Opportunity source research

Research date: 4 October 2026. Prepared by Open AI agent: Researcher.

Canonical project repositories: YouthOpp/data-pipeline, YouthOpp/youthopp.github.io and YouthOpp/.github. Dated fork workflow/release identifiers below record development-fork execution evidence, not canonical project identities.

Scoped access reviews include NASA, IKY and exact programme metadata; entries retain their own recorded review dates. The Spain renewal of 5 October 2026 updates three existing publishers without activating an adapter. See also the primary-source access review.

What this register means

This is a transparent discovery register, not a claim that every source has a working adapter or open application. USA and all EU27 countries are represented. Source country identifies the publisher, not applicant eligibility or opportunity destination. Original languages are retained.

Acquisition and quality policy

1. Prefer an official RSS/Atom feed or documented public API. A tested feed means an HTTP response was parsed into real items.

2. Otherwise inspect public HTML and publisher terms/robots; build a fixture-based adapter only after determining reliable list/detail selectors. Do not bypass blocks, logins or CAPTCHAs.

3. Publish concise factual excerpts with canonical source attribution; never copy whole articles. Source terms remain authoritative.

4. Never infer age, countries, funding or deadline when absent. Unknown dates are not "open".

5. Archive known expired calls; retain last good data on fetch failures and expose last successful check.

6. Distinguish publisher country, host country and eligible countries. Annual schemes and university guidance are not continuously refreshed catalogues.

Tested feed shortlist

All feed checks are dated 4 October 2026. Three newly tested commission feeds contain alumni, partnerships, webinars and other notices. Feed recency is not a currently open scholarship catalogue; selection and detail-page checks are required. This research documentation does not control adapter activation; the runtime manifest and verified release are authoritative. Scambieuropei and ONEK previously returned HTTP 403; no block workaround was attempted.

Reviewed HTML shortlist

Country register

Austria (AT)

Belgium (BE)

Bulgaria (BG)

Cyprus (CY)

Czechia (CZ)

Germany (DE)

Denmark (DK)

Estonia (EE)

Spain (ES)

Finland (FI)

France (FR)

International (GLOBAL)

Greece (GR)

Croatia (HR)

Hungary (HU)

Ireland (IE)

Italy (IT)

Lithuania (LT)

Luxembourg (LU)

Latvia (LV)

Malta (MT)

Netherlands (NL)

Poland (PL)

Portugal (PT)

Romania (RO)

Sweden (SE)

Slovenia (SI)

Slovakia (SK)

United States (US)

Evidence and reproducibility

source-registry.json records URL, country, operator, category, discovery date, check date, HTTP result, freshness notes and retrieval status for every entry. Recorded feed-review observations appear in the dated tables above; active collection health is published in each release collection report. Page checks sampled up to 150 KB with an identified review user-agent and 18-second timeout. Accessibility from the review environment can differ from GitHub Actions. Fixtures and automated adapter tests are required before enabling a source.

Feed access review — 4 October 2026

A blocked robots request is not permission and a guessed terms URL is not proof that no terms exist. robots rules describe crawl access, not copyright or licence. Feed integration tests establish technical compatibility only. The runtime source manifest is authoritative for which adapters are enabled; this research register never marks ingestion approved. Automated fetches should preserve minimal factual fields, never republish complete source articles, and provide an immediate correction/removal contact.

Country quality follow-up — 4 October 2026

The resumed pass incorporates the previously recorded Fulbright commission shortlist, freshly re-tests three feeds and reviews additional primary application pages. Research entries identify publishers, not applicant eligibility. Counts use distinct operators with dated or clearly cycle-specific opportunity evidence; primary application rounds in 2026 (including the 2025/26 cycle) and annual calls already closed are valid activity evidence and never treated as open applications. Conflicting source text is preserved as unresolved.

28 of 28 countries meet the editorial activity threshold; 0 remain below three. The register contains 110 entries, of which 94 have recent opportunity evidence. Of these, 93 are eligible evidence entries before country-level operator deduplication: the Croatian science ministry co-publishes the same bilateral call as AMPEU and is explicitly excluded despite being a distinct operator. No country has three operational, policy-reviewed automated adapters. Registry rights status after the Spain renewal: 100 pending, six scoped metadata access-policy reviews, three permission-required publishers (IKY, Fundación Carolina and la Caixa), and one reviewed manual-only publisher. A scoped review is not blanket publisher approval. Existing operational integrations remain nine; no Spain adapter is activated.

Application state distinctions

Recent programme evidence is separate from its application state. The new last_activity_evidence.application_state field records reviewed examples: open_confirmed, closed, some_calls_open, programme_specific_review, conditional_late_applications, conditional_funding_available, conflicting or conflicting_deadline. It is research metadata, not a runtime catalogue deadline calculation. Older entries without this field remain unclassified rather than implicitly open.

