LinkedIn Outlier Engine

Find the posts that beat their own author's average — then study why.

Most LinkedIn advice is guesswork because nobody knows what's actually working in your market right now. This fixes that. Point it at the creators worth watching in your space, and it pulls their last 6 months of posts, finds the outliers, and gives you a searchable wall of what worked.

This repo is a blueprint, not an app. There's no code in it. It's a spec that Claude Code reads and builds into a working system on your machine, wired to your own accounts. That's the point — you own the whole thing, and it's shaped around your niche instead of mine.

What you need

• Claude Code
• A Supabase project — free tier is fine
• An Airtable account + a personal access token
• An Apify account with credit on it (this is the only part that costs money — scraping a few hundred creators is a few dollars)
• A list of 30–50 creators in your niche. Nobody can give you this. It's the whole ballgame.

Setup

1. Clone this and open the folder in Claude Code.

2. Make your accounts. Create a Supabase project, an Airtable base, and an Apify token. Copy .env.example to .env and fill in the keys. Claude Code will walk you through any of these if you ask it to.

3. Pick your creators. Copy seeds/creators.example.txt to seeds/creators.txt and put 30–50 real people in it — the ones in your space whose posts you'd actually want to have written. Names or profile URLs both work.

4. Find a scraper. Search the Apify store for a LinkedIn profile-posts actor. Several exist; they vary in price, reliability, and whether they return media. Try two, compare the output, keep the better one. Put its id in .env.

5. Build it. In Claude Code, say:

   Read BLUEPRINT.md and build this system.

   Then let it work. It'll scaffold the code, apply the database schema, and set up your Airtable base. Expect to go back and forth with it — the blueprint is a spec, not a script.

6. Run it. Ask Claude Code to seed your creators, scrape them, and sync to Airtable. Your first real look at the data is the Gallery view in Airtable, sorted by outlier score.

Then what

Open the Gallery. Look at the top 20. You'll see the pattern inside an hour — the hooks that repeat, the formats that carry, the structures that keep showing up. Group the good ones into Collections, write your own version, and check back monthly.

The system finds the outliers. Reading them is still your job.

Fair warning

LinkedIn scraping is finicky and lives in a grey area of their ToS. Some posts come back without media or engagement counts. Rate limits are real. This is a research tool for studying public posts — use it accordingly, and don't point it at anything you shouldn't.

---

MIT. Do what you want with it.