The PRINT Stack — Claude × LinkedIn (2026)
▸ The PRINT Stack — Issue No. 14

Stop prompting Claude for DMs. Build the stack.

You commented PRINT. Here it is. The full 2026 stack: tiered signals, the right Claude for the job, and the ABM loop that books calls. No fluff. Steal it.

The prompt is the easy part. The reason 99% of "AI LinkedIn outreach" sounds like AI is the inputs are wrong and the outputs aren't matched to volume. Three layers. Skip one, the whole thing breaks.

Layer 1 — The signals stack

Not all signals are equal. Most people drown in Tier 3 noise and wonder why their reply rate is 3%. Stack by intent strength. Top tier first.

TIER 1 Active intent — the gold

They're literally researching a solution right now. Comments on competitor posts. "Anyone use X for Y?" questions in groups. Public pain rants ("our current tool is killing us"). RFP-style asks.

If you can find a Tier 1 signal, you barely need a prompt. The message writes itself: "Saw your comment about [competitor]. Had the same problem last year. Here's what worked."

DATA FROM: Clay Signals · Common Room · LinkedIn Sales Nav saved searches · Reddit/Slack/X mentions
TIER 2 Trigger events — warm

Something changed in their world that creates an opening. Job changes (new role = new budget = re-evaluating vendors). Team expansion in your buyer function. Public "we're moving off X this quarter" posts. New tool installs or uninstalls.

DATA FROM: LinkedIn Sales Nav · Apollo · Clay job-change triggers · Crustdata · BuiltWith
TIER 3 Surface signals — table stakes

Recent posts. Funding rounds. Hiring. Useful for color and context but not enough to drive a reply on their own. Everyone has access to them. Reference them after Tier 1 or 2, never as the lead.

DATA FROM: LinkedIn profile · Crunchbase · company blog · X feed

The reply rate gap between Tier 1 and Tier 3 outreach is roughly 8x. Most people work Tier 3 and blame the prompt.

Layer 2 — Pick the right Claude

Same brain. Three hands. The mistake is picking the wrong one for your volume.

Claude
web / mobile
Manual mode. Create a Project, upload your context, paste leads, run the prompt, review, send. You stay in control of every message. PICK WHEN: under 50 DMs/week, you want eyes on each one
Cowork
desktop app
Agent mode. Claude opens a browser, mines leads, pulls signals, writes DMs, queues them. Sandboxed VM. 132 pre-built skills. Zero terminal needed. PICK WHEN: 50-200 DMs/week, you want speed without setup
Code
terminal / cli
Full automation. MCP servers (Clay, Apollo, LinkedIn). Custom skills. Scheduled runs. Pipe data wherever. The system you build once and resell. PICK WHEN: 200+ DMs/week or building this for clients

I run all three. Web for low-volume founder outreach. Cowork for the ABM loop below. Code for the system I built for an agency client that runs nightly.

Layer 3 — The ABM loop prompt

This is what I paste into a fresh Cowork session. 15 target accounts in, 5 booked calls out. The whole thing runs in about 90 minutes.

cowork / abm-loop.md
# ABM loop — 15 accounts to booked calls
CONTEXT: Read icp.md, offer.md, tone.md, winners.md.
INPUT:
15 target company LinkedIn URLs (pasted below).
STEP 1 — Mine decision-makers:
For each company, visit the People tab.
Find 3-4 contacts matching icp.md role criteria.
Output: spreadsheet (name, title, profile URL).
STEP 2 — Score signals for each contact:
Tier 1: recent comments on competitor posts,
"anyone use X" asks, public pain posts.
Tier 2: job change in last 90d, team expansion,
"moving off X" posts.
Tier 3: recent posts, funding (context only).
Tag each contact with highest tier found.
STEP 3 — Draft outreach:
For each contact, write 1 connection note.
Lead with their highest-tier signal (verbatim).
Max 280 chars. No pitch. End with a question.
Match the voice in tone.md exactly.
OUTPUT:
Final spreadsheet: name, title, signal tier,
signal text, drafted DM, send/skip rec.
Recommend SKIP if no Tier 1 or 2 signal found.

The "skip if no Tier 1 or 2" rule is the unlock. It keeps you on a 15-account list instead of bloating to 200 weak DMs. Quality compounds. Every reply trains your winners.md and every skipped account is one you can warm up with content instead.

The 4 docs that power it

The prompt above assumes you have these. If you don't, build them tonight. 60 minutes total.

icp.md — your ideal client

Role, company size, industry, tools they use, what keeps them up at 2am. Specific roles, not "decision-makers."

offer.md — what you sell

One-sentence outcome. Three case-study one-liners with real names and numbers. The price (or "starts at"). Claude never pitches it, but knowing it shapes the tone.

tone.md — how you actually talk

Paste 5-10 messages you've written that sound like you. Texts to friends are fine. Tell Claude: "match this voice. Casual but precise. Specific over generic. No corporate-speak."

winners.md — your 10 best DMs

The 10 messages that booked calls or got real replies. Don't have 10? Use 5. Don't have 5? Use the four DM examples in last week's issue as starters and replace them as you collect your own.

Want the 4 doc templates?
Reply to this email with "DOCS" and I'll send the fill-in-the-blanks Notion templates I use with every client.