My LinkedIn posts itself. Every week. In my voice.

An autonomous weekly workflow, run by Claude Cowork, that researches what performed, drafts a full week of posts trained on my writing, scores every draft against a quality bar, and schedules the week. 100 posts published on this system. I review; I don't write.

ONE WEEKLY CYCLE · content-engine ● runs every week

FRIperformance review written to Drive: what worked, what died, what changes done

SATbatch run drafts the week against my voice profile + Friday's findings done

SATevery draft scored, weak hooks rewritten done

SUNdrafts hold for my final approval, Sunday night. Nothing moves without it enforced

M-Fapproved posts publish, 7:30 AM PT done

FRIresults feed the next report. Loop closes done

Stack behind the run: Claude Cowork weekly routine → Supergrow Content DNA, scoring + scheduling → performance data feeds the next run.

100posts published on this system

5/wkaverage cadence, Apr–Sep 2026

19weeks running

The system

Three pieces. None of them works alone.

The outcome is a scheduled week of posts in my voice, produced without me writing. No single tool can deliver that. Three of them, each feeding the next, can.

Claude Cowork

THE OPERATOR

Runs the weekly routine end to end: reads performance, decides angles, drafts, triggers scoring, schedules. Alone, it has no memory of my voice and no publishing arm.

Supergrow

THE PUBLISHING LAYER

Holds my Content DNA voice profile, scores every draft, schedules the slots, and reports what performed. Alone, it waits for a human to show up and write.

Google Drive

THE SOURCE OF TRUTH

My real stories, numbers, and frameworks, plus the system's own paper trail: friday-reports, weekly-batches, pending-revisions, lainie-note. Alone, it's a folder. This is what keeps the posts from being generic.

The loop: Drive feeds the raw material, Cowork drafts and schedules, Supergrow scores and reports what performed, and that report shapes Saturday's batch. Nothing publishes until I give final approval on Sunday night. Human in the loop, every week.

How it runs

A closed loop, not a content calendar.

Most "AI content" is a person pasting prompts into a chatbot every morning. This is a system: each week's performance data changes what next week writes. My job shrank to one review pass.

1. Performance in, not vibes inThe run starts by reading last week's report: which hooks earned comments, which formats died. Next week's drafts are shaped by that, automatically.
2. A voice profile, not a style promptDrafts are generated against a Content DNA profile built from my actual writing, plus a knowledge base of my real stories, numbers, and frameworks. That's why the posts sound like me and not like AI oatmeal.
3. A quality gate the machine enforces on itselfEvery draft is scored on hook, clarity, originality, and voice alignment. Anything under the bar gets rewritten before I ever see it.
4. A human gate before anything shipsI review the scheduled week. Approve, tweak, or kill. The system does the labor; I keep the judgment.

Inside the loop

The paperwork is real.

Every run leaves artifacts in Drive. Two of them, lightly excerpted:

friday-reports/2026-09-04-report.md · written by the Friday review run

What worked: the false-victory hook is generating consistent 1.5 to 2x multipliers on vertical posts. Prioritize it.

What didn't: the Aug 25 hook didn't stop the scroll. Fix: lead with the cost of the failure, not the mechanism.

Adjustment for next batch: no more than 1 post per day. Two posts published 90 minutes apart cannibalized each other.

The system critiques its own output weekly and Saturday's batch drafts against these findings.

scheduling-run/2026-08-31.md · written by the scheduler

Result: NO POSTS SCHEDULED. 10 drafts in pipeline, all awaiting Lainie's approval. Auto-publish will proceed once posts are approved.

The human gate isn't a promise. The scheduler enforces it: no approval, no publishing.

The voice test

Read two. Tell me they sound like a bot.

Excerpts from posts this system wrote and published, unedited.

Published Sep 4, 2026"The win at a dental practice this year is not the AI assistant that reads x-rays. It is the assistant that drafts the appeal letter for the claim the front desk already wrote off. Two thirds of denied dental claims are never resubmitted."

Published Jul 20, 2026"Your HVAC company runs a 24-hour on-call rotation. You are still losing the 11pm emergency call to whoever answers first."

Why this matters for you

If it works for my business, it works for yours.

I run this system on my own company first, which means every rough edge got fixed on my time, not a client's. The same architecture adapts to any founder or firm that needs a consistent presence without a ghostwriter's invoice: your voice profile, your knowledge base, your approval gate, your cadence.

Built and operated by an enterprise PM with 20+ years of process automation behind it, and paired with the voice agent work it promotes.

Contact

Want your content running on rails?

Tell me your platform and your cadence. I'll tell you what a system like this looks like for your business, and show you mine running first.

Email me

Or find me on LinkedIn · lnlaiagency.com · Upwork