Synapse

Free. No account needed.

Health questions, answered from real studies

Ask in plain English. Every result is ranked by our custom Research Evidence Rank (RER): study design, sample size, and relevance, scored 0–100.

Synapse summarizes published research for learning. It is not medical advice. Talk to a doctor about your own health decisions.

How it works

Research Evidence Rank (RER)

I built a custom ranking algorithm for Synapse, not a generic LLM score. Every paper gets a transparent 0–100 evidence score from PubMed metadata and abstract text alone. It runs in under a millisecond on every search, sorts results strongest-evidence-first, and powers the overview verdict.

Study design

34%

Meta-analysis and RCTs rank above case reports, modeled on GRADE / Oxford CEBM levels.

Sample size

16%

Parsed from abstracts (n=…); larger trials score higher on a log curve.

Query relevance

16%

BM25-lite match between your question and title + abstract.

Recency

12%

Newer publications score higher; missing dates get a neutral default.

Rigor flags

14%

Blinding, placebo control, preregistration, minus conflicts or tiny samples.

Journal tier

8%

High-impact venues (Lancet, NEJM, JAMA, Cochrane, etc.) get a boost.

Study design ladder (strongest → weakest)

  1. 1.Network meta-analysis
  2. 2.Meta-analysis / systematic review
  3. 3.Randomized trial (RCT)
  4. 4.Clinical trial
  5. 5.Cohort
  6. 6.Case-control
  7. 7.Observational
  8. 8.Case report
  9. 9.Preclinical
Deterministic
Same paper, same score. Reproducible and auditable.
No LLM for ranking
Regex + heuristics only; AI is separate for summaries and Q&A.
After deep analysis
A second custom scorer refines the number once a paper is fully analyzed.