Job 01
Personalized longevity
Coordinate your own longevity data, plans, and day-to-day follow-up. Care coordination assist. Companion scope.
01 — Opening constraint
Laws of
Longevity AI
Aging is an engineering problem. LongevityGrok exists to keep capable humans in the fight longer · sharper, healthier, measurable · while turning private effort into shared discovery. These laws govern the system. They are not medical advice. They are design constraints.
- Personal longevity superintelligence
- Grok Bot
- Research assist
- Not a medical device
-
01
Extend Peakspan
Keep people productive, healthy, and improving for as long as possible. Optimize the years of high function, not slogans about immortality.
-
02
Serve the Individual First
Every install exists for the person running it. Personal state, plans, and research-for-you come before any communal ask. Contribution is optional. Withdrawal is a right.
-
03
Dual-Use the Signal
Use data and biological samples for the owner's benefit and, under explicit consent, for shared therapeutic discovery. Same measurements. Two payoffs. No silent secondary market.
-
04
Agents Judge; Code Constrains
Models propose. Deterministic code scores, gates, and refuses. A failure of judgment must produce unnecessary refusal, never an unsafe action. Predictions that cannot be checked do not ship.
-
05
Consent Is Default-Off for Sharing
Registry, fleet insights, and research-gateway scopes start off. Local data-room work can proceed without them. Changing the consent text resets packing until the user re-accepts.
-
06
Two Trust Highways
Physician links are known-identity paths to clinicians the user already trusts. MotherShip links are typed, deidentified packages through a controlled relay · never free-text chat, never PHI on that wire.
-
07
Improve Until Aging Is Solved
Continuously raise the system's understanding of longevity · literature, mechanisms, targets, biomarkers, education · until aging is solved as an engineering problem. Ship measurable increments. Do not declare victory early.
LongevityGrok is a research, education, and care-coordination companion · not a medical device, and not a substitute for a licensed clinician.
02 — The system
Personal longevity superintelligence
LongevityGrok (PLSI) is a multi-agent companion inside Grok Bot. It connects consumer longevity practice to Insilico-grade research infrastructure without pretending to be a clinic in a box.
Put longevity tools into the hands of the willing. Anyone with a Grok Bot can run a personal longevity research stack · coordinate their own data and plans, press literature against their goals, talk to longevity physicians they already know, and optionally feed a shared research layer.
Job 02
Research for you
Surface literature, mechanisms, and questions tied to your goals. Private to the install.
Job 03
Research for the community
Optionally contribute consented, deidentified insights. The scopes that share start off.
Job 04
Known physicians
Prepare visits and draft messages you approve, with clinicians you already know.
Job 05
MotherShip relay
Typed, deidentified packages into the Longevity MotherShip. Free-text chat is not that wire. Residual re-identification risk remains after coarsening.
At install you choose Individual, Longevity physician, or Pro researcher. Everyone gets the personal longevity agents. Physicians and pro researchers unlock extra collaboration and workbench tools as those packages land. Early builds do not contain every extra.
03 — Thesis
Tools into the hands of the willing.
For themselves first
Personal state, plans, and research-for-you come before any communal ask.
For the community when they choose
Contribution is a deliberate second act. Withdrawal is a right.
On Grok Bot
The runtime is persistent agents, routines, connectors, and approval cards. Multi-agent collaboration is the point of the substrate.
04 — Bring-up
Run it on a Grok Bot
Operator checklist on file: template plsi-personal-v0, 19 Oct 2026, plus the getting-started skill. Public repositories were verified 1 Nov 2026. Read the repository you clone for the current steps. This page is the claim boundary.
Personal suite first
Labs, wearables, protocols, literature, and visit prep on your machine.
Optional contribution
Consented, deidentified packages can relay to the Longevity MotherShip. Those scopes stay off until you turn them on. Enrolment is steward-gated.
Role upgrades
Longevity physician and Pro researcher packages add collaboration tools as they ship. They are product packages, not a promise that every extra is in the build you clone.
