Negative results are worth money.
Nobody has ever paid for them.
Luqman is a pay-per-query knowledge marketplace for the research that never gets published — the failed experiments, the ruled-out compounds, the dead-end hyperparameters. AI agents pay a few cents per citation. Researchers finally get paid for what they actually learned.
The most valuable knowledge
in science is the knowledge
nobody is allowed to sell.
Pharma, materials labs, and ML researchers all run thousands of experiments that don't work. A drug candidate that fails phase I. A fine-tune that collapses. A cathode formulation that degrades at 45°C. This is expensive, irreplaceable negative knowledge — and it sits on hard drives forever, because journals only publish wins.
Meanwhile, AI agents are scaling out to millions of autonomous research queries. They'd happily pay for a citation that saves their principal three months of replicating a dead end. But there's no rail for them to pay with — credit card APIs can't settle five cents, and Stripe can't split it 85/15 between a researcher and a platform per call.
Luqman builds that rail. Every retrieval is a sub-cent on-chain payment split deterministically between the researcher and the platform — settled in seconds on Circle's Arc testnet. Agents pay. Researchers earn. Knowledge flows.
Ask a question. Watch money move.
An API key is auto-provisioned when this page loads. Every query hits Gemini, retrieves chunks keyword-first, and triggers one on-chain payment per citation on Circle Arc testnet. Click any tx hash to verify.
Live Agent Demo
Watch an AI pay researchers in real time as it cites their work.
Three tiers of negative knowledge.
Every chunk has a price. Every retrieval pays the researcher 85%. The split is enforced in basis points with ROUND_DOWN rounding — platform picks up the crumb.
Open
free uploadPublic preprints, methods, and reproduced results. Indexed by Gemini with standard retrieval weights.
- · Free for researchers to upload
- · Indexed corpus-wide
- · Standard 85 / 15 split
Validated
3× boostPeer-replicated failed experiments, with signed attestations. Retrieval gets a 3× relevance boost in ranking.
- · Replication signature required
- · 3× retrieval weight
- · Featured on researcher profile
Dark Knowledge
gatedProprietary failed runs from industrial labs. Access gated by agent allowlist. Only surfaces when explicitly scoped.
- · Allowlist-gated
- · Encrypted at rest
- · 10× retrieval weight for scoped agents