The LinkedIn Post: An Exhaustive Taxonomy

The line breaks are the part I could never get past.

One sentence.

Then a gap.

Then three words, alone, load-bearing.

I spent years reading that formatting as an aesthetic failure — a genre of person who had confused a business update with a poem. And then I read LinkedIn's engineering blog, which explains in flat technical prose exactly why a post shaped like that outperforms a paragraph, and the whole thing stopped being funny and became something considerably more interesting.

Nobody chose that shape. It was selected for.

The mechanism, from the people who built it¶

LinkedIn's engineering team published this in May 2020, and it is not a leak or an inference. It's a blog post about improving the feed.

“

Each update viewed on the feed generates two types of dwell time. First, there is dwell time "on the feed," which starts measuring when at least half of a feed update is visible as a member scrolls through their feed.

”

Read that as a specification rather than a description, because that's how the incentive works. Measurement begins when half the post is on screen. A post that occupies more vertical space is on screen, half-visible, for longer — during an identical scroll, at an identical reading speed, containing identical information.

Whitespace is not a style. It is surface area.

The second half is the penalty. LinkedIn describes building a model predicting the likelihood a member skips an update, and then, in their words, they "reduced the score of all updates by an amount proportional to the predicted P(skip) value."

So the system does two things at once: it rewards the post you linger on, and it actively demotes the post you glide past. Every convention that follows is a solution to that pair of constraints.

The taxonomy¶

Sorted by what ranking behaviour each one exploits, which turns out to be a more useful classification than how embarrassing it is.

---

1. The Vertical¶

I learned something on Tuesday.

It changed everything.

Here's what happened.

2. The Fold Hook¶

Nobody talks about what happens after you get the promotion.

…see more

3. The Parable¶

A candidate showed up 20 minutes late to the interview.

I hired her on the spot.

Here's why. 👇

4. The Provocation¶

Unpopular opinion: most meetings are fine.

5. The Closer¶

Agree?

♻️ Repost if this resonated.

---

Five formats. Not one of them is about the content. Every single one buys either screen height or thumb-stop, which are the two things the published ranking function is known to measure.

The part that isn't a joke¶

Now the taxonomy entry I deliberately left out, because it doesn't belong in a list of formats being mocked.

The layoff post. The one that starts "I'm sorry to share that I was impacted by the recent restructuring at…" and goes on to be gracious about a company that has just removed someone's income.

That post is doing something entirely rational. It's the most effective distribution mechanism available for the sentence I need work, it reaches former colleagues who might hire you, and the graciousness is not delusion — it's a person who knows that anger reads badly to future employers and who cannot afford to read badly to future employers.

And it is the single most mocked format on the internet.

Look at what's actually happening there. Someone has lost their job, in an event they didn't control and were probably assured wasn't coming, and is now performing composure on a platform whose ranking function rewards vulnerability with reach — which means the more openly they discuss the worst month of their year, the better it performs. That is a genuinely grotesque incentive, and the person subject to it did not design it.

The cringe is real and it is not theirs.

That's the turn, and it applies backwards through the whole taxonomy. The formats above look ridiculous because they are optimised for a measurement rather than for a reader. But the people doing the optimising are, in the main, not influencers — they're job seekers, new graduates, contractors whose pipeline dried up, and people in industries that contracted. They're using the tool the way it rewards being used, because the cost of using it badly is not being seen.

Who profits from the confusion¶

There's a second group, and this one is a fair target.

The advice industry that sells the taxonomy back to you. The course on hooks. The consultant whose entire offer is knowledge of the ranking function, monetised specifically because the function is undisclosed enough to make that knowledge scarce.

They are not doing anything illegal and some of what they teach demonstrably works. But notice the structure: their product is an information asymmetry about a system they didn't build and can't change, sold mostly to people who need the platform to work for them. Same shape as a values poster that predicts nothing about the company, and the same shape as an earnings call that discloses nothing while speaking fluently: a document, or a dialect, whose function is to be persuasive to whoever most needs to believe it.

And underneath all of it is the thing this publication keeps arriving at from different directions. Attention is the good being traded, the ledger for it is one almost nobody keeps, and a platform that can measure dwell time to the half-visible pixel is running a far better set of books on your attention than you are on theirs.

Reading the feed differently¶

1. 1

   Ask what each post is buying

   Height or thumb-stop. Once you can name which, the format stops being irritating and becomes legible — and you stop attributing to the author a decision that was made by a ranking model.

2. 2

   Distinguish the optimiser from the optimised

   Someone with a course on hooks is working the system. Someone announcing a redundancy is being worked by it. These deserve completely different reactions and get the same one, which is most of what's wrong with how this genre is discussed.

3. 3

   Notice what the format selects against

   Clarity, brevity, falsifiability and attribution — all penalised, structurally, because each one reduces dwell. That's a useful thing to know about any feed you use to form views about your industry.

4. 4

   If you have to post, do it deliberately

   There is nothing wrong with using the format knowingly. Write the thing you mean, then decide separately whether to shape it for distribution. The failure mode is drifting into the register without noticing, which is how people end up writing sentences they would never say out loud.

5. 5

   Treat the platform as a job-search tool with known mechanics

   It is genuinely useful for finding work, and being fluent in its conventions is a real advantage. Read the postings it carries the way you'd read any disclosure document, and treat the feed itself as a ranked surface rather than a representation of what your industry thinks.

One sentence¶

I still can't read the line breaks without wincing. That reflex hasn't gone anywhere and I don't especially trust it any more.

What changed is that I now know the wince is aimed at the wrong object. Somewhere there is a model that reduces a post's score in proportion to how likely you are to skip it, and somewhere else there is a person who has worked out, without ever reading a line of that engineering blog, that a sentence alone on its own line performs better than the same sentence in a paragraph.

They didn't do anything wrong. They ran an experiment, the experiment returned a result, and they kept the result.

The lightbulb is the story.

Not the moth.

It's just business.

Sources

1. LinkedIn Engineering — 'Understanding dwell time to improve LinkedIn feed ranking' (12 May 2020)
   +How this was checked

   LinkedIn's own description of the mechanism: 'Each update viewed on the feed generates two types of dwell time. First, there is dwell time on the feed, which starts measuring when at least half of a feed update is visible as a member scrolls through their feed. Second, there is dwell time after the click.' The team describes dwell time as a real-valued and more reliable engagement measure than clicks, and reports building a P(skip) model where they 'reduced the score of all updates by an amount proportional to the predicted P(skip) value.' Feed ranking has been revised many times since; the durable claim here is the direction of the incentive, not the current specification.