Newsletter Stack: Claude, Perplexity, and Substack

This stack is for a solo newsletter operator doing weekly sends to ~10K subscribers on a $40/month tools budget.

Hands-on time per issue: 3 hours. Research runs on Perplexity Max, drafting and editing on Claude Pro, publishing on Substack.

System Verdict

Pick this stack for a weekly 1,200-word research newsletter with citations. Perplexity owns discovery, Claude owns drafting and fact-checking in three separate chats, Substack owns delivery.

Skip it for daily cadence (Perplexity credits cap out) or visual-heavy formats (no image generation in scope). Substitute Beehiiv or Ghost if Substack monetization is blocking.

Total cost: $40/month. Swap Claude Pro for Max 5x at $100/mo only if drafting volume exceeds ~8 issues/month.

Key Facts

The short version

• 30 minutes on Perplexity Max for research. Cited synthesis across 19 models including Opus 4.7 and OpenAI frontier models.
• 20 minutes on Claude Pro for a 1,200-word draft. 1 hour edit. 30 minutes fact-check.
• 40 minutes pasting into Substack, adding links, scheduling.
• Hands-on time replaced: roughly 10 hours/week of manual research and copy work.

The stack

Perplexity Max ($20/mo)

Owns research. Queries like “latest on AI agent workflows April 2026” return synthesized answers with inline citations in about 2 minutes.

Deep Research mode handles multi-step queries better than a browser with tabs open. Typical output is an 800-word cited report.

Claude Pro ($20/mo)

Owns writing and checking. Opus 4.7 drafts from outlines, edits for voice, and fact-checks against sources.

Use three distinct chats per post: one for draft, one for edit, one for fact-check. A single chat mixes the roles and bloats the output.

Substack (free)

Owns publishing. Paste Markdown, embed links, schedule. No learning curve.

Lacks native AI integration, but the workflow does not need it. Analytics on the reference channel show ~45%+ open rates.

The workflow, step by step

1. Sunday 9 AM. Research (10 min). Open Perplexity Max. Prompt: “Research top 5 AI workflow tools for solo operators in April 2026. Include pricing, versions like Claude Opus 4.7 and OpenAI frontier models, and real user savings from sources post-Feb 2026. Synthesize into 800 words with citations. Focus on newsletters.” Save 600 words of output to /Newsletter/2026-W15/Research.txt.

2. Draft (15 min). New Claude Pro chat labeled “Draft-W15”. Paste the Perplexity research. Prompt: “Write a 1,200-word newsletter draft on AI workflows for solo operators. Structure: hook, 3 examples with savings, editorial take, call to action. Voice: blunt, specific, no hype. Base only on pasted research.” Save to /Newsletter/2026-W15/Draft.md.

3. Edit notes (5 min). Read draft aloud. Log issues to /Newsletter/2026-W15/Edits.txt. Example from the prior issue: “Too many lists; tighten to 2 examples. Voice drifts salesy in para 4.”

4. Edit pass (45 min). New Claude chat “Edit-W15”. Paste draft and edit notes. Prompt: “Edit this draft: short sentences, concrete examples, cut 200 words, fix salesy tone. Output only the revised draft.” Write back to Draft.md.

5. Fact-check (25 min). New Claude chat “Fact-Check-W15”. Paste final draft and original Perplexity research. Prompt: “Fact-check every claim against the research. List errors as bullets with quoted source text. Suggest fixes. Flag uncited inferences.” Apply fixes manually.

6. Publish (25 min). Open Substack new post. Paste final Draft.md (Markdown works). Bold key terms, convert Perplexity citations to inline links, add one stock image. Preview, schedule Monday 8 AM.

7. Archive. Zip /Newsletter/2026-W15/ to Drive. Log time in Notion.

Where it breaks

Opus 4.7 drafts repeat Perplexity phrasing verbatim if the prompt omits “use your own words.” First run without that guardrail sounded copied.

Perplexity Max cites outdated info pre-April 2026 if the query is not date-scoped. Prior issue pulled February announcements that had since been superseded.

Substack Markdown strips nested lists occasionally. Flatten them or paste twice.

Claude fact-check flags editorial opinions as “missing citations.” Prefix the prompt with “ignore first-person opinions” to silence that.

Monthly cost

Prices verified 2026-04-17 via Perplexity pricing and claude.com/pricing.

Who this is for

Copy this stack for a weekly solo newsletter, ~1,200 words per issue, research-backed, no team.

Skip it for daily cadence (Perplexity credit caps bite at scale) or visual-heavy formats (no image gen in the loop).

FAQ

How is edit history tracked? Three Claude chats per post plus Draft.md git commits in the issue folder. Typical word counts: draft 1,480, edit 1,200, final 1,205.

Why not Perplexity Pro’s Deep Research at $200/mo? It executes multi-step agentic research but is overkill for newsletters. Voice control matters more than automation depth at this cadence.

Claude vs OpenAI frontier models for drafting? Opus 4.7 edits with more nuance; OpenAI’s frontier stack is wordier on the same prompt. Claude wins blind draft tests ~8 out of 10 runs on this stack.

Substack open rate for reference? The reference channel averages ~47% over the past three months. Research-led hooks from Perplexity lifted it from a ~32% baseline.

System Notes

This page documents an operational stack verified by the aipedia.wiki editorial pipeline. Last verified 2026-04-17.

Related

• Tools: Claude · Perplexity
• Workflows: Research Assistant Stack · YouTube Content Stack