Three emails, one thread. . Email 2 and 3 reply in the same thread, so the subject stays.
Merge fields are highlighted. {{observed_fact}} is one verifiable thing about them from their site, LinkedIn or a job ad; if there is none, drop the line. {{ps_fact}} is a second, different fact for the P.S. of email 2: a named project or page of theirs, AI-filled per contact. {{lawful_touch}} names the inbound request, LinkedIn acceptance, trade-fair meeting or partner intro that makes the email lawful in consent countries. The warm sequence adds {{new_fact}}, {{shipped_since}} and {{calendar_fact}}, all taken from the register's calendar and the release notes.
Two subject lines on every email 1. Send A and B to 100 to 150 people each and the winner to the rest. Emails 2 and 3 are real replies in the thread, never a typed "Re:".
Non-English emails have an EN toggle on the card that shows the English version for review.
Built on lists the finding IDs in business/gtm/pain-research-2026-10-21/ that each fact traces to. Nothing in any email is unverified.
Never say is the per-country do-not-claim list from the weakness register. It applies to replies and demos too.
Not now
What the corpus says to build around the emails
Read after the sequences were written: the full cold-email articles, the LinkedIn playbook, the Clearscope outreach story and sixty tactic records. These are the moves that change the result more than any line of copy.