01unique
Discovery Gaps
Maps populations, geographies, and validations that nobody covers. Ranked by opportunity score.
enginecontinuous
v2.0 · institutional preview launching q4 2026
Scientific intelligence
beyond search.
Yubetsu Researcher detects methodological weaknesses, emerging discoveries, hidden contradictions, and unexplored scientific opportunities — across the global research landscape, in real time.
Researcher home · live preview buildresearcher.yubetsu.com
500M+
Papers indexed
3B+
Citation edges
49k+
Journals & venues
240+
Funders tracked
live · research feed
gapOpen gap: pediatric dosing data missing for 14 of 22 new LNP formulations·contradictionConflicting result: tirzepatide cardiac remodeling — 3 trials disagree on EF Δ·methodologyMethodology flag: n=24 mouse study reports human-equivalent CI without scaling·gapReplication gap: room-temp superconductor — 0 of 37 preprints with full protocols·contradictionContradiction promoted: LNP hepatic clearance — biodistribution vs PK kinetics·methodologyUnderpowered: primary CRISPR off-target screen, post-hoc power = 0.41·gapMissing control: 6 of 11 microbiome → mood papers lack diet-matched cohort·contradictionDiscovery gap: no LMIC validation cohort for the 2024 sepsis ML triage model·methodologyOpen gap: pediatric dosing data missing for 14 of 22 new LNP formulations·gapConflicting result: tirzepatide cardiac remodeling — 3 trials disagree on EF Δ·contradictionMethodology flag: n=24 mouse study reports human-equivalent CI without scaling·methodologyReplication gap: room-temp superconductor — 0 of 37 preprints with full protocols·gapContradiction promoted: LNP hepatic clearance — biodistribution vs PK kinetics·contradictionUnderpowered: primary CRISPR off-target screen, post-hoc power = 0.41·methodologyMissing control: 6 of 11 microbiome → mood papers lack diet-matched cohort·gapDiscovery gap: no LMIC validation cohort for the 2024 sepsis ML triage model·
/01the problem
Tens of thousands of papers a day. Zero of them read by the same person.
Disciplines drift. Replications fail silently. Contradictions hide between subfields. The instrumentation researchers need does not exist — until Q4.
tens of k
new papers indexed on a busy day
38%
of high-impact claims lack independent replication
1 in 4
methodological weaknesses missed at peer review
6 mo
median lag between contradiction and retraction
/02product tour
Eight surfaces. One instrument.
Every module shares the same corpus, the same audit trail, the same keyboard. Built so a principal investigator can move from a daily digest to a contradiction debate without breaking flow.
Contradiction Engineresearcher · preview
/03core intelligence system
How the machine thinks.
Eight interlocking engines run continuously over the global corpus. Each is inspectable, queryable, and accountable to its evidence.
02rubric v3.2
Methodology Review
84-point rubric scoring controls, power, blinding, replication — at the per-paper level.
enginecontinuous
030–7 day lead
Novelty Detection
Surfaces frontier papers 0–7 days before citation velocity is visible to the field.
enginecontinuous
04always on
Research Agents
Standing watch lists that draft full briefings, not just notifications.
enginecontinuous
05~3B edges
Research Graph
Almost 500M papers, ~3B citation edges, recomputed every four hours.
enginecontinuous
06research preview
Contradiction Engine
Surfaces opposing claims across methodology, cohort, and pre-processing — an active research direction.
enginecontinuous
/04discovery gaps
What nobody has done yet — ranked.
Pre-computed continuously for newly published research — every freshly indexed paper is scanned for missing cohorts, absent controls, and untested populations. Open opportunities are scored on novelty, impact, feasibility, and funder fit.
/05contradiction engine
12 studies support it. 4 contradict it. Here's why.
Every disagreement is decomposed into the methodological choices that produced it — not flattened into a meta-analysis average.
/06my library
Your lab, your sublibraries, one audit trail.
Sublibraries, shared collections, custom tags, agent attribution. Every action is logged for grant reporting and replication.
/07request access · q4 2026
Reserve your institutional seats.
Researcher rolls out to a small cohort of institutions in Q4 2026. Seats are scoped per lab, with shared sublibraries and full audit trail on every agent run. Tell us about your group — we'll get back within five business days.
q4 2026
rollout
per lab
scoped seats
audit trail
every agent run
sub 5d
response time