The one-person newsletter operating system.

• 01 Beehiiv — host, segments, analytics daily
• 02 Claude Sonnet 4.6 — drafts & edits daily
• 03 Notion — ideas vault weekly
• 04 Granola — interview & call notes weekly
• 05 Stripe Payment Links — sponsor & product placements monthly

Two daily tools. Three that surface at the right time. That's the whole shelf. The rest of this guide is how we use them.

1. 01vault · 5 min

   Ideas vault — raw, not written.

   Open Notion. The vault is one database, four columns: title (one line), source link, why-now, category. Two filters on top: this week and shelf for later.

   The rule: if you can't write the why-now in one sentence — why this matters this week, not next month — it goes to the shelf. Don't fight that rule. The shelf is where 80% of ideas live and that's the right ratio.

   Drop three to five entries a week. You'll use one. The other two to four are how you stay sharp on what's not the angle.

2. 02angle · 10 min

   Pick the angle — headline first.

   Look at the vault. Pick one. Not the available one — The right one.

   Write the headline before you write anything else. Pattern: noun + verb + sharp claim. Avoid questions. Avoid "how to" unless the how is genuinely unusual. Avoid colons — they're a tell that you couldn't decide between two ideas.

   Three tries. If none land, the angle isn't ready — close the doc, come back to the vault later. Discard the two losers; don't archive them. Saving losing headlines is how you write four mediocre issues in a row.

3. 03draft · 30 min

   Draft with Claude — pressure-test, don't replace.

   The setup: voice file open, tight outline pasted, rough first draft from you. Don't ask Claude to write the issue. Ask it to put pressure on the writing you've already done.

   Three prompts we rotate, in order:

   1. "Find the three sentences in this draft that don't earn their place. Quote them and explain why."
   2. "What's the strongest claim I'm making that isn't backed by a number, a date, or a name? What backing would close it?"
   3. "Where am I using an adjective when a number would do? Flag each one with the swap."

   Never paste the whole issue at once. Paste sections of 400 words. The model is sharper on a section than on a full draft. Run all three prompts on each section. Takes ten minutes. Worth it.

   The first ~70% should still come from you. Claude is editing pressure, not first-draft authorship. Readers can tell the difference in two sentences — the giveaway is rhythm: AI drafts have evenly-weighted sentences. Yours shouldn't.

4. 04edit · 25 min

   Real edit — out loud.

   Read it out loud. Every word. If your throat catches on a sentence, the reader's eye will too — cut it or rewrite it.

   Then four passes, in order:

   1. Adverbs. Search the doc for words ending in ly. Most come up empty after you delete them — that's the point.
   2. Adjectives. Replace each one with a number where you can. "Massive growth" becomes "3.4× in six months" or it gets cut.
   3. Italic budget. One italic phrase per section, maximum. More than that and the prose looks anxious.
   4. Summary cut. If a sentence summarizes the sentence right after it, delete the summary. This rule alone trims 10–15% of a typical draft.

   Twenty-five minutes is enough. If you go longer, you're not editing — you're rewriting. Stop and ship.

5. 05send · 10 min

   Ship through Beehiiv — same time, every week.

   Same send time every week. Mine is 8:05am CT, Fridays. Five past the hour reads less like an alarm than 8:00 sharp. Consistency beats clever.

   Two segments only: free and paid. Not eight. Beehiiv defaults give you more — ignore them. You'll know when you need a third segment, and that week is not this one.

   Subject lines: declarative, under 55 characters, no emoji, no questions. The headline you spent ten minutes on earlier usually works without modification. Preview text gets the second-best sentence in the draft, not a recap of the subject.

   Monetization: Stripe Payment Links sit in the footer, never inline. Inline ads break the read and convert worse anyway. Rotate them monthly so the same paid placement isn't sitting on the same readers six weeks in a row.

   One excerpt tweet scheduled at the same minute as the send. Skip every other platform until one of them moves the list — most never do.

6. 06reply · 10 min

   Read replies — they're next week.

   Open the inbox sixty minutes after send. Read replies for ten minutes, no more. Set a timer if you have to.

   The good ones go straight into Notion — title, link, why-now. Half of next week's vault populates itself this way. The other half comes from your own reading.

   Reply to three personally. Not a thank-you — a one-sentence reaction to what they actually said. This is what makes readers feel like they're talking to a person and not a mailing list. It also quietly improves your sender reputation, which Beehiiv uses to decide whose inbox you reach next week.

• ×AI-written first drafts. They sound like everyone else's first drafts. Readers notice in two sentences.
• ×Automated subject-line A/B tests. Send time matters more. Your fifth subject line was already fine.
• ×Subject-line emoji. They look like marketing, the inbox sorts them harder, and they age badly. Skip.
• ×"Personalization at scale." If it's at scale, it isn't personalization. It's a merge field.
• ×Eight segments. Two is the right number until you have staff. Free and paid.
• ×Seven-email welcome series. The first one matters. The rest are noise. Write one good welcome, ship it, move on.
• ×Cross-posting everywhere. One excerpt, one platform, same minute as the send. The rest is noise tax.
• ×"Optimizing" the publish day. Pick a day, pick a time, hold both for six months before you touch them. Switching is what kills habits — yours and the reader's.

A newsletter operating system isn't a workflow doc. It's a set of constraints that make the writing inevitable on the day you don't feel like writing. Tools second. Routine first.

If you only steal one thing here, make it the headline-first rule — pick the angle, write the headline, refuse to draft until both land. Everything else is downstream of that one decision.

If you steal two, make the second one the summary cut in step 04. It will shape every piece of writing you do, not just the newsletter.

Once per quarter, block one production day and run the routine four times back-to-back. Pull four headlines from the vault, draft four outlines, write four ledes. Walk away.

The four issues aren't finished — they're staged. The next three sends, you finish one in 30 minutes instead of 90 because the angle and the lede are already there. The fourth one stays in reserve for the week the news cycle eats the schedule.

Net: three Fridays freed per quarter. No quality cost, because nothing went out faster than it should — only earlier.

Small agent. Claude project, voice file in context, one prompt: "Given this title and source link, draft the why-now in one sentence in my voice."

When you drop an idea in the vault, the agent fills the why-now column. You review and edit it before the draft session. The friction of "I'll write the why-now later" disappears, and the shelf fills with stronger material because the test gets applied at capture time, not draft time.

Cost: an afternoon to build, ~$5/month to run. The clearest ROI move in the whole OS.

Tweet the lede sentence midweek. If it earns engagement, the angle is sharper than you thought — leave the headline alone. If it lands flat, you have two days to fix the angle before send.

Faster, cheaper, and more honest than a Beehiiv subject-line A/B test. Twitter tells you whether the idea moves people. The subject test only tells you whether your existing subscribers clicked one of two near-identical openers.

End of each quarter, pick the weakest tool in the stack. Drop it. Try the runner-up for one cycle.

Half the time you switch back. Half the time you don't, and the routine gets a little sharper. Either outcome is good — what's bad is letting the stack ossify around a tool that stopped pulling its weight last August.

Stop pasting the voice file every draft session. Make it a pinned project context. Now every prompt starts inside the right voice — the three rotating prompts get sharper, the rhythm test gets stricter, the press-release risk drops to near zero.

Update the project once a month from the source doc. The doc stays the truth; the project is the operating copy.