Content research that starts with what her market passes on

Impact Creator helps impact founders build and monetise personal brands on LinkedIn, and Julia writes the posts, carousels and newsletter that sell it. We moved her content engine out of chat onto her own laptop and built its first skill: a LinkedIn analyser that ranks her market's posts by reshares, not likes.

LinkedIn growth and personal brands for impact foundersUnited KingdomDWY Coaching
At a glance
Who uses it
Co-founder-led, content in-house
Engagement
DWY Coaching
Timeline
Built in April 2026
Reshares, not likeshow the analyser ranks her market
On her laptopClaude Code with her own folder, rules and context
The starting point

Where the week went

What the work looked like before we built anything. One problem, one system to fix it.

Impact Creator content, beforefrom the screen-shares and calls
everything waited onJuliare-explaining her voice to Claude every session
1The prototype
A content research engine, built in Claude Chat in February.
It worked, but not the results she wanted.
Still making things up more often than I'd like!Julia
2Every session
Claude forgot her voice and her frameworks.
Copy slid back between sessions.
Every session started from scratch.
3The team
Her copywriter was leaving.
Posts, carousels and a newsletter still to write.
How we worked

Find. Fix. Automate.

Connected along the way
  • Claude Code logo
  • VS Code logo
  • Playwright logo
  • LinkedIn logo
  • Stripe logo
  1. 1Find: we watched the work

    We went through the engine she built in chat and found where it broke: it forgot her voice between sessions and filled the gaps by making things up.

  2. 2Fix: we made the job simpler

    Moved the work out of chat into a folder on her laptop, with rules and context files Claude reads before it starts. Read-only rules for her Stripe connection, so no client data is shared.

  3. 3Automate: we built what runs

    Built the first skill with her, step by step: a LinkedIn analyser that collects posts in her market and ranks them by reshares against likes.

The system

The content research board

A LinkedIn analyser that ranks her market by reshares, running on her own laptop.

The problem

A content engine in chat that forgot her voice and made things up.

What we built

Claude Code on her laptop, starting with research: what her market reshares, ranked.

Before

How a post got researched

  1. Step 1Open a new Claude chatJulia
  2. Step 2Explain her voice and frameworks again
  3. Step 3Ask for topics and research
  4. Step 4where it hurtCheck what it made upJulia
  5. Step 5Write the post
After

How the system runs

Each line is a hand-off the system now does. The orange line is the result a person checks.

what comes in
Posts in her market logo
Posts in her marketthe accounts she points it at
Her folderCLAUDE.md, rules, context
what does the work
Claude Code logo
Claude Codeon her own laptop
Post collector logo
Post collectorlikes and reshares, per post
Rankerreshares against likes
what it makes
Ranked list of what performsby topic
the result
What her market passes onnot only what it taps
who signs off
Julia picks the topicsshe decides what gets written

A like is a tap. A reshare means someone put their name on it.

  1. 1Claude Code installed on her laptop: Node.js, VS Code, the Claude Code extension and a folder with CLAUDE.md, rules and context files.
  2. 2The LinkedIn Content Analyzer: collects posts with the Playwright browser tool and ranks them by how often they get reshared compared with liked.
  3. 3Read-only safety rules for her Stripe connection. No client data is shared.
  4. 4A first Stripe finance skill, built in Plan Mode, still in iteration.
  5. 5A voice template, so the system writes the way she does.
  6. 6A memory pattern (collect, promote what works, clean up) with the prompt to set it up.
What they see every day

The content research board

Impact Creator · Content research boardclaude code · linkedin-content-analyzerIllustrative data
Collecting posts from your market…
What your market passes on
Ranked by reshares, not likesYou pick the topics
0posts collected
8topics ranked
10 per 100best reshares per 100 likes
3topics on the shortlist
Reshares per 100 likes, by topicFirst 96 posts
1AI for small teams
10
2Pricing your offer
9
3Tool reviews
7
4Founder hiring
6
5Productivity
5
6Launch lessons
4
7Client stories
3
8Industry news
2
Write nextJulia decides. The board only ranks.
TopicWhy nowFormatStatus
Founder hiringTop reshare rate this weekCarouselJulia's pick
Launch lessonsClimbed 3 places since MondayPostShortlist
Client storiesLow likes, high resharesNewsletterShortlist
Tool reviewsLiked, rarely passed on·Skip
Ranked list saved to your folder. Run it again whenever you want a fresh read.
What changed

Before and after

A content engine in chat that made things up more often than she would like.
In her words: "a LinkedIn analyser that looks at my market and helps identify what content is actually working."
Claude forgot her voice and frameworks every session.
A folder on her laptop with rules and context that Claude reads first.
Research by asking chat and checking the answers.
Posts collected from her market and ranked by reshares.
An idea she wanted to turn into a workflow.
In her words: "how to take any repetitive, time-consuming task and turn it into a system that can run with AI."
in their words
Back in February, I built a content research engine MVP inside Claude Chat.
Read the full recommendation

Back in February, I built a content research engine MVP inside Claude Chat. It worked, but it wasn't giving me the results I wanted and it was still making things up more often than I'd like! I was looking for someone who actually had a system and could help me turn the idea into a proper content workflow. That's exactly what Saba helped me do.

Impact Creator logo
Julia KotulaCo-founder, Impact Creator · LinkedIn recommendation
LinkedIn recommendationJulia Kotula's recommendation

Run this pattern on your own data

This case is now a pattern the Unstuck MCP is trained on. Connect it to your Claude and it looks for the same signals in your own tools, picks the one worth fixing first, and builds it with you.

Pattern

Content research that ranks what the market reshares

The owner's voice and rules written down once so Claude stops starting from scratch, then posts from the market collected and ranked by how often they get reshared, not liked, so topics come from evidence.

Connect your Unstuck MCP How the audit works
What they own now

Theirs, for good

  • Claude Code on her laptop, with her own folder, rules and context.
  • The LinkedIn Content Analyzer skill.
  • The read-only rules for her Stripe connection.
  • The voice template and the memory setup prompt.

Planned next: turn the ranked topics into drafts in her voice, and finish the Stripe finance tracker.

Start here

Find the work that should not need you

The AI Audit Partner maps where your team's week goes and names the first three jobs worth building.

Book the audit