Trust should be inspectable.

Decision dossier layer / 2.1.1

Deeper evidence, with a separate review clock.

29 selected summaries across 12 initial destinations connect entry/residence research, institution checks and scoped cost references. 27 sources have a content-review date; 2 additional pointers remain unconfirmed. Coverage differs by country and layer. A government requirement, a university estimate and your own budget are never interchangeable.

Requirements are normally due for another source review after 30 days; other summaries after 90 days unless a shorter interval or reference-period boundary applies. These are TGPI operational review intervals, not guarantees of legal validity. Overdue or expired summaries are labelled. Source authority does not remove the need to confirm personal applicability.

Source Watch retrieves exact registered pages, with six-hour caching, time and size limits and at most one transient retry. Access restrictions and redirects are not bypassed. HTTP status, selected text markers and normalized-content fingerprints are recorded. A changed fingerprint or missing marker is a review signal, not proof of a changed law. Successful automated checks never renew editorial dates or rewrite rules. A configured daily workflow archives checks and compares available fingerprints with its previous cache when present; the initial run has no historical baseline.

The Decision Lab stores one active worksheet per signed-in account, only after an explicit save. It preserves profile, shortlist and learning metadata. Numbers, quote notes and checkpoint completions are user-authored assumptions, not verified evidence or legal approval. Changing currency clears amounts rather than silently relabelling them.

One source of evidence. Several useful decisions.

Confidence without false precision.

Our labels describe evidence handling, not the probability a statement is true. “Source checked” requires a registered source, matching indicator and country identifiers, a valid numeric value, complete pagination and an acceptable reference-year age. It does not mean independent factual verification of the producer's methodology.

Automatic collection. Controlled publication.

The application revalidates its shared statistical cache after six hours when requested. A daily scheduled check is configured to request the published API and archive its public snapshot for 90 days. Scheduling can be delayed; this is not a real-time feed. The visible collection date, not the schedule, is the evidence of a successful refresh.

Only four registered, read-only World Bank API queries are allowed. They contain no user profile information. All four series must pass schema, range, country, year, pagination and minimum-coverage checks before replacing the cached snapshot. A slow or unavailable source never removes the last validated snapshot: TGPI serves retained observations with their original collection date and marks the graph as degraded until a later refresh succeeds.

Each snapshot has a content revision, methodology version and per-series response digest. Raw response digests make changes detectable; they are not a guarantee that the source is correct. Statistical revisions can change historical values.

What remains outside automatic verification.

The new dossier summaries do not establish personal visa eligibility, professional licensing, admission, affordability, safety or tax obligations. They are not complete legal dossiers for all 195 countries. Local housing series, programme-by-programme accreditation and work-route coverage still need expansion. A linked government page does not certify eligibility. Legacy model-generated country budgets and ratings are not used in the new research core.

The same source-first standard must be applied as we expand to institutions, programs, cities, documents and multilingual content. Education from TGPI is independent: no affiliation, accreditation or endorsement by Harvard, the UN, the World Bank or another university is implied.

Attribution: World Bank, World Development Indicators, with upstream sources named in each indicator's metadata; United Nations Statistics Division, M49. Values are displayed with rounding; no missing values are imputed. Editorial summaries attribute their individual government, agency or university source. Source-specific reuse terms and third-party rights apply; the World Bank indicator license does not license every linked page.

Found an issue? Send the country, indicator, source URL and reference year — no personal documents — to contact@theglobalpolymath.com.

TGPI Learning / Evidence literacy / 1 of 3

Learn to challenge a number.

A short practice in interpretation, dates and missing evidence. Completion is a learning exercise, not an accredited qualification.

A country has 120% gross tertiary enrollment. What is the responsible interpretation?