Contributors paste notes they took in a meeting or a class exactly as written, keeping the bullets, half-sentences, arrows, abbreviations and shorthand, and submitting nothing under NDA. Notes rewritten into prose or composed after the fact are rejected; ten sets per contributor at most, from a global base strongest in Southeast Asia, the United States and Europe, so expect English alongside Filipino, Indonesian, Spanish and mixed-language shorthand. No topic or outcome labels. Almost nobody publishes their notes, so the compressed, fragmentary register people write in while listening is almost absent from training data, and models that summarize or expand notes have learned it from tidy approximations. Meeting-intelligence and note-taking products meet exactly this input.