Assertion and citation pairs
Each conclusion is an assertion linked to the published records it cites.
Design partner pilot · invitation only
TCI is an experimental evidence-review tool that retrieves and organizes published literature associated with dietary-supplement ingredients.
Manually assembled illustration of the artifact format, using a fictional product. Not generated by TCI's pipeline and not a customer artifact.
Submitted claim
Example Creatine Powder
Supports strength gains during resistance training.
Assertion / citation pair
The cited literature includes a pooled analysis reporting strength gains with creatine plus resistance training and a small trial in which some between-group improvements were not statistically significant.
“Creatine supplementation with resistance training enhances upper- and lower-body muscle strength in adults aged < 50.”
“the greater relative improvements in PT and ACC for the CRE group were not statistically significant”
Artifact contract
Each conclusion is an assertion linked to the published records it cites.
Where recorded, supporting, contrary, no-effect and context excerpts are quoted beside the citation, with a link to the record.
Where recorded, the dossier lists the searches it ran: provider, query and date. A missing search record is shown as a coverage gap.
Where TCI lacks the reference data, the dimension is marked not assessed. It is never averaged away.
Safety is reported as explicit hazard states, never as a single number.
A stage that did not complete stays visible with its state. It is never shown as no evidence, complete or safe.
Process
A partner submits a product and the label claims to review.
TCI searches PubMed, OpenAlex and Europe PMC for the product's ingredients, and for the label claims whose outcomes it covers.
Retrieved records are screened for eligibility. Scanned, screened, eligible and analyzed counts are shown; an unrecorded count is marked unrecorded, never zero.
Findings are taken from the source text and written as assertions paired with their citations.
The software includes automated checks that cited PubMed records resolve and that quoted snippets match retrieved source text.
The dossier is stored as a versioned, append-only record, with its coverage gaps and failures intact.
Result states
MEASURED_FINDING
The required check ran and recorded eligible evidence.
NO_ELIGIBLE_EVIDENCE_IDENTIFIED_AFTER_COMPLETED_SEARCH
The search completed and identified no eligible evidence. This is not a weak positive.
NOT_APPLICABLE
This check does not apply to this subject.
NOT_ASSESSED_COVERAGE_GAP
TCI lacks the required reference or execution record here. No conclusion was made.
PROCESSING_FAILURE_RETRY_REQUIRED
The check did not complete. This is not no evidence, complete, or safe.
HAZARD_IDENTIFIED_REVIEW_REQUIRED
NO_LISTED_HAZARD_FOUND_IN_COVERED_CHECKS
NOT_ASSESSED
PROCESSING_FAILURE
A hazard state describes the cited evidence. It is not a safety determination.
Design partner pilot
We are working with a small number of supplement teams. Accounts are invitation-only, and pilot engagements are scoped and billed directly.
Share your work email, company, role, and the product work you want to review.
TCI reviews fit and capacity before any account is created.
Accepted partners receive an invitation and work from their signed-in dashboard.
Design partner pilot
Tell us about your team and the product work you want to review.
Pilot access is by invitation. Engagements are billed manually.