Living ecosystem observatory · Release 2026-08-01-r2

From scientific information to better decisions.

A connected map of organizations, infrastructure, care models, consumer products, and evidence systems spanning scientific production and the lived decision loop: encounter, explore, ground, evaluate, translate, contextualize, decide, act, observe, and revise.

Baseline classifications held stickyClaim-level source linksEvents separated from classification changesFull URLs in source library
146mapped entities
25verified additions
34functional categories
14dated events
193source records
Two connected lifecycles

Scientific production

The upstream lifecycle describes how evidence is funded, generated, reviewed, published, translated, and monitored.

Funding & scoping4 entities
Discovery & grounding14 entities
Evidence synthesis16 entities
Protocol & registration6 entities
Study design12 entities
Data collection & measurement10 entities
Experiment execution4 entities
Participant recruitment3 entities
Analysis & reproducibility10 entities
Authoring & submission7 entities
Submission readiness6 entities
Peer review & production5 entities
Post-publication integrity4 entities
Guideline & decision translation6 entities
Clinical evidence18 entities
Regulated devices1 entities
Regulatory & HTA10 entities
Public & media translation20 entities
Post-market learning19 entities
Provenance rail8 entities
Governance rail7 entities

Consumer decision capacity

The downstream lifecycle asks whether information becomes a calibrated choice, feasible action, observed outcome, and revised decision.

Encounter5 entities
Explore24 entities
Ground52 entities
Evaluate54 entities
Translate41 entities
Contextualize26 entities
Decide8 entities
Act21 entities
Observe31 entities
Revise12 entities
Current strategic read

Integration signals

Systematic review is a separate market

Discovery tools find and summarize papers. Review platforms manage protocol-driven screening, duplicate decisions, extraction, risk of bias, meta-analysis, and audit trails.

The clinical market splits at the regulatory surface

Data substrates, authorized AI devices, and clinician-facing evidence assistants have different buyers, evidence gates, and failure modes.

The content moat is visible at point of care

Clinical evidence products differentiate through licensed or curated corpora, citation verification, institutional distribution, and workflow access—not only model quality.

Execution is becoming a product boundary

K-Dense–Litmus, Potato, Lila, CuspAI, and Bunsen connect reasoning to computation, laboratory work, or industrial validation.

Post-publication and post-deployment remain underbuilt

Integrity tools are stronger before publication than after it, while deployed-model drift lacks a global surveillance substrate comparable to drug safety.

Nutrition evidence infrastructure is sparse

Examine is the clearest AI-enabled evidence anchor; most nutrition AI products focus on logging, recognition, or personalization rather than evidence synthesis and translation.

Primary care is the longitudinal integration layer

Primary care organizations can connect external evidence and personal data to decisions, action, monitoring, and plan revision; care models vary substantially in accessibility, incentives, and data depth.

Supplement decisions span multiple independent functions

Efficacy, interaction risk, product identity and purity, personal relevance, fulfillment, adherence, and stopping rules are separate jobs with different conflicts and evidence requirements.

Consumer health is converging around measurement, meaning, and action

Conference and product signals show dense investment in sensing, diagnostics, and engagement, while defensible interpretation, option comparison, clinical handoff, and outcome learning remain fragmented.

Acquisitions reveal missing workflow edges

Function–SuppCo is a concrete attempt to connect measurement with supplement quality and action; the integration should be evaluated as a decision pathway, not only as corporate news.

Searchable landscape

Organizations and systems

Each record has one primary job, no more than two scientific-stage assignments, and no more than two adjacent consumer-decision stages.

Potential collaborators

Primary-care pilot anchors

Rhythm Medical

Membership-based direct primary care with longer visits, direct physician access, chronic-condition and weight-management support, and recurring laboratory monitoring.

Official site ↗Evidence ↗

PrimaryMD

Membership primary care combining medical history, biomarkers, genetics, wearables, imaging, exercise physiology, and specialist coordination into one evolving physician-led plan.

Official site ↗Evidence ↗
Integration opportunities

Where the chain still breaks

Story-level evidence grading

No scaled successor clearly performs the former HealthNewsReview / Behind the Headlines function: comparing a health story with its underlying paper for causal overclaiming, absolute benefit, harms, costs, and conflicts.

High-value FRESH adjacency

Protocol-to-publication consistency

Registration systems timestamp commitments, but few products compare protocols, analysis plans, outcomes, and manuscripts as a continuous provenance chain.

Trust infrastructure opportunity

Model post-market surveillance

Clinical AI performance can degrade through dataset shift, workflow drift, and over-reliance without producing a reportable adverse event.

Regulatory and health-system gap

Nutrition-specific living evidence

Dietary evidence has severe exposure-measurement, substitution, confounding, and translation problems, but little purpose-built AI infrastructure.

Domain-specific opening

Independent scientific-agent benchmarks

Vendor-run comparisons dominate. Hidden tasks, fresh data, executable environments, preregistered scoring, and expert adjudication remain uncommon.

Metascience infrastructure

End-to-end provenance

Source → hypothesis → protocol → data → code → result → claim → review → decision remains split across incompatible systems.

Cross-market moat

Measurement-to-meaning integration

Wearables, labs, and imaging generate longitudinal observations, but consumers still need calibrated interpretation of what deserves action, monitoring, professional escalation, or no response.

Cross-category integration opportunity

Evidence-to-primary-care handoff

Evidence tools rarely export decision-ready, patient-specific option and uncertainty structures into longitudinal primary-care workflows.

