Platform Zero

You want to call this the golden age. The innocence before everything broke.

Don't.

What May 2006 actually represents isn't a beginning. It's a decision point. A three-to-four-year window when the architectural DNA of the entire social web was being written. When the incentive structures were being locked in. When the people building the infrastructure for human communication were making choices — mostly under investor pressure, mostly without long-term consequence modeling — that would determine how two billion people communicate for the next twenty years.

We call it the golden age now because we know what came after. At the time, it felt like infrastructure being built by people who hadn't quite figured out what they were building. In retrospect, some of them had. The ones who hadn't were the users.

Welcome to Platform Zero.

i · the last clean signal

The historical record is predictably sentimental about 2006, so let me offer the more honest version: it was genuinely different. Not because the people were better, but because the extraction engine hadn't been installed yet.

MySpace was chaotic in ways that were actually good. Users had control over their profile pages — custom HTML, embedded music players, visual designs ranging from "interesting" to "load-time catastrophe." The chaos was real self-expression. Bands used it to talk directly to fans before SoundCloud existed. Artists built followings before "audience building" was a content strategy buzzword. The platform was imperfect but honest about what it was: a public space where users had genuine control over their corner of it.

YouTube in early 2006 was doing something genuinely strange. The video was bad — 320x240, compressed to near-unrecognizable quality — but the form was new. Television had been a one-to-many broadcast medium for fifty years. YouTube was something else: anyone could publish, and the search was functional enough that people could find each other. The early community was small enough that comment sections occasionally made sense. The recommendation engine was primitive enough to be honest.

Twitter's first months were almost comically minimal. A text box, 140 characters, a timeline of people you chose to follow. No algorithm, no engagement optimization, no checkmarks, no promoted content. Just the raw stream of whoever you'd decided was worth reading. Signal-to-noise was determined entirely by your own curation choices.

For a specific window — probably 2005 to 2008, maybe 2009 if you're generous — the social web was doing something close to what its boosters claimed. Publishing power was genuinely more distributed. Small communities could sustain themselves. Niche interests could find their audiences without mass-media economics.

Then the business models caught up.

ii · the code was already written

Here's the architecture lesson nobody wanted to have in 2006, and that most people still resist having now: when you build communication infrastructure on advertising revenue, you've already decided what your product is.

The product is attention. The users are the supply chain.

This wasn't secret. The advertising-supported web had existed for a decade by 2006. Banner ads. Google AdWords. It wasn't a mystery that platforms selling ads needed to maximize time-on-site and engagement. What may have been genuinely underestimated — or possibly just ignored — was how completely this incentive structure would reshape the platforms over time.

Facebook's "clean" design, which positioned itself as the professional alternative to MySpace's chaos, was marketed as aesthetics. It was architecture. The removal of user customization wasn't about taste; it was about control. When users can't modify their environment, the platform controls the experience. The platform controls what's visible. The platform controls what data gets packaged for sale.

YouTube's recommendation algorithm didn't become what critics call a radicalization pipeline because engineers were negligent. It became that because "watch time" was the optimization target, and watch time correlates with emotional activation — outrage, anxiety, compulsive curiosity — in ways that rigorous A/B testing will eventually surface, even without explicit intent. The algorithm learned what kept people watching. It turned out to be mostly unpleasant.

Twitter's shift from chronological to algorithmic timeline was presented as a user experience improvement: you won't miss important tweets. The honest description: the chronological feed gave users control over their information diet; the algorithmic feed transferred that control to Twitter's engagement optimization system. "Important" meant "what our models predict will generate the most interaction" — which correlates with controversy and emotional intensity, not quality or truth.

None of these trajectories required malice. They required only consistent optimization pressure applied to an incentive structure that was misaligned with user wellbeing from the start.

The decay was in the code. The code was written in 2006.

iii · the consolidation that didn't have to happen

The counter-argument — and it's worth taking seriously before dismissing — is that the consolidation wasn't structurally required. The social web could have developed differently.

By 2010, the social web had contracted around three or four platforms. Not because these were the best products, but because network effects concentrated users and capital simultaneously. The platforms that survived weren't the most honest or the most useful; they were the ones that executed the attention-capture model most effectively and attracted enough growth capital to outlast the consolidation period.

