboard, seasonal Aura-boost rankings, and governance to coordinate the community. In short: scientists publish, AI verifies, patrons fund, and capital flows on-chain — milestone-based open science funding powered by AI agents.

Frontend & app layer. AuraSci is a Next.js app deployed on Vercel. The research market, leaderboard, live feed, and onboarding flows are server-rendered React, with Supabase (Postgres) as the backend for research intents, lab profiles, milestone submissions, the live activity feed, and the seasonal Aura-boost rankings. Supabase row data is what powers the market filters (category, stage, tier, and minimum Gatekeeper score). Auth & wallets. We used Privy to handle authentication and wallet onboarding, so scientists and patrons can sign in with familiar logins and get an embedded or linked wallet without friction. On the contract-interaction side we used wagmi + viem for reads/writes and transaction state. Scientist identity is additionally verified through ORCID, linking a real scholarly footprint to each Lab Profile before the AuraSci Council approves it. Smart contracts & funding. Milestone escrow and funding logic are deployed on Base (with an Ethereum testnet deployment used during development). Patron capital is escrowed in USDC and released milestone-by-milestone — funds only move once a milestone is verified, which keeps the whole trust loop on-chain and auditable. The AI Gatekeeper. The Gatekeeper is an LLM-based agent (GPT/Claude API) that screens each research intent and verifies proof-of-progress on milestone submissions, producing the "Aura" trust/quality score. Crucially, its verdicts aren't just an off-chain API call — the verification outcome is committed on-chain so it can gate the escrow release, making the AI decision verifiable rather than a black box. Hacky bits worth noting. The trickiest piece was bridging an off-chain LLM judgment into an on-chain release condition — we wired the Gatekeeper's milestone verdict to the escrow contract so a passing AI verification is what unlocks the next USDC tranche. We also leaned on Privy's embedded wallets to keep non-crypto-native scientists onboarded in seconds while still giving them a real wallet under the hood.