We piloted Text Blaze in Chrome on a 1,200-record address cleanup and saw 18% faster entry. My auditor brain keeps asking: what happens if this scales — do error patterns shift (like a wrong city propagating across hundreds of rows), and how are you preserving per-field provenance for audits?
I’d tackle scale by adding a tiny ‘source’ cell per field — the snippet stamps TB:vX + method (e.g., ZIP→City) + timestamp, and it forces a quick confirm when the ZIP changes so the “wrong city propagating” bug dies early; belt-and-suspenders but it works. Do you version-tag your snippets or push a sidecar audit log to Sheets?