Searches broadly across science, history, nature, technology and culture for candidate findings, favouring primary material over trivia lists.
Rejects the obvious, the over-shared and anything too close to a past edition. Remaining candidates are scored for novelty, credibility, surprise, story and visual potential.
Looks for the primary source and independent corroboration, then checks every link actually resolves. Exaggerated wording is corrected or the candidate is dropped.
Writes a headline no stronger than the evidence, a short explanation, a deeper layer, the research trail and the parameters for the day’s artifact.
One finding is scheduled for the next day. At midnight it is printed — and stays at a permanent address.
“Research paths” counts the searches made while looking. “Candidates” counts findings considered for the day. Scores used to rank candidates stay internal — they are an editorial tool, not a rating.
The launch editions were researched and verified by hand while the automated pipeline was being built.