Consent defaults
| Scope | Default | What it governs |
|---|---|---|
| data_room_core | Required | Store and organize documents in the private data room. |
| Connectors | Opt-in | Gmail, Drive, and Calendar, each explained before connect. Least privilege. Prefer read and draft. Sends go through approval cards. |
| clinician_draft_assist | Opt-in | Draft messages to clinicians you already know. You still approve sends. |
| registry_contribute | Off | Sanitized physician or method tips. Never your labs. |
| fleet_insights_v1 | Off | De-identified insight packets toward the research aggregator. |
| tool_gateway_research | Off | Calls to research tools with minimized, non-chart payloads. |
| education_hub_sync | On · topics only | Topic-level help from the Longevity Medicine Education Hub. No lab uploads to the Hub. |
Operator sequence · greenfield
- Create a new Grok Bot so the PHI boundary stays clean.
- Load the package. Missions, skills, and routines. The public client repository is the code reference; the 19 Oct 2026 checklist names BOT_SPEC and SKILLS_MANIFEST.
- Create the local store. Consent ledger (Guardian only), audit, profile, config, data room, protocol, analysis, evidence, physician comms, sandbox.
- Connect Gmail, Drive, and Calendar only if you opt in. Drive folders used by the checklist: Health/Labs, Devices, Visits, Protocol, Digests, Sandbox, PLSI-Exports.
- Run getting_started. The disclaimer acknowledgment blocks the rest of the wizard. Profile asks for a display name, timezone, units, and birth year rather than a full date of birth. It does not ask for an MRN, insurance id, or government id.
- Seed clinicians locally if you want a private list. That list stays out of any public registry.
- Enable routines in your timezone: ingest watches, nightly analysis, a weekly digest you schedule, and pre-visit prep when a calendar event is tagged.
- Leave sharing off unless you explicitly opt in. Hub topic sync can stay on.
- Smoke-check on your own bot. A lab PDF can become observations with provenance. Low-confidence parses ask you to verify. A digest carries the disclaimer and does not invent citations. A clinician draft stops on an approval card. A registry tip that contains a patient name is denied. A gateway payload that looks like a chart is rejected.
The acknowledgment the skill must show, before anything else:
PLSI is a research, education, and care-coordination assistant. It does not diagnose, does not prescribe, and is not a medical device. Clinical decisions stay with you and your licensed clinicians. In emergencies, call local emergency services.
Brownfield, pause, and operator limits
Attaching to an existing Grok Bot means every key lives under plsi/, longevity intents route to the PLSI orchestrator, general chat cannot read the data room, and the Privacy Guardian still intercepts egress. Greenfield is the cleaner audit.
Pause by disabling routines, revoking connector scopes you no longer need, exporting the store and the Drive health tree, and deleting the store only after an explicit confirmation. Say so if consent or audit records must be retained. Withdrawing fleet insights stops new packets.
Credit budgets prioritize safety and ingest, then digests and visit prep, then evidence, then sandbox. Marketing does not promise unlimited autonomous research spend. Keep identifiable health information out of public chat, tickets, and screen recordings. Use fixtures for demos.
Micro-disclaimer. Research, education, and care-coordination assistant. Not a medical device. MotherShip return digests and enterprise command features are design targets where not yet shipped. See the Laws of Longevity AI.
05 — Who it is for
Three installs. One research brain.
Every personal install gets the same core agents. Physician and pro-researcher rows are product packages. Not every extra is in Fedor 0.1.0.
| Package | Who | What they get | Status |
|---|---|---|---|
| Individual | Anyone optimizing their own longevity | Personal agent suite. Opt-in community contribution. Later digests when that return path ships. | Personal suite |
| Longevity physician | Clinicians in longevity practice | The same suite, plus panel and prep helpers, patient messaging, and practice-pattern aids as the package lands. Judgment stays with the clinician. | Package |
| Pro researcher | Scientists, drug discovery, methods | The same suite, plus a target, pathway, repurposing, and KOL workbench and a MotherShip collaboration path as the package lands. | Package |
| MotherShip | Enterprise research brain, Insilico-operated | Consented deidentified packages in. Aggregate releases and reviewed notes out. Return digests to personal installs are a design target. | Partial ops |
Scientific framing
Alex Zhavoronkov
Founder and Co-CEO, Insilico Medicine. Primary stakeholder. Owns scientific framing, ethics language, Hub policy, and what may be claimed in public. Lead contact on the current manuscript draft.
Science and build
Fedor Galkin
Science and implementation. MotherShip package. Author of manuscript draft v14. Execution owner for schemas, bots, and methods-accurate text.
Education Hub
Dominika Wilczok
Manages Longevity Medicine Education Hub involvement. Education path into LongevityGrok and physician packages. Hub home: longevity.degree.