Confirmed current application coverage

The 3–5 active-source target is not established as currently open applications. After the scoped 5 October renewal, explicitly classified examples establish at least one open call or open-call subset for 22 distinct national publishers across 20 target countries, plus one GLOBAL publisher (Visegrad Fund): 23 publishers total. The Spain renewal adds only la Caixa as an explicitly confirmed open publisher; INJUVE remains unclassified for live availability and Carolina is closed. GLOBAL evidence does not establish national coverage. None of the 28 target countries has three confirmed current publishers in this conservative review. This is a lower bound from open_confirmed or some_calls_open; unclassified, conditional, conflicting and closed examples are excluded. A zero count means insufficient reviewed evidence, not proof that the country has no open opportunities. No country has three operational policy-reviewed adapters either.

Per-source reviewed activity

Explicit exclusions

General guidance, footer dates and unrelated institutional news are excluded from activity counts. Eurodesk Polska and FRSE share a publisher; grants.at/OeAD, Tempus programmes and SAIA programmes also share publishers. The Croatian ministry and AMPEU co-publish a bilateral call. The original RTU state-grant and KTU national-grant notices duplicate VIAA and Study in Lithuania. Their updated evidence instead documents independently administered RTU doctoral research grants and KTU Talent/Breakthrough scholarships. Their own programmes now count; the national re-listings still do not.

The Belgium/Luxembourg Fulbright commission is a single source, stored under Belgium, even though Luxembourg citizens can apply. Visegrad is regional and not repeated as a national publisher for every service country. No cross-country count is inferred from an opportunity destination.

Germany Fulbright has contradictory open/closed text on the same current student page; no open application is asserted. Austria lists an October 7 closing date while national US Fulbright pages use October 6; the discrepancy remains explicit. Swedish Institute portal notice contains mixed academic-year labels; no new date is inferred. Luxembourg scholarship evidence was retrieved from the primary publisher search index while direct page content was challenged; this is editorial evidence only.

Fulbright metadata adapter follow-up — 4 October 2026

Access, reuse language and opportunity relevance were reviewed separately. Fresh HTTP requests returned 200 for the three official homepages, robots files and RSS feeds. All three homepages advertise their exact feed through rel="alternate" with RSS content type. Each feed contains ten items. Hourly syndication metadata describes feed updates and does not establish a reuse licence. No feed channel/item copyright, rights or licence element was found.

No express reuse licence or publisher approval is claimed. Absence of an express licence is an uncertainty, not invented permission and not an automatic requirement for new user approval. Ordinary copyright reservations alone are not treated here as an express title/link indexing ban. This is an engineering acquisition-policy assessment limited to the inspected pages, not a blanket legal conclusion. Source terms remain authoritative.

The Czech allowlist contains only these exact canonical item URLs:

Preserve original Czech title, canonical original link and valid publication date. Set summary to an empty string, deadline to null, availability to unknown and host/eligible country arrays to empty. Do not publish article prose or images. The generic Novinky category and broad scholarship keywords also match excluded webinar/alumni content; they cannot determine selection. Unreviewed items remain excluded until primary-content review expands the exact allowlist. Sanitized fixtures retain title/link/date/categories only.

Czech runtime source fulbright-czech-programmes, adapter reviewed-rss, is operational. Hosted main run 37263518031 (fork verification) passed 19 tests, durable-state recovery and collection; immutable release catalog-37263518031-1 (fork verification) was published on 5 October 2026 at 04:26:39 UTC with 32 total records, including the two reviewed Czech metadata records, and all four sources healthy. This verifies data collection and publication; hosted website acceptance is separate. No publisher-contact or owner-approval action is presently required by documented policy for this scoped metadata operation. Full descriptions/images, newly restrictive terms or blocked access would require a separate acquisition decision. NL/PT currently have a relevance blocker, not missing consent. Poll conservatively with a clear project User-Agent, honor rate limits and keep at least ten seconds between Netherlands same-host requests.

Remaining implementation gaps

The user’s target is 3–5 high-quality, active researched sources per country where genuinely verifiable. The register supports at least three distinct publishers with dated editorial activity in all 28 countries; this includes closed annual calls and is not equivalent to three currently open application publishers. The conservative confirmed-open coverage gaps remain explicit above. Adapter counts are a separate implementation measure: all 28 countries remain below three independently tested, operational, policy-reviewed automated adapters with health reporting. The researched-source target does not require three adapters per country, nor does it require enabling every candidate in this discovery register.

Exact Portugal programme-page selection — 5 October 2026

The separately reviewed master's programme now has a trusted exact canonical/Open Graph title/link adapter; this does not activate the unrelated Portugal RSS. The dated access review records GET200, robots scope, privacy-only policy text, all-rights-reserved footer and the absence of express reuse permission. This scoped project assessment covers factual metadata only, not source prose or publisher approval. Original publication is unavailable and stays null; the observed modification timestamp is never publication. The record stays a scholarship programme overview with unknown application state, deadline and eligibility. Current window evidence is separate research evidence. Actual producer run 37270813364 (fork verification) and immutable release catalog-37270813364-1 (fork verification) verify the Portugal metadata record, with all six integrated sources healthy at 06:07:29 UTC. This is dated production evidence; current source health comes from each later release. Website acceptance remains separate.

