# How Moonshi MohsenRuddin handed his firm's LinkedIn to a team of Claude Code agents, with a person approving every post

Moonshi MohsenRuddin runs Scaling-Up! Ventures, a Singapore consultancy, and says he isn't a coder. He set up four Claude Code agents to brief, write, edit and analyse LinkedIn posts, and a person approves every post. He published the before-and-after, including the number that makes it look less impressive.

| | |
|---|---|
| **Who** | Moonshi MohsenRuddin, Scaling-Up! Ventures, Singapore ([scalingupventures.com](https://www.scalingupventures.com), [LinkedIn company page](https://www.linkedin.com/company/scalingup-ventures/)) |
| **Problem** | "Every post waited on me." He was the bottleneck on his firm's marketing |
| **Stack** | Claude Code, four role agents, scheduled skills, an idea sheet, a weekly metrics dashboard |
| **Confidence** | High for the system. He says part of the lift is confounded (see Caveats) |
| **Last verified** | 2026-10-15 |

## The problem
"I would draft something, sit on it, forget it, and three weeks would pass with nothing published." He didn't want to hire an agency, because, in his words, "when you outsource your own marketing, you outsource your Clarity."

## The blueprint (as he describes it)
1. He wrote down the brand rules, five audience pain points and content pillars once. "The system reads them every time for every post."
2. Four agents, one job each: a strategist writes briefs, a writer drafts posts in his voice, an editor reviews them against performance data, and an analyst reads the numbers.
3. Skills on top: a daily skill checks his idea sheet, a weekly skill scans Singapore business news for stories worth reacting to, and another schedules approved posts.
4. A fixed pipeline: idea → draft → review → approved → scheduled → published. Each week he pulls in metrics to see what worked and refine the next round.
5. Every draft goes through an "adversarial editor gate", then waits for him or his colleague Mark to approve it. "No post ships without a person saying yes."

He says he's describing the system "at the operating-system level, not the config level". No prompts or config are published.

## Stated results
Before: 1 Jan to 9 Apr 2026, manual, 21 posts over 14.1 weeks. After: 10 Apr to 24 May 2026, with the system, 38 posts over 6.4 weeks.
- Total monthly reach: about 2,200 → about 54,000 ("roughly 25x")
- Cadence: 1.5 → 5.9 posts a week
- Average reach per post: 338 → 2,113. With the two viral posts removed, it's 942
- **Median post: 242 → 309 (about 1.3x).** He calls this "the most important number in this whole piece"
- Record post (S$50K AI grant for SMEs, 10 April): **26,874 impressions** in the results section. His TL;DR and intro say the system's first post hit **25,560**. Both figures appear in the source. The 26,874 figure is dated and itemised, so 25,560 may be an earlier snapshot. **[UNVERIFIED]** which is final.
- Comments per post: 0.9 → 4.4. Reactions per post: 4.8 → 8.8
- Time saved: he says, "I have not put a verified hours-saved figure on it, so I will not invent one."

## Caveats
- The account was "already climbing out of an October-November 2025 algorithm dip", so he says some of the lift is recovery, not AI.
- The two windows are different lengths, so he only compares per-post and per-month figures.
- He says some of the newer posts' impressions were still growing when he measured.
- This is an SME's company and founder account, not a solo creator. The post also promotes his consultancy's services.
- X: no verified handle found (checked 2026-10-15). @MOONSHI is an unrelated account, and his site links only LinkedIn. Tag him on LinkedIn instead.

## Sources
- https://www.scalingupventures.com/claude-code-linkedin-marketing-case-study/ (first-person, published 10 Jun 2026, updated 25 Jun 2026)