Shelf
Catalog Newsletter
Starter · Outreach

Cold-to-Reply

The 2-minute fix for cold outreach that doesn't sound like AI. Drop in your offer, the prospect's specific signal, and a few of your own writing samples. Get back a multi-touch sequence calibrated to that exact prospect, with reply handlers and a built-in spam-pattern check.

Interactive browser tool Works with ChatGPT, Claude, Gemini Instant download · no setup
$18
One-time. Lifetime updates. 7-day refund if it doesn't work as described.
Buy

Powered by Gumroad

Already own this? Access your purchases at gumroad.com/library.

Who This Is For

Not For

See It Work

Drop in what you're selling and who you're reaching out to:

Selling: Northwind cost-monitor for Postgres-heavy AWS stacks
What it does: cuts AWS bills 40% on average, two-week setup
Pain points: engineering teams spending 6+ hrs/week on manual cost reviews

Sender: Pat Chen, Founder at Northwind

Recipient: Sam Rivera, VP Engineering, Acme SaaS
Signal: Just raised Series B (announced last week)
Knowledge: Sam wrote a LinkedIn post in Q1 about Postgres scaling pain at 3am

Plus 1–3 short samples of your own writing:

Sample 1 (email to a coworker):
Hey Pat, the strategy doc isn't useful in its current shape. Three layers of generic. Section 2 says "optimize engagement" six times without one example. Can we cut to one page and put the actual moves in?

Sample 2 (LinkedIn post):
Three things I learned running a 6-month vendor selection: 1) The cheapest option costs the most. 2) The fastest sales cycle hides the worst onboarding. 3) Trust the references that DON'T sound rehearsed.

Click Build my outreach prompt, paste into your AI, get back a 4-touch sequence calibrated to that exact prospect:

Touch 1 — Email
Subject: Series B and the Postgres bill

Hey Sam,

Caught your Q1 post on the 3am compaction pain. The pattern you described usually doesn't get cheaper after a Series B. It gets noisier: more traffic, more migration churn, more uncertainty about which line on the AWS bill is "normal."

Northwind sits in front of Postgres-heavy stacks and answers "is this normal" in real time. Two-week setup. Cuts about 40% of the spend by killing the queries you forgot existed.

Curious what you're using right now (or if you're in the "we'll-figure-it-out-after-this-quarter" phase). Either way, helpful context.

Pat
Founder, Northwind
Touch 2 — Email (4–7 days after touch 1)
Subject: Two Postgres patterns we see at Series B

Hey Sam, two specifics from the last six Series B engineering teams we worked with:

First, the IO-bound query that worked fine at 50K MAU breaks at 200K. Usually it's a join that nobody flagged because it never showed up in slow-query logs at the old volume. Northwind catches these in week one.

Second, the post-Series-B hire wave doubles the rate of accidental long-running migrations. We auto-flag any migration that touches more than X% of a table.

If either rings true, happy to share the full pattern doc.

Pat
+ Touch 3 (different angle), Touch 4 (dignified break-up), reply-handler kit for 5 reply types, optional LinkedIn variant of touch 1.

Then run the optional ship-check on any draft for an independent verdict before sending:

Ship-check verdict
Verdict

Ship as-is.

Voice match
  • Sentence-length rhythm matches sample 1's mix of short and medium
  • Specific-number pattern ("40%," "two-week," "Q1") matches the writing samples
  • No corporate-warm filler ("hope you're doing well," "circling back")
AI tells flagged
  • Zero em-dashes
  • Zero hedging words ("perhaps," "potentially," "I'd love to")
  • Zero generic transitions ("furthermore," "ultimately," "additionally")

How To Use It

Step 01
Open the tool

Double-click the HTML file. It opens in any browser. No install, no login, no account.

Step 02
Fill five sections

Product, sender details, recipient + signal, 1–3 writing samples, sequence shape (length, tone, optional outputs). About 2 minutes.

Step 03
Build and copy

Click Build my outreach prompt. Copy the output, or use the one-click links to open ChatGPT, Claude, or Gemini.

Step 04
Paste into your AI

Works in ChatGPT, Claude, or Gemini. Multi-touch sequence comes back in 30–60 seconds, each piece labeled and paste-ready.

How it works under the hood

The tool runs entirely in your browser. Your writing samples and prospect details never leave your computer until you paste them into your AI of choice. No Shelf server in the middle. No account, no API key, no data collected. Download once, use forever.

Straight Talk

What it does well. Produces sequences that sound like you wrote them, calibrated to a specific prospect's signal. Each touch in the sequence delivers a new reason to reply, not the usual "just following up" filler. The bonus ship-check catches AI-spam patterns you'd miss on your own.

What it does reasonably well. Building a base for moderate-volume outreach (run several prospects through, edit each lightly before sending). Always worth a human pass per email, especially around facts and names.

What it can't do. Guarantee replies. Replace sending infrastructure if you're sending 1,000+ emails a day. Replace having something genuinely useful to offer the recipient. Calibrated outreach replies measurably more than generic; that's the limit.

How we make it. AI drafts. A human reviews and edits everything before it ships. Built for ChatGPT, Claude, and Gemini. Shipped with a 19-check internal quality rubric. If it doesn't work as described, use the 7-day refund, no questions.

$18
One-time. Lifetime updates. 7-day refund if it doesn't work as described.
Buy

Powered by Gumroad

Already own this? Access your purchases at gumroad.com/library.