Every page began as a question.
Some questions stayed unfinished. Some disappeared after a night's sleep. A few refused to leave.
Those became essays.
What you'll find here isn't a collection of conclusions. It's a record of ideas that remained interesting long enough to deserve careful thought, slower writing, and the patience to let them become something worth sharing.
Begin anywhere. Leave whenever you've found enough.
LinkedIn started as a filing cabinet. That's not an insult. It was the whole point.
In 2003, the premise was almost aggressively boring: your professional network is an asset you own, but you have nowhere to keep it. Your employer's CRM doesn't count. Your Rolodex doesn't survive a job change. So here's a database. Upload your résumé, connect to people you've actually worked with, and leave.
And leave. That part matters more than anything else in this story.
Early LinkedIn didn't want your attention. It wanted your accuracy. The connection request literally asked how you knew someone — colleague, classmate, did business together — and lying about it was discouraged. Trust was the moat. The revenue model matched: recruiters paid for search access, which works fine even if you log in twice a year, because a static profile is still a valuable index entry.
A product that helps you find a job perfectly is a product you leave. LinkedIn was, briefly, okay with that.
LinkedIn went public in 2011. Public markets don't reward accuracy. They reward usage.
A database where users show up twice a year has a catastrophic DAU/MAU ratio and no ad business worth the name. So the product started manufacturing reasons to come back. The feed arrived in 2012. Influencer posts the same year. Open publishing for everyone in 2014.
But the real tell was endorsements.
Think about what an endorsement actually is. Someone who has never watched you write a line of code clicks a button asserting that you know Python. It carries close to zero information. It is, as a signal, worthless — and everybody knows it's worthless, which is why it became a running joke almost immediately.
So why does it exist?
Because it generates a notification. The notification generates a session. That's it. That's the entire function. Endorsements are a feature built for a metric, not for a person, and they were shipped anyway — which tells you exactly when the company's compass changed, and to what.
Microsoft's $26B acquisition in 2016 didn't cause the drift, it just funded it. The feed hardened. Algorithmic ranking replaced chronology, and the moment posts started competing, a genre emerged — because a competition selects for whatever wins.
What won was broetry. Short lines. One sentence each. Building to a lesson you could see coming from three scrolls away. "I rejected a candidate for wearing a hoodie. Then I realized I was the problem." Nobody sat down and designed this. It was reverse-engineered, by thousands of people independently, from a ranking function that rewarded dwell time and comments and had no mechanism whatsoever for punishing emotional fabrication. The platform never asked for it. It just paid for it, and the market delivered.
Meanwhile, quietly, the original thing decayed. InMail became noise. Connection requests decoupled from actual acquaintance. That "how do you know this person" question vanished, because it was friction, and friction shrinks the graph, and the graph feeds the feed.
Then came vertical video. TikTok in a blazer.
Here's the part worth being precise about, because "LinkedIn has video now and I don't like it" is a weak complaint and it's not the one I'm making.
A résumé database is a tool. You arrive with intent, you complete a task, you leave. There's a finish line. You can be done updating your profile.
An infinite vertical feed is a casino. Variable-ratio reward, no completion state, no natural exit. There is no version of "I'm finished scrolling" the way there's a version of "I'm finished with my profile" — because a finish line is a lost session, and lost sessions don't ship.
That's not a feature addition. That's a change in what kind of object the product is.
And then generative AI removed the last remaining constraint. Writing a fake vulnerability post used to cost twenty minutes and a small deposit of self-respect. Now it costs one prompt. The supply of engagement-optimized content went effectively infinite overnight, which means the median post's information content is converging on zero while volume converges on infinity. AI-written posts, read by AI summarizers, answered by AI-generated comments. A closed loop that produces metrics and nothing else.
The obvious objection: every platform has this problem. Why single out LinkedIn?
Three reasons.
It's a worse casino than the actual casinos. TikTok has a decade of ranking infrastructure and a creator economy that genuinely selects for talent. LinkedIn is running the same game with worse tooling, staffed by people whose real skill is B2B sales. It's competing on the one dimension where it has no advantage, against opponents who have nothing but that advantage. Of course it loses.
Context collapse does the actual damage. Instagram's clickbait costs you nothing — you're already in leisure mode, the frame is "waste time," nothing is violated. LinkedIn is the one place where your identity is real. Your name. Your employer. Your actual career. So when hollow motivational sludge shows up in that frame, it doesn't just fail to entertain — it contaminates the signal you came for. Same content, different container, completely different harm.
The incentives are inverted. On TikTok, engagement bait is the end — the creator wants views, and at least wants them for the thing they made. On LinkedIn, it's instrumental. The poster wants leads, credibility, a consulting pipeline. Which means the content is produced by someone with no interest in the craft, aimed at an audience they privately think of as a funnel. That's a uniquely soulless output, and you can feel it even when you can't name it.
Here's the tell that it isn't working: nobody says "I saw a great video on LinkedIn." They say "I saw this insane thing on LinkedIn." Screenshot-and-mock is the platform's real viral loop now. It generates engagement by being embarrassing — and a dashboard cannot tell the difference between being good and being humiliating. Both are impressions.
The users aren't victims here.
The engagement-bait post gets written because it works. Reach converts to clients. Clients convert to revenue. Everyone in the loop — the poster, the algorithm, the advertiser, the guy leaving "Great insights, Sarah!" on 40 posts a day — is behaving rationally given what they're paid for.
That's what makes it stable. And that's why complaining doesn't fix it. You cannot shame your way out of an equilibrium where defecting pays.
LinkedIn didn't lose its way. It found a more profitable one.
It had a rare escape hatch, too — a real non-attention business. Recruiters and subscribers paid for access to accurate data, not eyeballs. It could have stayed a utility forever. It chose the feed anyway, because a utility's growth curve is bounded by the number of professionals alive, and a casino's isn't.
So today it's two products stapled together. Underneath: a genuinely useful, boring, irreplaceable professional index — still the best in the world at the thing it was built for. On top: an attention machine that degrades that index a little more every quarter.
The index is why you can't leave. The machine is why you don't want to be there.
And the version that isn't a casino still exists. It's just been demoted below the fold — available to anyone who arrives with intent and leaves when they're done. Which is, of course, exactly the behavior the business model was built to eliminate.