What science
actually shows.
The scientific literature - structured into claims and evidence AI can reason over.
Why Claims is needed
$3T1
spent on R&D every year
300M
scientific papers —
humanity’s most valuable knowledge
humanity’s most valuable knowledge
0%
in a structured form machines can reason over
Frontier models hallucinate references
RAG systems parrot what authors say
Semantic similarity can't weigh evidence
Claims will enable AI to do precise scientific reasoning.
Built by researchers working across scientific replication, mechanism design and AI-for-science.
Philipp Koellinger
Professor of Economics, VU Amsterdam. Published in Nature and Science; helped resolve the replication crisis in his field.
Christian Roessler
Professor of Economics, CSU East Bay.
Game theory and mechanism design in leading economics journals; built the framework for optimal incentive mechanisms on Bittensor.
Game theory and mechanism design in leading economics journals; built the framework for optimal incentive mechanisms on Bittensor.
Ogban Ugot
AI Engineer.
Built SciWeave with Philipp —
an AI for science that beats frontier models on the Nature-published ScholarQA benchmark.
Built SciWeave with Philipp —
an AI for science that beats frontier models on the Nature-published ScholarQA benchmark.
From paper to answer.
1T+ words of prose
SNP rs9320913
explains
0.022% of variance in years of schooling
n = 101,069
n = 25,490
confidence interval
ask("genetic effects on education?")
rs9320913: +0.022% variance
confidence: HIGH
source: Science 340, 2013
confidence: HIGH
source: Science 340, 2013
The pipeline runs on every paper in the literature.