The problem
He wanted to post daily on X without a scheduler, a social media manager or Sunday batching. The work behind each post is research, drafting, formatting and a follow-up CTA, and that is what he handed to Claire.
The blueprint (as he describes it)
- One agent, many skills. Claire runs 24/7 on OpenClaw, connected to Discord, Gmail, Google Calendar and his file system. Each task is a markdown skill with formatting rules, voice guidelines and quality checks.
- Quote tweets on request. He scrolls X for 5-10 minutes, sends a URL to Claire in Discord, and she reads the tweet and drafts a quote tweet: a bold hook, a numbered breakdown of 3-7 steps, then specific numbers or examples.
- X articles in one shot. A topic (his, or researched with last30days), a 700-1000 word body as a numbered list of 5-10 items, an intro and CTA, 7 headline options, and 3 thumbnails built as HTML and rendered with headless Chrome at 1250x500. About 10 minutes of his time, mostly picking a headline.
- Every episode becomes a tweet. He sends the podcast transcript and Claire writes a promo that leads with the tactic, not “New episode dropped!”
- A CTA reply on every hit. For each high-performing tweet a skill reads it and drafts 2-3 CTA options that bridge the topic to an offer.
- Research first. The open-source last30days skill pulls the past 30 days of Reddit, X and web posts with engagement numbers, and Claire picks the 3-5 strongest signals.
- A voice file under everything. The brand-voice skill keeps every draft sounding like him, and the skills get refined each time he gives feedback.
Stated results
- One quote tweet on a Kevin O’Leary clip “hit 471K views and 7,600 bookmarks.”
- “20 million views per month” on X. The page title says 21M.
- 3-5 pieces of content a day with “maybe 30-45 minutes of active work.”
- Setup cost: one OpenClaw instance on a Mac mini, a Claude subscription and the time spent building skills.
Caveats
- View counts are self-reported; the post shows no analytics.
- First published April 2, 2026 and updated October 29, 2026; it doesn’t say which numbers are from which date.
- He sells an academy and services and gives away his brand voice template in exchange for an email.
- He still reviews every draft and picks the options; Claire drafts, he posts.
- Handles: X @coreyganim, LinkedIn and YouTube linked from his site. No Bluesky, Threads, Instagram or TikTok profile verified.
FAQ
How does Corey Ganim use an AI agent for X content?
His OpenClaw agent Claire runs on a Mac mini and takes requests in Discord. Markdown skills tell her how to draft quote tweets, X articles, podcast promo tweets and CTA replies in his voice.
How much time does he spend?
He says 3-5 pieces a day take maybe 30-45 minutes of active work, mostly reviewing drafts and picking options. An X article takes about 10 minutes of his time.
What results does he report?
One quote tweet hit 471K views and 7,600 bookmarks, and he says the account gets 20 million views a month on X. He shows no analytics.
Sources
- https://www.coreyganim.com/blog/my-openclaw-gets-me-21m-views-per-month-on-x-steal-my-full-system-step-by-st (primary: Corey Ganim’s post “My OpenClaw gets me 21M views per month on X”, April 2, 2026, updated October 29, 2026)
- https://github.com/mvanhorn/last30days-skill (the open-source research skill he links)
Want agents like these working your distribution? Get It Built. Get new teardowns weekly: subscribe (double opt-in).
Get one teardown a week.
Same format, a new operator every issue. Double opt-in.