The Broadcast Layer

They were right for about eighteen months.

Twttr launched publicly on July 15, 2006 — an escape pod welded together inside Odeo, the podcasting startup Apple had just flattened by putting podcasts into iTunes. Jack Dorsey, Noah Glass, Biz Stone, Evan Williams. The prompt was "What are you doing?" The answer had to be short because SMS gave you 160 characters and they reserved twenty for your username. That's the whole design story. The most consequential constraint in the history of public speech was a rounding error against a telecom spec written in the eighties.

Here's the part the retrospectives skip: the thing they shipped in 2006 is not the thing that ate the next two decades.

In 2006 the timeline was a pipe. Everyone you followed, in the order they said it. No ranking. No promotion. No "in case you missed it." If you wanted more, you followed more people. If you wanted less, you unfollowed. The system had exactly one lever and the user's hand was on it. That is not a small property. That is the entire difference between a medium and a machine.

Then, eight weeks later, Facebook shipped News Feed. September 5, 2006. Users revolted immediately — hundreds of thousands piled into protest groups within days, which is a hell of a thing when the entire user base was college kids. Zuckerberg posted "Calm down. Breathe. We hear you," and kept the feed. Of course he kept the feed. The feed was the business.

That's the real anniversary. Not one launch — two, in the same summer, from opposite ends of the same idea. A broadcast layer and a ranking layer, born within two months of each other, and we spent twenty years finding out which one was load-bearing.

The obvious objection is that this is nostalgia, and partly it is. So let me answer it rather than wave at it. The chronological feed was not virtuous. It was unfinished. Twitter didn't ship an algorithmic timeline until February 2016 — a full decade — and the reason wasn't principle. Nobody had built the ad machine that made ranking pay yet. Hold onto that, because it turns out to be the entire story.

Meanwhile the features that actually made the thing work came from the users: the @-reply, the retweet, the hashtag. Chris Messina proposed the hashtag in a tweet in 2007 and the company adopted it years later because resistance was pointless.

Sit with that ledger. Nearly everything that made the platform useful was invented by the people standing in it. Nearly everything that made it unbearable was invented by the people running it. Messina published a convention in public, for free, and it became a core interaction primitive of a company he had no claim on. That isn't irony, and it isn't only an incentive gradient. It's an enclosure — value generated in common, captured privately. The commons kept the invention. The company kept the deed.

So what did the 2006 architecture actually have? Not auditability. I want to be careful here, because the appealing version of that claim doesn't survive contact. You can audit the list of accounts you follow. You cannot audit the process that produced it — who got retweeted into your view, what search surfaced, whose follows seeded yours, or the plain fact that the early accounts compounded their advantage forever, which traces back to no decision you ever made.

What the pipe had was smaller and more durable: the distortion was load-visible. The loud were visibly loud. You could watch a pileup happen, count the retweets, see a bad idea travel. Ranking doesn't remove that distortion — it hides the hiding. The thing shaping your picture of reality now has no surface you can put a finger on, and you can't miss what you were never shown.

The lesson isn't "bring back chronological." Chronological is a checkbox now. X has one. Instagram has one. They're two taps deep, default off, and they quietly reset themselves every few updates — the software equivalent of a fire exit that locks from the outside.

And the lesson isn't that whatever layer gets built first becomes the ground, either. Twitter's pipe was first and it lost anyway. It just took ten years, because that's how long the money took to arrive. Facebook shipped ranking in week eight because News Feed was ad-shaped from birth. Twitter held out a decade on a missing revenue model and folded the month the model showed up. Architecture didn't decide this. Accounting did. Whatever layer pays becomes the ground, and the first layer survives only as long as nobody has worked out how to charge for the second one.

Twttr shipped a pipe because a pipe was all they could afford to build. Twenty years later the pipe is buried under a funnel with a paid tier and a rebrand, still down there, still holding up a structure it was never engineered to hold — and still owned by nobody who built it.

Somewhere right now there's another underfunded side project shipping the next broadcast layer, and it looks harmless, and it is harmless, and it will be harmless for approximately eighteen months, or until someone works out how to charge for it, whichever comes first.

Happy birthday. Ask me again in 2046.