Luqman: The Dark Corpus of Unpublished Science

90% of scientific research never gets published. Failed drug trials, ruled-out compounds, dead-end ML experiments, and inconclusive materials tests disappear into lab notebooks — never shared, never compensated, endlessly repeated by other researchers around the world. The scientific community wastes an estimated $28 billion per year replicating experiments that someone else already ran. Luqman is a knowledge marketplace that fixes this. Researchers upload their unpublished negative results and earn USDC micropayments every time an AI agent cites their work. No grants, no journal gatekeepers — just automatic, per-chunk compensation the moment their insight prevents someone from repeating a mistake. The system has three tiers. Open papers are fully public. Validated papers carry verified researcher identity. Dark papers are the most powerful: researchers at pharmaceutical companies, government labs, or sensitive institutions can publish anonymously — their identity committed to a Zeko Mina L2 zero-knowledge proof so they are paid correctly without ever being identified. Under the hood, an AI agent built on Google ADK and Gemini 2.5 Flash queries the corpus, synthesises an answer, and renders a payment receipt showing each researcher paid, their paper, the amount in USDC, and a blockchain transaction link on Arc testnet. Circle Developer-Controlled Wallets handle the actual money movement — 85% to the researcher, 15% to the platform — settled on-chain with every query. Luqman is live on Railway with 17 seeded corpus documents across drug discovery, materials science, and machine learning. Every citation is a receipt. Every failed experiment is finally worth something.