2011–2016
Portfolios and pathway tools
Aging research portfolios turned into machine-readable maps with SVM and related methods, then signaling-pathway tools for ranking geroprotectors and the curated Geroprotectors.org database.
The science behind Longevity Grok
Research
Insilico Medicine has spent fifteen years building AI for longevity, from aging research portfolios and deep aging clocks to PreciousGPT and LongevityBench. Longevity Grok brings that work into a personal companion.
Fifteen years of AI for longevity
Grants, papers, patents and products, each step backed by a primary source in the bibliography below.
2016–2021
Deep aging clocks
The first deep-learning aging clock from routine blood tests (Aging.AI), population-specific clocks, the first microbiome clock, PhotoAgeClock with Haut.AI, and granted U.S. patents for transcriptomic and proteomic clocks.
2019–2025
From biology to drugs
PandaOmics for target discovery and Chemistry42 for generative chemistry, leading to rentosertib, a program that reached a phase 2a trial.
2023–2026
Transformers and benchmarks
Precious1GPT, Precious2GPT and Precious3GPT brought multimodal transformers to aging research. The 2026 Cell paper released LongevityBench, longevity language models and Longevity Claw.
Today, Longevity Grok integrates this stack as a personal research companion on Grok Bot, so these tools keep improving and reach more people.
Psychological aging
Longevity isn't only blood and methylation. Insilico's work treats psychological age as something you can measure, and shows that it affects biological aging. That's why Longevity Grok keeps biology and psychology in the same picture.
Psychological factors substantially contribute to biological aging: evidence from the aging rate in Chinese older adults
Galkin F, Kochetov K, Koldasbayeva D, Faria M, Fung HH, Chen AX, et al. · Aging 14(18):7206 · doi:10.18632/aging.204264
The most direct published link between psychological state and the rate of biological aging in this series. Psychological factors are measured alongside aging clocks and treated as real contributors to how fast we age.
- Measure psychological and subjective age, not slogans
- Relate psychological factors to the rate of biological aging
- Keep sharing and clinical action in human hands
PsychoAge and SubjAge
Deep markers of psychological and subjective age using AI. Psychological age becomes something a model can learn, like blood and image clocks.
Zhavoronkov A, Kochetov K, Diamandis P, Mitina M. Aging 12(23):23548. doi:10.18632/aging.202344
Psychological aging, depression and well-being
Maps psychological aging against depression and well-being, and shows why mental health belongs in a longevity plan.
Mitina M, Young S, Zhavoronkov A. Aging 12(18):18765. doi:10.18632/aging.103880
Islands of emotional stability
Self-organizing maps for future well-being: a person's life path as a space you can navigate. An early precedent for making personal state navigable, as Longevity Grok does.
Galkin F, Kochetov K, Keller M, Zhavoronkov A, Etcoff N. Aging 14(12):4935. doi:10.18632/aging.204061
Psychogenic aging
Brings together the psychobiological hallmarks of aging. Psychological factors are among the strongest predictors of healthspan, yet rarely part of medical aging frameworks.
Faria M, Ganz A, Galkin F, Zhavoronkov A, Snyder M. Translational Psychiatry 14:226. doi:10.1038/s41398-024-02919-7
Physician education
The Longevity Medicine Education Hub trains physicians and researchers in evidence-based longevity medicine. Its alumni and chapter physicians are the natural community for Longevity Grok's physician and researcher tools. Longevity Grok shares topics with the Hub, but no lab results are ever uploaded to it.
Longevity Medicine 101 and 201
Free, CME-accredited courses for medical professionals covering biogerontology, geroscience, aging clocks, geroprotectors, AI and clinical translation, through the Medical Society of Delaware's ACCME-accredited framework. More than 13,000 medical professionals had enrolled by October 2025.
Deep Longevity Medicine Education
Available on the NHS Learning Hub since 2021, authored by Alex Zhavoronkov, and through HDR UK Futures as Longevity Medicine 101.
In the MD curriculum
Indonesia: Healthy Longevity Medicine is mandatory in the MD curriculum at Universitas Negeri Makassar (about 516 hours; inaugural cohort of 47; first graduates expected with the class of 2030).