Collaboration opportunity

Supplement efficacy-to-product-quality chain

Evidence of ingredient efficacy, interaction risk, product testing, commercial recommendation, adherence, and biomarker follow-up remain split across organizations.

Consumer decision infrastructure

Decision-to-behavior feedback

Behavior platforms can sustain action, but often cannot preserve the evidence rationale, decision thresholds, or conditions under which the user should revise or stop.

Outcome-learning opportunity

Accessible longitudinal decision support

Data-rich concierge models demonstrate integration but may not establish a scalable or equitable route to population decision capacity.

Distribution and equity gap
Dated developments

Timeline

Historical backfills are labelled and are not counted as newly occurring August developments.

2026-07-27
Schrödinger Bunsen
Product launch
Bunsen announced in early access
Schrödinger introduced an agentic co-scientist designed to plan and interpret computational molecular-discovery workflows.
Baseline event ↗
2026-07-23
K-Dense + Litmus Science
Partnership
Hypothesis-to-experiment bridge announced
The companies described a workflow connecting agentic research and hypothesis generation to outsourced physical experiments.
Baseline event ↗
2026-07-23
CuspAI
Network
AI Materials Foundry network expanded
The University of Cambridge announced participation in CuspAI’s founding laboratory network for materials discovery.
Baseline event ↗
2026-07-22
Lila Sciences
Award
Selected for DOE Genesis Mission projects
Lila reported selection for three Phase I lighthouse projects with national-laboratory and university partners.
Baseline event ↗
2026-07-10
OpenEvidence
Product launch
EvidenceGrade launched
OpenEvidence added a displayed assessment of the strength or certainty of evidence supporting clinical answers.
Baseline event ↗
2026-07-05
CMS ACCESS Model
Payment model
Outcome-aligned digital chronic-care model began
CMS began a ten-year model paying technology-supported chronic-care organizations partly on measured outcomes.
Baseline event ↗
2026-07-02
DeepEvidence
Publication
DeepEvidence published in Nature Machine Intelligence
The paper describes multi-agent evidence exploration and synthesis over a biomedical knowledge graph.
Baseline event ↗
2026-06-30
Claude Science
Product launch
Claude Science launched in beta
Anthropic introduced a scientific workbench integrating tools, compute, artifacts, and audit history.
Baseline event ↗
2026-06-16
FDA
Registry update
AI-enabled medical-device list updated
FDA refreshed its structured registry of authorized AI-enabled medical devices and reiterated work on identifying LLM-based functionality.
Baseline event ↗
2026-05-12
Function Health + SuppCo
Acquisition
Function Health acquired SuppCo
The transaction connects longitudinal biomarker measurement with supplement routine management, product-quality intelligence, recommendations, and follow-up.
Historical backfill ↗
2026-03-17
Cochrane
Independent evaluation
LASER.ai and Nested Knowledge selected for platform study
Cochrane selected two tools from 48 submissions for prospective evaluation inside review workflows.
Baseline event ↗
2026-02-21
Stanford Medicine Consumer Health Conference
Conference ecosystem
Consumer-health ecosystem convened at Stanford Medicine (self-described)
The conference linked founders, investors, scientists, clinicians, wearables, diagnostics, behavior platforms, longevity services, and preventive-care companies.
Historical backfill ↗
2025-11-11
Cochrane + Campbell + JBI + CEE
Governance
Joint responsible-AI position published
Four evidence-synthesis organizations aligned around RAISE and human accountability for AI-assisted reviews.
Baseline event ↗
2025-06-02
FDA Elsa
Government deployment
Agency-wide generative AI launched
FDA launched Elsa for internal scientific review, document access, coding, inspection, and operational support.
Baseline event ↗
Reference library

Complete source URLs

Dates are explicit where available; otherwise the record states the date basis. Verification reflects what was actually retrieved in this cycle.

accessed 2026-08-01
Zenodo
fetched
Official repository
https://zenodo.org/
Official or evidentiary page
accessed 2026-08-01
Yseop
fetched
Official product
https://yseop.com/
Official or evidentiary page
accessed 2026-08-01
USP Dietary Supplement Verification
blocked
USP Dietary Supplement Verification official site
https://www.usp.org/
Official site
accessed 2026-08-01
SuppCo
fetched
SuppCo official site
https://supp.co/
Official site
accessed 2026-08-01
Scite
fetched
Official platform
https://scite.ai/
Official or evidentiary page
accessed 2026-08-01
ORCID
fetched
Official infrastructure
https://orcid.org/
Official or evidentiary page
accessed 2026-08-01
NSF Certified for Sport
fetched
NSF Certified for Sport official site
https://www.nsf.org/
Official site
accessed 2026-08-01
ICER
fetched
Official organization
https://icer.org/
Official or evidentiary page
accessed 2026-08-01
Guidelines International Network
fetched
Official network
https://g-i-n.net/
Official or evidentiary page
accessed 2026-08-01
Ground News
fetched
Official product
https://ground.news/
Official or evidentiary page
accessed 2026-08-01
Full Fact
fetched
Official organization
https://fullfact.org/
Official or evidentiary page
accessed 2026-08-01
Elicit
fetched
Official product
https://elicit.com/
Official or evidentiary page
accessed 2026-08-01
DataCite
fetched
Official infrastructure
https://datacite.org/
Official or evidentiary page
accessed 2026-08-01
CuspAI
fetched
CuspAI official site
https://cusp.ai/
Official site