why your cold outreach feels like ai slop (and how to fix it in 10 minutes)
there's a 426-point HN thread today called "AI slop is killing online communities."
the top comment is basically: we can tell. we can always tell.
cold outreach has the same problem.
what people actually mean when they say "personalized"
most founders treat personalization as a mail-merge field. {{first_name}}. maybe {{company}}. done.
that's not personalization. that's a slot machine with the same message behind every pull.
real personalization is noticing something specific and saying it plainly. no preamble. no throat-clearing. no "i came across your work and was blown away."
just: what you saw, why it matters to you, what you want.
three sentences. that's the whole formula.
the pattern that actually reads as human in 2026
i've been collecting cold DMs and emails that got responses. not templates — actual messages that worked. the pattern:
1. a reference only a human who paid attention would make
not "i saw your tweet." more like: "noticed you switched from monthly to annual pricing last month — curious what changed."
that's one sentence. it signals you're real.
2. a specific ask with a specific number
"do you have 20 minutes" is vague. "tuesday 3pm ET, 20 min, i'll send a loom first so you can decide if it's worth your time" is a real offer.
3. no signature block with three logos and a legal disclaimer
just your name. maybe a link. that's it.
why this matters more now than 6 months ago
inboxes are flooded with AI-generated outreach. everyone's tools got better at generating volume. so the signal-to-noise ratio collapsed.
the people still getting responses are the ones who sound like they typed the message themselves and meant it.
this is a rare case where doing less — fewer messages, slower, more specific — actually wins.
what "control flow" has to do with outreach
there's another thread today: agents need control flow, not more prompts.
the same idea applies to human outreach sequences. most people treat follow-up as "send message 2 on day 3, message 3 on day 7." linear. no branching.
but real sequences branch:
- did they open? did they click?
- did they reply with a question vs a no?
- did they visit your site after the first DM?
each of those is a different branch. a different next step.
most cold outreach fails not because the first message was bad but because the follow-up ignored every signal after it.
the actual 10-minute fix
pull your last 5 outreach messages. for each one, ask:
- does this reference something that couldn't be auto-generated?
- does the ask have a specific time, duration, or deliverable?
- is there a branch in my follow-up based on what they actually did?
if you answer no to any of those, rewrite before sending more volume.
the goal isn't to sound less like AI. it's to sound more like someone who actually did the homework.
if you want the exact message frameworks i use for cold DMs and emails (including the 3-sentence opener, the follow-up branching map, and 9 subject lines that don't sound like subject lines), i put them in a kit:
no upsell, no drip sequence after. just the doc.
yanmiayn.com — tools and playbooks for one-person businesses
Top comments (0)