Thailand: mandatory in 4th-year Neuro-Geriatrics at KMITL (King Mongkut's Institute of Technology Ladkrabang); 29 students passed in Jan–Mar 2025.
National chapters in Brazil, Spain, Thailand, Indonesia, China and Japan, with more universities and clinics running the curriculum or CME pathways.
Under the hood
Every part of the ecosystem and where it stands today.
| Part | Layer | What it does | Status |
|---|---|---|---|
| Grok Bot runtime | Personal | Persistent agents, routines, connectors and approval cards | Live |
| Personal longevity agent suite | Personal | Chief Orchestrator, 8 core bots and 13 specialists you pick | Live · evolving |
| Personal Longevity Researcher package | Personal | The personal setup everyone starts with | Live · early builds |
| Longevity Physician package | Personal | Clinic team: Patient Coordinator, Visit Assistant, Evidence Analyst, Clinic Liaison | Early access |
| Institutional Researcher package | Personal | Lab team: Target Hunter, Cartographer, Cohort Analyst, Grant Writer | Early access |
| Research Autopilot | Personal | Daily research on one question, inside the budget you set (new in v0.2.0) | Live |
| Visit prep and approval cards | Clinical | Drafts for clinicians that wait for your approval | Live |
| Longevity physicians you know | Clinical | A private list of your clinicians; every draft gets human review | Live · human review |
| Cloudflare relay | MotherShip | Forwards shared data from your Longevity Grok to the MotherShip | Live |
| Longevity MotherShip | MotherShip | Insilico's research hub: the relay, the Mother and its data stewards (v0.2.0) | Live · partial operations |
| Insilico research tools | Network | PandaOmics and generative chemistry | External |
| Biomarkers and aging clocks | Network | The coordinates of your longevity map | Methods |
| Longevity Medicine Education Hub | Network | Physician education at longevity.degree | Live |
| ARDD and the research network | Network | Conference and collaborator network | Network |
From clocks to personal agents
How AI in longevity got here, in five stages.
- Describe
Clocks and panels
Blood and methylation clocks and population biomarkers turned aging into coordinates you can plot. Longevity Grok uses them as the axes of your longevity map.
- Predict
Trajectories
Organ-age maps, multi-omics risk models and research digital twins ask where a body is heading. In Longevity Grok every prediction must be checkable; forecasts that can't be scored don't ship.
- Generate
Discovery
Target identification and generative chemistry: PandaOmics, Chemistry42, and rentosertib, a program that reached a phase 2a trial.
- Act
Personal agents
Records move from scattered portals into one reasoned personal picture, with persistent agents running for each person on Grok Bot.
- Share
Longevity Grok
The same Longevity Grok coordinates your personal care and, only if you opt in, community research. Build, measure, improve, until aging is solved.
Insilico's drug discovery is the upstream science. Longevity Grok uses graded evidence from it to support your personal research, and human clinicians keep clinical judgment. The personal install doesn't inherit any right to prescribe from a discovery platform, researcher tools grade outside model output carefully by default, and no clinical outcome figures are claimed here.
How findings are made
Professional access and the preregistered findings loop: how verified researchers turn MotherShip releases into checked, published findings.
- 1
Access
A researcher requests access with an ORCID iD or licence, an institutional email and a data use agreement.
A steward checks them and approves the role and scopes. The Mother emails a one-time invite link, valid 72 hours, to the institutional address only.
Scoped token shown once, kept in the Grok secret store, expires within a year, revocable at once.
StewardMotherRelay - 2
Monthly release
Cells by coordinate, age band and sex, one value per contributor. Cells with fewer than 10 contributors are withheld.
A steward approves the counts before anything is published. The catalog is public; releases are readable only with the releases:read scope.
MotherSteward - 3
Preregistered analysis
The Cohort Analyst drafts a hypothesis from the catalog and fixes the release version without reading its values.
The preregistration is recorded first in an append-only findings ledger; only then is the release opened. The analysis runs the same code as the Mother, and null results are reported.
LabMother - 4
Review and publication
The Mother reruns it on the stored release and must get the same result within 10−6. Then an independent review (self-review refused) and the false discovery rate over all preregistrations.
A steward accepts or rejects it with a reason. Accepted findings are published with q-values, readable by physicians and researchers.
