Data (L1)

PII-safe dataset cut — author handles anonymized. Raw handles stay internal at DP.

• emi-jay-voc-l1.zip — corpus (3,356 records, JSONL) + study config + coverage log + Wave-0 files (~1 MB)
• corpus_pii_safe.jsonl — browse the corpus directly

Record shape: {source, kind, id, text, created_iso, url, community, score, rating, brand, ...}