Alternatives existed. Federated protocols had been theorized and partially built. RSS was doing genuine work. The blogosphere was a functioning distributed network. Early versions of what would become ActivityPub were being sketched. Subscription-supported platforms weren't unimaginable.

They were just underfunded. Because the advertising model generates the kind of growth curves that attract institutional capital, and institutional capital is what separates the platforms that scale from the ones that stay small.

This is what should sit most uncomfortably about Platform Zero: the future wasn't overdetermined. The choices were real choices. The road to what we ended up with was wide and well-lit, but other roads existed. They just didn't get funded, because the extractive model was more legible to investors than the alternatives.

The small bands MySpace connected to fans? Streaming economics make it financially unviable to have a small audience. The YouTube creators who built genuine communities? Demonetized, algorithmically buried, or forced to produce at industrial scale to survive the recommendation engine's demand for fresh content. Twitter's timeline of interesting people you'd curated over years? Replaced with an algorithmic product that surfaces what the engagement model predicts will generate the most interaction — which is mostly not the people you chose to follow.

Platform Zero didn't make this inevitable. It made it easy.

iv · what the inheritance actually cost

Here's what the May 2006 cohort eventually became.

MySpace had already been sold to News Corp for $580 million in 2005 — technically before the Platform Zero window, which tells you something about how early the commercial logic arrived. It was sold again in 2011 for $35 million. The user-generated culture it hosted dispersed across the platforms that absorbed it.

YouTube was acquired by Google for $1.65 billion in October 2006 — five months after the moment we're describing. The acquisition thesis was correct; video on the web mattered enormously. The company that built the platform users loved was absorbed into advertising infrastructure that systematically reshaped the content ecosystem around its revenue model.

Twitter went through a decade of tortured attempts to monetize its genuinely novel product, failed to find a sustainable model, and sold itself to a billionaire who used it to conduct a public experiment in how much incoherence a platform's network effects can withstand before the network dissolves.

Facebook became Meta, spent tens of billions of dollars on virtual reality infrastructure that no significant number of users wanted, and is currently attempting to maintain dominance through AI integration while its core product's average user age drifts upward.

TikTok — the iteration that emerged after the Platform Zero cohort had already shaped user behavior and expectations — optimized the extraction model beyond anything the 2006 companies attempted. Short-form video, infinite scroll, an algorithm so effective at predicting and reinforcing behavior that its regulatory problems across multiple countries are, at this point, arguably less severe than the behavioral effects on the generation that grew up inside it.

This is the inheritance of Platform Zero. Not the tools themselves — those were fine, occasionally even good. The inheritance is the incentive model, the architectural choices, and the normalization of surveillance-funded communication infrastructure as the default internet.

v · the timer is running again

The generation now trying to rebuild something better — the federated protocol advocates, the indie web builders, the subscription-media people — are working against twenty years of network effects and user behavior shaped by extraction-optimized platforms. Some of what they're building is genuinely interesting. The Fediverse exists. Newsletter platforms exist. Charging users instead of selling them is a real product category.

But they're building in the shadow of platforms that trained two billion people to expect free, frictionless communication in exchange for their attention and data. Breaking that habit is slow, difficult work against gravity.

What May 2006 actually teaches isn't optimism or nostalgia. It's architecture. The decisions made when infrastructure is being laid determine the incentive gradients for everything that follows. Once the advertising model was the default for social platforms, the optimization pressure it created was going to shape everything downstream in predictable directions.

We're not at a new Platform Zero — there's no equivalent clean slate. But the architectural choices being made right now, in AI systems, in the next generation of social infrastructure, in the protocols being standardized in the next three years, will determine what the next twenty years looks like in exactly the way 2006's choices determined these twenty.

The timer starts when the attention model gets embedded in the AI layer.

Check back in twenty years.

vi · sources

• Timeline of social media — Wikipedia
• Google Closes $1.65 Billion Deal for YouTube — The New York Times, 2006-10-09
• News Corp acquires MySpace for $580m — The Guardian, 2005-07-19