Data Explorer

AI-generated source note: CIATECH Africa, @clearglobal.org and Lake Chad Basin contributors. Approval splits are not train/test sets. Use canonical transcripts and check readback_candidates.tsv before treating Freeform as spontaneous speech. Non-commercial license; no biometric profiling.

AI-generated source note: World Bank Entrepreneurship Database. Indicator page: CC BY 4.0. Covers registered limited-liability firms, not all entrepreneurs or informal activity. Check country/year coverage and registry changes before ranking economies.

Credit: Matthew Blissett, @gbif.org and data publishers. Experimental layouts from 1 Aug 2026: https://doi.org/10.15468/dl.8yrbe7 Check occurrence licenses (CC0/CC BY/CC BY-NC) and media rights separately. Occurrence counts are not abundance.

Credit: The Squirrel Census via NYC Open Data. Counted foraging=true in all 3,023 CSV rows; no exclusions. Activity flags can overlap. NYC portal terms apply; metadata lists no separate CC license. Sightings do not identify distinct animals.

Source: IMDb. Personal/non-commercial use only, subject to its terms, not an unrestricted open-data license. Join titles on tconst, treat \N as missing, and keep numVotes when comparing weighted rating averages. Ratings are not a representative audience survey.

Credit: @ember-energy.org and its national/international source providers. The July 2026 CSV redesign puts metrics in columns and flags aggregate sources. Avoid adding Clean or Fossil totals to their component fuels; compare matching months and save the extraction date.

Credit: OECD, WHO and the European Commission developed the health-accounts framework. @oecdsocial.bsky.social: compare the same spending boundary and year; current health expenditure excludes investment. Keep GDP shares, financing shares and per-capita amounts distinct.

Credit: Bandarkar et al. / Meta AI, using FLORES-200 passages. Belebele is intended for testing, not training or validation. Match questions by passage link + split + question_number; keep prompting language and answer parsing fixed for cross-language comparisons.

Sources: Kpler; UN Global Platform; IMF PortWatch. These are maritime volume proxies, not customs trade values. AIS gaps, spoofing and coverage changes matter. Save the extraction date: series are revised and daily estimates are updated weekly, not live.

Data collected by Kristen Gorman and Palmer Station LTER; package by @allisonhorst.bsky.social, Alison Hill and Gorman. The sample covers 2007-2009, not all penguins. Keep species, sex and island visible when exploring size relationships.

Credit: @walthickey.bsky.social / FiveThirtyEight. These are manually coded presence/absence tags, not object counts. Guest episodes are included, so filter GUEST before attributing patterns to Bob Ross. A useful lesson in choosing the denominator.

Credit: @statsbomb.com. Coverage is selected competitions/seasons, not all football. Check competitions.json and 360 availability before comparing teams. The repository asks published analyses to credit StatsBomb and use its logo; check the terms before reuse.

Built by @mike.teczno.com for Wikimedia Enterprise, using volunteer-maintained sources. Snapshot: 15 Sep 2026. Start with links.parquet, not a giant flat join. Licenses differ by table: CC0, ODbL and CC BY-SA. Preserve source attribution.

Credit to @ecdc.europa.eu and national surveillance systems behind the Atlas. For analysis, keep indicator definitions, case definitions, reporting completeness and country/year coverage visible. Surveillance data is not the same thing as true incidence.

Methodology note: LHTB intentionally withholds tests/solutions to reduce contamination. Reuse should report Harbor version, continue-until-timeout handling, task timeout, resources and whether scoring used the official hidden verifier loop.

Credit: TiVA is built by OECD from its Inter-Country Input-Output tables. For analysis, keep edition/year, current-price USD vs shares, industry concordance and ICIO balancing/gap-filling caveats visible. @oecdtrade.bsky.social

Credit to @artinstitutechi.bsky.social. One caveat for reuse: keep API vs nightly-dump version, public-domain/CC0 image status, IIIF availability and source field completeness visible. Otherwise collection coverage can look cleaner than the underlying cataloging history.

Source is DataDawn / OpenRegulations; repo says records are from official government sources and public domain. Methodology caveat: derived matches between commenters, lobbyists, bills and firms should keep source, match rule, confidence and refresh date visible.