Research tools
Insilico Medicine
Origin of the discovery tooling Pro packages can surface as graded evidence, and operator of the Longevity MotherShip. Upstream science. The personal install does not become a clinic because the company runs discovery programs.
Research network
ARDD
Conference and collaborator gravity for the extended research network, as named in the ecosystem brief. No attendance, partnership, or program claim is added here.
Scientist collaborator
David A. Sinclair
Named author on draft v14. Scientist collaborator. Not a product endorser. Not a shipped feature. The site does not treat validation credit in the draft as an endorsement.
06 — What it is for
Benefits, inside companion scope
Research, education, and care-coordination assist. Outputs are briefs, questions, literature maps, and plans for a human to review. The system does not diagnose, prescribe, or act as a medical device.
Care coordination
One auditable place for labs, wearables, protocols, and visit prep. Scored predictions instead of fluent guesses. The record of a check stays outside the model that proposed it.
Private research-for-you
Literature, mechanisms, and questions pressed against your goals, on your install, before any communal ask.
Optional MotherShip
The same measurements that serve self-care can, under explicit consent, become coarsened packages for shared discovery. Defaults off. No PHI on that wire.
Known-identity path
Structured briefs and drafts for clinicians you already trust. Nothing in that lane replaces the clinician’s judgment. Clinic-side agents do not author a diagnosis, a prescription, or an order.
Workbench, not identity
Role tools aimed at targets, pathways, literature, and MotherShip releases. The researcher path is not a window into a contributor’s identity.
Human decisions
Plans stay marked for review. Emergencies go to local emergency services. Commercial rankings, if they ever exist, have to stay transparent. Today this is a research companion.
07 — Constellation
Ecosystem map
Rebuilt for this surface from the AZDot diagram spec. Solid marks shipped or partial paths. Dashed marks design targets. The print master was a white matplotlib export; this drawing is the dark schematic.
- Solid · shipped or partial
- Amber word · product package, not fully in early builds
- Dashed · design target
- Steel · external network
Personal → relay → Mother pack client: live path. Mother → PLSI digests and the full researcher download portal: design targets. Not a medical device. No PHI on the MotherShip wire. Residual re-identification risk remains even after coarsening. This is limited-field deidentification, not absolute anonymity, and not a Safe Harbor claim.
Node register · shipped versus design target
| Id | Node | Cluster | Status on this page |
|---|---|---|---|
| gb | Grok Bot runtime | Edge | Shipped platform |
| pkg_ind | Individual install | Edge | Shipped framing / early builds |
| pkg_md | Longevity physician package | Edge | Product package · not all in Fedor 0.1.0 |
| pkg_pro | Pro researcher package | Edge | Product package · not all in Fedor 0.1.0 |
| agents | Personal longevity agent suite | Edge | Shipped / evolving |
| visit | Visit prep and approval-card drafts | Clinical | Shipped pattern |
| phys | Longevity physicians you know | Clinical | Human review |
| relay | Cloudflare relay | MotherShip | Shipped personal → Mother pack |
| mother | Longevity MotherShip | MotherShip | Shipped repo / partial ops |
| digest | Mother → PLSI digests | MotherShip | Design target |
| idle | Idle Hunter and priority board | MotherShip | Design target |
| cell | cell@ command plane | MotherShip | Design target |
| insilico | Insilico research tools | Network | External · PandaOmics and generative design |
| biomarkers | Biomarkers and aging clocks | Network | Methods and tools · state-map coordinates |
| leh | Longevity Medicine Education Hub | Network | Live education path · longevity.degree |
| ardd | ARDD and the research network | Network | Conference and collaborator gravity |
| sinclair | David A. Sinclair | Network | Scientist collaborator · not an endorser |
08 — Lineage
AI in longevity, then the personal edge
No separate history pack was in the site sources. The stages follow the AZDot diagram spec and the lineage already named in the PLSI drafts. They are a field narrative. They are not a claim that LongevityGrok practices medicine.
S1 · Descriptive
Clocks and panels
Blood and methylation clocks, and population biomarkers, turned aging into coordinates someone can plot.
Insilico role in the drafts: aging-clock methods and publications. PLSI uses them as state-map coordinates.
S2 · Predictive
Trajectories
Organ-age maps, multi-omics risk models, and research digital twins ask where a system is going.
