ColdOpen: an open-weight model that learns how you write, then drafts the email you'd actually send
This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend What I Built I built ColdOpen for my sister, Aishwarya. She's a software engineer with 5+ years of experience, and she's good at her job. She still puts off one task for weeks: writing the first message to someone she doesn't know. That could be a referral request, a note to a recruiter, or a question to someone who…
Aishwarya, a software engineer with over five years of experience, faced a common challenge when reaching out to people she didn't know. She would spend hours rewriting the first line of an email before finally closing the tab. Existing AI email tools produced polished drafts that lacked her personal touch, so she never sent them.
ColdOpen, built for her, starts from her own writing rather than using a template. Users paste 2-3 messages they have actually sent, and specify their intent (referral, recruiter intro, alumni intro, resume feedback, mock interview, mentoring, follow-up, thank-you, reconnect, reference, switching roles). ColdOpen generates a short "voice learned" note (greeting, sign-off, sentence length, habits) for review.
It then creates three emails and three subject lines, presented in a side-by-side comparison with non-voice drafts. Green highlights show her original phrasing, pink marks generic template phrases, and amber indicates placeholders that need to be filled before sending. The demo is live at https://coldopen-fe.onrender.com/ and the code is available on GitHub.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
