The blank LinkedIn box, solved

The blank box has beaten me more times than I want to admit

July 24, 2026

In partnership with

You know the one. Cursor blinking, week slipping away, and you have nothing. Not because you have nothing to say, but because finding the angle from scratch every single week is its own separate job nobody warned you about.

So when I hit a post from someone who built an actual system for this, I read the whole thing twice. Not a prompt. A system that takes real post data, figures out what genuinely works, and hands you a finished caption in your voice every week.

They were upfront about the catch too, which is why I trust it. This only pays off if you post weekly. Post once a month and you will spend around a hundred bucks over time for almost nothing back. Post often and the math flips hard in your favor.

Here is the whole build.

Win AI Search Without a Big Team

92% of VCs use AI to find companies. 58% of buyers start there, too. If you're not showing up in AI answers, you're invisible before the conversation even starts. Join HubSpot for Startups, Anthropic, and Marketing Against the Grain on July 30 (11 am ET) for a live AEO teardown. Real startup. Real recs. Register and unlock the free Startup Visibility Bundle.

*Ad

The first move made me put my coffee down

Almost everyone starts by training the AI on their own posts. That feels right. It is your voice, after all.

They flip it. And the reasoning is hard to argue with.

If your posts already perform, sure, train on them. But if they are middling, feeding an AI 400 average posts just teaches it to be reliably average. You have built a machine for producing exactly the thing you were already stuck producing.

So instead: find the best creator in your niche first. The person whose feed you genuinely wish was yours. Ideally 30 or more posts with real engagement behind them.

You are not copying anyone. You are showing the AI what great looks like in your specific corner of the world before it ever touches your material. Calibration first, imitation never.

Getting the raw material for about the price of nothing

They use a tool called Apify to pull public posts. The flow is four steps:

Make a free Apify account.

Open the "LinkedIn Profile Posts" agent.

Paste in the profile URL.

Hit Save, then Start.

A free signup gets you around five dollars in credits, which covers up to 2,000 posts. Day one costs you nothing.

One safety note they stressed, and I am repeating it because it matters: never automate anything from your own logged-in account. These posts are already public. The tool pulls them without ever touching your login, and that distinction is the difference between a clean workflow and a suspended profile.

Give it about two minutes. The raw output looks terrifying at first glance. It is fine. Export it as an Excel file and move on.

How Jennifer Aniston’s LolaVie brand grew sales 40% with CTV ads

The DTC beauty category is crowded. To break through, Jennifer Aniston’s brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category.

*Ad

Where 99% of people quietly ruin it

This is the step they flagged as the big one, and I think they are right.

The spreadsheet lies to you.

Not maliciously. It just does not say what you think it says. So rather than letting the AI eyeball the data and pattern-match its way to a vibe, you upload the Excel and give it one blunt instruction. Their exact wording is worth stealing:

Compute the exact stats in code. Don't guess.

Three specific traps sit behind that instruction:

The real content format is not in the caption. It is hidden in which media columns are filled. Read the caption and you will misclassify half your data set.

The post date is buried inside each post's activity ID. The AI can decode it. You never could by looking.

For images and carousels, you have to explicitly tell the AI to open the media and actually look at it. Left alone, it will judge a post by its caption and confidently tell you the wrong thing.

Eight to ten minutes later you get two files out. One is a report explaining what wins and why. The other is an SOP, which is really just the exact recipe for your next post.

Turning a one-time experiment into a command

Here is the clever bit. That recipe becomes a Claude skill.

Which means the entire analysis stops being a thing you did once and becomes a thing you can call. The logic lives somewhere permanent. You never rebuild it, never re-explain it, never dig through an old chat trying to remember how you did it in March.

That is the difference between having done something impressive and owning something useful.

The best voice models, now with full orchestration. Build real-time voice and chat agents on one low-latency stack: any LLM, your tools and knowledge, testing, Guardrails, and omnichannel deployment.

*Ad

The weekly loop, which is honestly the whole point

Five steps, and it is beautifully boring:

Open a fresh chat in Claude, type a slash, invoke your skill.

Answer three quick questions.

Get five angles back, or ask for far more if you are feeling greedy.

Pick the one you like.

Let it write the caption and tell you what image to make.

One invocation, one finished post. Every week. That is the loop, and there is nothing else to it.

Two warnings I would not skip

What sold me on this post is that they did not oversell it. Two cautions came attached.

The first: a viral post is like a burger. Delicious, dangerous every day. Clone your best-performing recipe occasionally, not on repeat, or your feed turns into one note played over and over.

The second is really the heart of the thing. Close the loop. Every good post you publish becomes new training data. Better posts feed a smarter system, which produces better posts. That compounding cycle is the actual asset. Skip it and you have a clever one-off. Run it and you have something that sharpens itself while you sleep.

What I keep circling back to

Content creation stopped being a weekly panic and became a system that gets better the longer you run it. You are not producing posts anymore. You are building something that improves itself, and the posts fall out the side of it.

Tonight, do just the first part. Open LinkedIn and find the one creator in your niche whose feed you actually envy. Not the biggest account. The one whose posts make you think "I wish I had written that." Save the profile URL somewhere.

That single choice determines the quality of everything the system produces afterward. Get it right and the rest is just plumbing.