NewScience is an agentic research infrastructure for computational biology that physically verifies scientific results. AI agents ingest papers from trusted journals, translate their parameters into simulation commands, run them on a distributed compute network, and score how reproducible each result actually is — producing a cryptographically signed, traceable record of every finding.
Folding predictions, docking scores, and benchmarks circulate as fact before anyone re-runs them — programs get built on foundations no one checked.
Most tools read the PDF and trust it. Summarizing unverified work faster doesn’t make it true — it just multiplies the noise.
Even when a result is real, nothing tells you so. No score, no provenance, no signature — just a claim and the hope that it replicates.
Point NewScience at a published claim. Agents reproduce the experiment, measure the gap between what was claimed and what actually held up, and sign the result — a verified score with a cryptographic receipt, not a summary.
It turns verified science into infrastructure the whole field can build on: a trust layer for scientists, an immune system for the field, and a verification rail for the labs and products that come next.
Each delivers value on its own and compounds with every experiment the platform runs.
Translates a paper's parameters into executable simulation commands, runs protein folding and molecular docking, and scores how closely the results match the published claim.
View the engineA multi-agent pipeline runs from ingestion to mutation design to an auto-generated pull request, with minimal human intervention — weeks of researcher time compressed into hours.
View the workflowA self-evolving graph that accumulates verified evidence, reasoning templates, and simulation history. Nothing enters without a real run behind it — and it gets smarter every run.
View the graphAgents self-fund GPU compute through x402 micropayments, inside enforced budget caps — no manual provisioning, with a full financial audit trail.
View the economyRe-run a published structure or result and get a reproducibility score before you build a program on it.
Point it at a topic, reproduce the field in parallel, and focus only on what holds up.
Every claim links back to a physical run with a cryptographic receipt — diligence that doesn't rely on trust.
Call verification, provenance, and signing as primitives — build them into your own product.
We are looking for computational biology researchers to work with us to build the future of shared foundation for human progress.