You describe your research once. The most relevant new papers arrive in your inbox with a note explaining why each one matters. Here’s what happens in between.
Each stage narrows the field with increasing precision across retrieval, ranking, and a fine-tuned scientific evaluator.
0+ papers indexed
Every major scientific source
Multi-angle retrieval
Hybrid search shaped to your profile
Top candidates surfaced
Merged and deduplicated
Deep relevance scoring
Proprietary ranker reads each paper
Fine-tuned evaluation
Judged on a four-dimension rubric
Your final three
Each paper explained
You describe your research once: the question you’re asking, the methods you use, the constraints you face. From that description, we build a structured profile of your work: entities, methods, mechanisms, context.
That profile becomes a set of multi-angle search queries that reach beyond your core domain into the adjacent fields where a technique or insight you haven’t seen yet might be waiting.
Your Research
“Investigating LRRK2 kinase inhibition as a therapeutic target for Parkinson’s disease using iPSC-derived dopaminergic neurons and mitochondrial functional assays...”
Your Interest Profile
We ingest every major scientific source daily. Citation graphs, multidisciplinary indexes, biomedical archives, and preprint servers are merged into a single unified corpus refreshed nightly.
250K+
Papers in the working corpus
14 days
Rolling window, refreshed nightly
Hybrid
Semantic + lexical retrieval
Unified & Deduplicated
The same paper is often indexed in multiple places. We resolve identities, merge records, and give papers found by multiple sources a stronger signal.
Search engines rank papers by their own generic algorithms. Our ranker reads your full research description alongside each paper’s title and abstract, scoring how precisely the two relate. We evaluated this approach head-to-head against the top commercial options across a hundred research projects and shipped the one that won most consistently.
Traditional Search
Keyword-matched, ranked by recency
Our Approach
Ranked by relevance to your research
Relevance alone isn’t enough. A paper can be on your topic but incremental, or use interesting methods but lack evidence.
Every finalist is read by our fine-tuned evaluation model, trained on hundreds of expert-graded examples to score papers on four dimensions that reflect how researchers actually assess the literature. Review articles are detected and filtered based on your preferences. The highest-scoring papers are selected for your digest.
For each paper you receive, a second pass writes a two-sentence note explaining exactly why it matters to your work. Not a generic summary, but a targeted connection between the paper’s findings and your own hypotheses, methods, or goals.
Scientific Rubric
LRRK2 Inhibition Restores Mitochondrial Function in iPSC-Derived Parkinson Neurons
Cell Stem Cell · Feb 2026
Relevance Note
Why this matters to you
“Demonstrated that LRRK2 kinase inhibition rescues mitochondrial membrane potential and complex I activity in iPSC-derived neurons. The dose-response data and mitochondrial assay approach directly inform the researcher’s experimental design.”
Every component you see above was chosen after hand-graded evaluation on real research projects across every major discipline.
100+
Research projects hand-graded
Independent human evaluation, pairwise comparison
600+
Expert-graded training examples
Used to fine-tune our scientific evaluator
3
Commercial rankers benchmarked
We shipped the one that won most often
Breadth of evaluation
Our evaluation set spans the real research landscape: high-temperature superconductors, base editing for sickle cell disease, nitrogen-vacancy diamond sensors, algorithmic fairness audits, urban gentrification studies, tonal text-to-speech, cancer immunology, protein language models, and dozens more. Wide enough that any system prone to niche-domain failure breaks visibly.
Pairwise
Blinded head-to-head
4-Dim
Scoring rubric
93%
Cheaper vs prior evaluator