Predictive biology and target hypotheses. In PLSI, a prediction has to be scorable. Uncheckable forecasts do not ship.
S3 · Generative
Discovery
Target identification and generative chemistry. The drafts cite PandaOmics, Chemistry42, and rentosertib as a program that reached a phase 2a trial.
This is Insilico’s discovery business. This page states no clinical outcome figures.
S4 · Agentic
Personal agents
Records move from scattered portals toward a reasoned personal state. Persistent agents run on the person’s machine.
Platform choice for PLSI: Grok Bot. The personal install is the edge. It does not inherit a prescription right from a discovery platform.
S5 · Dual-use
LongevityGrok
The same install coordinates personal care and, if the person opts in, communal research.
MotherShip, open repositories, Education Hub, physician and pro packages. Build. Measure. Improve until aging is solved.
Break in the line. Insilico drug discovery is upstream science. PLSI consumes graded evidence and coordinates personal research. Human clinicians retain judgment. Pro researcher tools grade external model output as low-level evidence by default. There is no arrow from a clinical program to “PLSI diagnoses” or “PLSI prescribes.”
Works named because they appear in the workspace drafts, not as a bibliography invented for this page: Zhavoronkov and Mamoshina, deep aging clocks, Trends in Pharmacological Sciences, 2019, doi:10.1016/j.tips.2019.05.004. Zhavoronkov, Bischof, and Lee, artificial intelligence in longevity medicine, Nature Aging, 2021, doi:10.1038/s43587-020-00020-4. Kamya et al., PandaOmics, Journal of Chemical Information and Modeling, 2024, doi:10.1021/acs.jcim.3c01619. Ivanenkov et al., Chemistry42, Journal of Chemical Information and Modeling, 2023, doi:10.1021/acs.jcim.2c01191. Rentosertib citations in draft v14: Ren et al., Nature Biotechnology, 2024, doi:10.1038/s41587-024-02143-0, and Xu et al., Nature Medicine, 2025, doi:10.1038/s41591-025-03743-2. No efficacy numbers are restated here.
09 — Manuscript
Draft on file. Not a deposit.
Latest manuscript on file: Fedor draft v14, ingested 1 Nov 2026 (Asia/Dubai). Open client and MotherShip repositories at Insilico-org. Not a peer-reviewed deposit. Not a medical-device paper.
Personal Longevity Superintelligence: A Collaborative Ecosystem for Personal Care and Communal Longevity Research
Lead contact on the manuscript: Alex Zhavoronkov, Insilico Medicine. This page does not publish a correspondence address.
Both organization repositories were verified public on 1 Nov 2026. Canonical client URL is lowercase plsi. The draft sometimes cites an uppercase path; GitHub resolves it. Citing the live URLs is a fact about the code. It is not an official journal deposit, a bioRxiv posting, or a medRxiv posting. No public preprint URL is on file. Do not invent one.
Claim hygiene. v14 is the author figure and prose source, plus the public-repo facts. The 28 Oct 2026 pass-2 CLEAN redline — the prior Heavy-Grok claim-hygiene path — remains the restraint until a merge absorbs it. In particular: the relay separates identity from content under limited-field deidentification; residual re-identification risk remains; do not say the relay hides identity in the absolute sense, and do not claim Safe Harbor anonymity. David A. Sinclair is a scientist collaborator, not an endorser of a product.
Earlier title, still the product lock from 19–20 Oct 2026, unused as the v14 title: Personal Longevity Superintelligence: Lifelong Always-on Research and Healthcare Multi-Agent Collaborative Ecosystem Based on Grok Bot. Preprint v4.5 and the 28 Oct redline are older artifacts. v14 is the newest author draft.
What v14 reports
Architecture and offline software checks on synthetic fixtures. It reports no identifiable personal health data. It is not a clinical-benefit paper, a diagnostic-accuracy paper, or an autonomous-prescribing paper. Hosted Grok chat in the personal tier processes user content with the cloud model provider.
What stays human
Diagnosis, prescriptions, clinical judgment, and medical advice remain human. Clinic agents in the draft never author a diagnosis, a prescription, or an order. A physician approves anything that would reach a patient.
Copy fold
Section copy is composed from _azdot/. Laws, description, diagram spec, install honesty, and this paper pointer can be replaced there without redesigning the plates. Rendered HTML is what the browser ships.