Staring at a blank LinkedIn box every week is one of the most draining parts of building an audience. You know posting consistently works, but coming up with angles and captions from scratch burns hours you don’t have. That’s why I got so excited when I found this post from an AI professional who built a full system to fix exactly that.
The creator lays out a way to turn your own LinkedIn content into a machine that writes posts in your voice. And here’s the honest part I appreciated: the original poster says straight up that this only pays off if you post weekly. Post once a month and you’ll spend around a hundred bucks over time for little return. Post often, and it becomes your unfair advantage.
I broke down the whole process below so you can follow it step by step.
Step 1: Don’t start with your own posts
This surprised me. Most people would train an AI on their own content first. The expert flips that.
The logic is sharp. If your posts already perform well, sure, train on them. But if they’re mediocre, feeding an AI 400 average posts just teaches it to be average. Garbage in, garbage out.
Instead, the author suggests finding the best creator in your niche. Pick the person whose LinkedIn you genuinely wish was yours, ideally someone with 30 or more posts and real engagement. You’re not copying them. You’re showing the AI what great looks like in your world before you ever feed it your own material.
Step 2: Pull the posts cheaply
Next, the creator uses a tool called Apify to grab public posts. Here’s the flow the post walks through:
- Make a free Apify account
- Open the “LinkedIn Profile Posts” agent
- Paste in the profile URL
- Hit Save and Start
According to the post, a free signup gives you around $5 in credits, which covers up to 2,000 posts. So you pay nothing on day one.
One safety tip the author stresses: never automate anything from your own logged-in account. These posts are already public, and the tool grabs them safely without touching your login.
After roughly two minutes you get the raw output. The original poster warns it looks terrifying at first. It’s fine. Just export it as an Excel file.
Step 3: Make the AI do math, not vibes
This is the step where the expert says 99% of people go wrong, and I think it’s the smartest part of the whole system.
The spreadsheet lies. So instead of letting the AI eyeball the data, you upload the Excel and give it a clear instruction. The author’s exact wording is worth copying:
Compute the exact stats in code. Don’t guess.
Why this matters so much comes down to three traps the creator points out:
- The real content format is hidden in which media columns are filled, not in the caption
- The post date is buried inside each post’s activity ID, which the AI can decode even though you never could
- For images and carousels, you have to tell the AI to actually open the media and look, never to judge a post by its caption alone
Around eight to ten minutes later, the post says you end up with two files. One is a report that explains what wins and why. The other is a SOP, basically the exact recipe for your next post.
Step 4: Turn the recipe into a reusable skill
Here’s where it gets clever. The creator turns that recipe into a Claude skill, so the whole analysis becomes a repeatable command instead of a one-time experiment. Once it lives as a skill, you never have to rebuild the logic again. You just call it.
Step 5: Run your weekly loop
This is the payoff, and it’s beautifully simple. The author describes the weekly rhythm like this:
- Open a fresh chat in Claude and type a slash to invoke your skill
- Answer three quick questions
- Get five angles back, or ask for far more if you’re feeling greedy
- Pick the one you like
- Let the AI write the caption and even tell you what image to create
One invocation equals one finished post. Every single week. That’s the loop.
Two honest warnings
What I respect about this post is that the creator doesn’t oversell it. There are two cautions worth repeating.
A viral post is like a burger: delicious, but dangerous if you eat it every day. Clone your best recipe occasionally, not on repeat.
The second warning is really the heart of the whole thing. Close the loop. Every good post you publish becomes new training data. Better posts feed a smarter AI, which produces better posts. That compounding cycle is the entire game.
I love this because it turns content creation from a weekly panic into a system that gets sharper the longer you run it. You’re not just producing posts. You’re building an asset that improves itself.
If you want the full breakdown with every detail the creator shared, go check out the original LinkedIn post. It’s worth the read.