Corrections log

Every fix to labels, rules or item data, newest first. Email a correction.

  • 2026-10-16, Title artifacts: Fixed 'C omputers' (from feed markup) and '--' in arXiv titles.
  • 2026-10-16, Policy frameworks: Frameworks, guidance and reports are typed Report / framework, not Launch, so they are not flagged 'No evidence cited'.
  • 2026-10-16, Feedback, Inclusion, Policy tags: Feedback requires feedback to learners; Inclusion requires learner-focused terms; Policy no longer matches bare 'guidance' or 'regulators' and now matches 'governance'.
  • 2026-10-16, Promising flag: Now also matches survey, 'new research' and report findings, matching the legend. 'Study habits' and similar phrases no longer count as a study.
  • 2026-10-16, Carl Hendrick posts: Retyped from Study to Commentary. Study now only means a journal or preprint feed.
  • 2026-10-16, Adult ed / CTE audience tag: The CTE match is now a whole, case-sensitive word. It had been matching inside words such as 'detected' and 'selected'.
  • 2026-10-16, Journal listings: Author lists no longer cut off mid-name; long lists end in 'et al.'
  • 2026-10-16, Product Hunt summaries: Removed trailing 'Discussion | Link' text copied from the feed.
  • 2026-10-16, Item dates: Some feeds report a last-modified date. Dates now come from the article page's published date when available. Product Hunt items use the listing's published date.
  • 2026-10-16, Reading and writing tag: Stopped matching 'AI literacy', 'digital literacy' and similar phrases; the tag now requires reading or writing instruction terms.
  • 2026-10-16, Evidence labels replaced: The pre-launch flags (Research-backed, Promising, No evidence cited, Context) were retired and replaced by evidence levels (Level 1 to 3, Technical test, Commentary, No study cited, Not yet classified) under methodology v0.3. Earlier entries below use the retired names.
  • 2026-10-16, Back catalog added: 110 studies from a 283-study arXiv back catalog were added after blind grading by two AI graders and a delegated AI review; 173 off-topic studies were dropped. One school-choice study was lowered from Level 3 to Level 2 in final review.