Proven & powerful
Solid but modest
High-potential, unproven
Speculative / weak
Impact / effect size →
Strength of proof →
An evidence-ranked atlas of 135 longevity interventions — drugs, supplements, herbs, diets, exercise, thermal therapies and frontier biotech — scored on the strength of the science behind each.
Educational only — not medical advice. Talk to a clinician before changing anything.
Every intervention is rated on four 0–100 dimensions, curated on one consistent rubric so they can be compared fairly. The headline Promise score is a weighted blend:
Evidence strength follows this hierarchy, strongest first:
These scores are an informed synthesis, not a clinical guideline. Evidence-grade ≠ personal recommendation: the right intervention depends on your goals, health and risk tolerance. The strongest-evidence items (exercise, not smoking, good diet, sleep) are unglamorous on purpose — that's what the data show. Frontier biotech scores low on current proof, not on future potential.
Aggregated live from the 135 interventions in this database. Compiled from a multi-agent web-research sweep of 2023–2026 literature (RCTs, cohorts, meta-analyses and the NIA Interventions Testing Program).
Geroscope is an evidence-ranked atlas of the things people take or do to live longer and healthier — from rapamycin and metformin to sauna, senolytics, fasting and gene therapy. The goal is to cut through hype with a transparent, comparable read on the quality of evidence behind each.
Each of the 135 interventions was researched across the 2023–2026 literature and scored on four dimensions — evidence strength, human translation, impact and safety (see How it's scored). Sources are linked on every card.
⚠ This is an educational tool, not medical advice. Evidence summaries are simplifications and the science moves fast. Nothing here is a recommendation to take, stop, or combine any intervention — talk to a qualified clinician first. Several listed items are experimental, unregulated, or carry real risks.
Open data & independent review: the full dataset is public — interventions.json · CSV (every key study links to its source, with the scoring methodology embedded). Fact-checking and corrections are welcome.
How this was built, who verified it, references & the equivalent human effort: How this was made →
Part of the 42-apps collection.
What it is. Geroscope is an evidence-ranked atlas of 135 interventions people use to try to slow aging — drugs, supplements, botanicals, diet, exercise, thermal/light, hormones/peptides and frontier biotech — each scored on the strength of the science behind it. It was conceived to cut through longevity hype with one transparent, comparable read on evidence quality, deliberately rewarding what is proven in humans over what is merely promising in mice or in marketing.
Compiled and audited through a structured multi-stage pipeline, with primary-source web research at each step (prioritising 2023–2026 literature — randomised trials, prospective cohorts, meta-analyses, and the NIA Interventions Testing Program):
Built by AI — stated plainly. Geroscope was compiled and written by an AI system (Anthropic’s Claude, orchestrating the agent fleets above), not by a named panel of human researchers or clinicians.
Verification. Internally, the adversarial fact-check and citation passes cross-examined the data. Externally, the public dataset was then independently fact-checked by three leading AI models — Google Gemini, xAI Grok and OpenAI GPT — each of which reviewed citations, effect sizes and grading and reported no factual errors (Gemini: “exceptionally accurate”; Grok: “one of the better-curated public syntheses available”). Their feedback was incorporated: 95% confidence intervals on headline figures, and six additional well-evidenced interventions.
No human clinician or formal peer reviewer has signed off on this work. AI cross-checking is not a substitute for expert human review. Treat the scores as an informed synthesis, not a clinical guideline — and consult a qualified professional before acting on anything here.
For transparency, a rough estimate of the effort an equivalent evidence-synthesis-plus-web-build would normally take a small specialist team — i.e. what it would require, not a claim that such a team did it. The AI pipeline produced it in a few hours of wall-clock time across dozens of parallel agents.
≈ 900–1,400 human-hours (~5–8 person-months) for a small specialist team.
Every intervention links its key studies to canonical sources. The complete dataset — 193 cited studies, scoring methodology embedded — is open for independent review: interventions.json · CSV. Landmark trials anchoring the top tiers include Jha (NEJM 2013), Arem (JAMA Intern Med 2015), Mandsager (JAMA Netw Open 2018), PREDIMED (NEJM 2018), the NIA ITP, EMPA-REG / DAPA-CKD, SELECT (NEJM 2023) and CALERIE-2.
Offered in the spirit of transparency about how AI-assisted evidence synthesis is produced. Corrections and expert review are warmly welcomed via the open data above. Educational only — not medical advice.