Exact Netherlands programme-page selection — 5 October 2026

The doctoral programme passed a fresh scoped title/link access and identity review. Current direct HTML provides matching canonical/Open Graph URL and one original title; publication stays null. Use the existing exact Open Graph adapter with the same programme-overview/unknown-application constraints as Portugal; see dated access review. The Netherlands news feed remains excluded for relevance. One exact programme adapter does not establish three operational publishers per country, unrestricted youth eligibility or publisher approval. Producer verification is recorded below.

Research and operational state

The 4 October country research snapshot and later recorded access dates describe editorial evidence. acquisition_state: connected is separately supported by a dated producer run/release, not inferred from a reachable page or local test. Czech, NASA and Portugal metadata have such evidence in run 37270813364 (fork verification) and release catalog-37270813364-1 (fork verification). The runtime source directory joins each adapter to the research source and reports the latest collection health; a later fetch failure preserves records and does not silently rewrite this historical verification. Operational access review is not express publisher permission, and editorial application-window evidence is not catalogue availability.

Finite follow-up scope and ongoing backlog

The named bounded follow-ups from this research stage are the reviewed Czech feed selection, NASA overview and exact Portugal/Netherlands programme pages. Each needs a verified producer release before an operational claim; the Netherlands producer release is now independently verified below. IKY remains disabled because its publisher terms explicitly require prior permission for public republication. This project has not requested or obtained that permission.

Further HTML adapters across the remaining discovery candidates, additional exact calls, and periodic evidence renewal are ongoing maintenance/backlog choices. They are not a claim that three open calls or three adapters now exist in every country, and they are not an indefinite gate requiring all 110 candidate publishers to be integrated. Newly restrictive terms, access challenges and contradictory application dates remain explicit limits rather than fabricated coverage.

Netherlands operational verification — 5 October 2026, 06:29 UTC

The actual merged-main producer run 37272654737 (fork verification) passed 44 tests and published immutable release catalog-37272654737-1 (fork verification) at 06:29:36 UTC. The downloaded Actions artifact and all three extracted assets match GitHub artifact/release digests and the integrity manifest. The catalog has 37 retained records, with all seven source collections healthy in this dated release; current collection health is established separately by each later run.

The exact Netherlands record 1f4def960c9cca6ae7111473 preserves the original title/link, null publication, empty summary, scholarships/programme-overview, unknown application status/deadline and empty host/eligible countries. NL is publisher geography only. Its exact research join and category/kind memberships were verified. Catalog size is 324,910 bytes and SHA256 is 8c93f5d91fb41e2c0eeb4bcd12f7ec2afd70240b4db457388040b30c260e9616. This is measured producer collection/publication; website consumption and public hosted acceptance remain separate. The general Netherlands RSS remains excluded for relevance, and no publisher approval or currently open youth call is inferred.

Scoped country evidence renewal — 5 October 2026

See the exact primary-page review. Bulgarian research grants and Malta maintenance grants have explicit current open evidence. Vienna homepage invitations conflict with detail-page application-closed labels; ONEK Open badges conflict with the expired September 16 call and December 10 next opening. Both remain excluded from confirmed-open counts. Malta footer-linked terms prohibit automated scraping, so it remains a manual source-link directory entry. Research Ireland nominations are institutional and subject to restrictive content-publication terms; no adapter or registry activation is implied.

Generic freshness text for recorded dated opportunity evidence now mirrors the existing evidence with its original review date; this is prose reconciliation, not a fresh whole-register verification. The refreshed open-evidence lower bound is 21 national publishers across 19 target countries plus one GLOBAL publisher (Visegrad Fund), 22 publishers total, including two newly classified publishers and previously classified entries. Research-source counts and operational adapter counts remain separate.

Luxembourg exact metadata production verification — 5 October 2026

Merged-main fork producer run 37274664011 passed 48 tests and published immutable release catalog-37274664011-1 at 06:53:16 UTC. Downloaded artifact/native asset hashes matched the manifest. The catalog retained 39 records across eight integrations: seven source collections were healthy, including one FNR exact record; Scholarships Corner reported a fetch error and retained ten last-good records. This is dated producer evidence, not blanket current health.

FNR record 8d24597c237a1ff602abeaed preserves AFdoc-FNR, original publication 2026-07-29T13:25:54.000Z, empty summary, unknown application availability, null deadline and empty host/eligible countries. No automated eligibility or application-window inference. Website acceptance is separate.

Browsing collected records by country

The Countries menu opens /opportunities/countries/, with destination collections and publisher-country collections. /opportunities/from/<country>/ displays actual collected opportunity/programme rows; /sources/countries/<country>/ remains the separate publisher directory. Country options open complete static country collections, including records from later pagination pages. Category pages retain the category when changing publisher country.

Country provenance is explicit: publisher country comes from the registry or normalized record; destination comes from the record's host_countries, supported by the original title or reviewed programme evidence. Publisher country never fills unknown destination or eligibility. /collection/ reports source collection errors and distinguishes classified records from unclassified metadata. build-report.json and the Actions post-build check verify that catalog records and country collection pages were generated.