MotherSteward
Bibliography
Key firsts and milestones, each linked to its primary source.
| Year | Work | Note | Source |
|---|---|---|---|
| 2011 | IARP: machine learning on aging research grants and papers | Taxonomy of aging research portfolios | 10.1371/journal.pone.0022597 |
| 2014 | Weighted kNN, then SVM/RF on AgingPortfolio grant text | Multilabel text classification | 10.1155/2014/781807 |
| 2014 | Signaling-pathway cloud for in silico geroprotector ranking | Pathway regulation tools | 10.3389/fgene.2014.00049 |
| 2015 | Geroprotectors.org curated intervention database | Database and Aging paper | 10.18632/aging.100799 |
| 2016 | Geroprotector evaluation criteria; GeroScope signalome screen | Pathway and product line | 10.1111/acel.12463 · 10.18632/aging.101047 |
| 2016 | First deep-learning blood aging clock: Deep biomarkers of human aging (Aging.AI) | Neural-network ensemble on routine blood tests | 10.18632/aging.100968 |
| 2018 | Population-specific blood aging biomarkers | Korean, Canadian and Eastern European cohorts | 10.1093/gerona/gly005 |
| 2018 | PhotoAgeClock with Haut.AI | Visual photographic biomarkers of aging | 10.18632/aging.101629 · haut.ai |
| 2018–2020 | First microbiome aging clock | 2018 bioRxiv preprint; Science news, Jan 2019; iScience 2020 | 10.1101/507780 · Science news · 10.1016/j.isci.2020.101199 |
| 2019 | Granted U.S. patent: deep transcriptomic aging clock | US10325673 | US10325673 |
| 2019 | Deep aging clocks review (Zhavoronkov & Mamoshina) | Trends in Pharmacological Sciences | 10.1016/j.tips.2019.05.004 |
| 2020 | Granted U.S. patent: deep proteomic aging clock | US10665326 | US10665326 |
| 2021 | AI in longevity medicine (Zhavoronkov, Bischof & Lee) | Nature Aging | 10.1038/s43587-020-00020-4 |
| 2021 | DeepMAge methylation aging clock | Deep-learning DNA methylation clock | 10.14336/AD.2020.1202 |
| 2023 | Chemistry42 (Ivanenkov et al.) | Generative chemistry platform; J Chem Inf Model | 10.1021/acs.jcim.2c01191 |
| 2023 | Precious1GPT | Multimodal transformer aging clocks and targets (Urban et al.) | 10.18632/aging.204788 |
| 2024 | PandaOmics (Kamya et al.) | Target discovery platform; J Chem Inf Model | 10.1021/acs.jcim.3c01619 |
| 2024 | Rentosertib: AI-discovered TNIK inhibitor (Ren et al.) | Nature Biotechnology | 10.1038/s41587-024-02143-0 |
| 2024 | Precious2GPT | Multi-omics transformer with conditional diffusion | 10.1038/s41514-024-00163-3 |
| 2024 | Precious3GPT preprint | Multimodal, multi-species, multi-omics transformer | 10.1101/2024.07.25.605062 |
| 2025 | Rentosertib phase 2a trial (Xu et al.) | Nature Medicine | 10.1038/s41591-025-03743-2 |
| 2026 | Proteomic aging clocks in the rentosertib phase 2a trial | Clinical aging-clock readout | 10.1038/s41587-026-03286-y |
| 2026 | An open benchmark and language models for AI in aging biology | Cell, 17 Sep: LongevityBench, Longevity-LLMs, Longevity Claw | 10.1016/j.cell.2026.08.026 · longevitybenchmarks.org |
| 2020– | Longevity Medicine Education Hub: Longevity Medicine 101/201 | CME curriculum; more than 13,000 enrolled by Oct 2025 | longevity.degree · 101 |
| 2021– | NHS Learning Hub: Deep Longevity Medicine Education; HDR UK Futures | UK physician learning platforms | NHS Learning Hub · HDR UK Futures |
| 2025 | Longevity medicine in university MD curricula: Indonesia and Thailand | Universitas Negeri Makassar; KMITL | Longevity.Technology · JMIR Med Educ |