The World Wide Web As A Failed Unity Experiment — And What Comes Next

Tim Berners-Lee has been remarkably consistent about what he intended. His 1989 proposal at CERN was practical — a system for physicists to share documents across incompatible computer systems. But the underlying architecture was radical: a hypertext system with no central node, no hierarchy, no gatekeeper. Anyone could link to anything. The web had no built-in concept of ownership.

By the mid-1990s, as the web scaled beyond academic use, Berners-Lee articulated the vision more explicitly. The web was supposed to be a universal medium for sharing knowledge. Not a publishing tool for elites. Not a commercial platform. A commons.

He founded the World Wide Web Consortium (W3C) in 1994 specifically to maintain open standards — to prevent any single company from capturing the web's protocols. That effort partially succeeded (HTML, CSS, HTTP remain open standards) and partially failed (the application layer built on top of those standards was captured entirely by private corporations).

By 2017, Berners-Lee was publicly stating that the web had gone wrong. He identified three problems: the loss of personal control over data, the spread of misinformation, and the lack of transparency in political advertising. In 2018, he launched the Solid project — a protocol for decentralized data storage that would give users control over their own information. As of 2025, Solid remains niche.

The pattern is consistent: the protocol layer stayed open. The application layer was captured. The business model layer was captured even more thoroughly. Every meaningful attempt to re-decentralize the web has struggled against the gravitational pull of capital concentration.

The theoretical framework here comes from multiple sources, but three are essential.

Tim Wu — The Attention Merchants (2016). Wu traces the business model of selling human attention back to the 1830s, when Benjamin Day's New York Sun discovered that selling newspapers below cost and making money on advertising created a more profitable model than subscription. Every subsequent media innovation — radio, television, social media — followed the same logic. The content is the bait. The attention is the product. The advertiser is the customer.

Shoshana Zuboff — The Age of Surveillance Capitalism (2019). Zuboff went deeper. The shift isn't just from subscription to advertising. It's from advertising to behavioral prediction. Google discovered, around 2001, that the "behavioral surplus" generated by search queries — the data about your intentions, desires, and patterns that exceeded what was needed to improve the search product — could be used to predict your future behavior. That prediction product is sold to advertisers. The better the prediction, the more valuable it is. This creates a structural incentive to extract as much behavioral data as possible and to manipulate the environments in which behavior occurs to make behavior more predictable.

Tristan Harris and the Center for Humane Technology. Harris, a former Google design ethicist, articulated the mechanism at the product level. Features like infinite scroll, pull-to-refresh, notification badges, autoplay, and recommended content are not neutral design choices. They are engagement-maximizing mechanisms designed to exploit psychological vulnerabilities — variable ratio reinforcement (the slot machine effect), social validation feedback loops, and fear of missing out. These features don't serve the user. They serve the advertising model.

The result is an information environment optimized for arousal, not accuracy. For division, not connection. For addiction, not understanding.

The research on social media's effects on political polarization, trust, and social cohesion is now extensive.

Bail et al. (2018, PNAS). A randomized experiment exposed Twitter users to opposing political content via bots. Counter to the "exposure hypothesis" (which predicts that seeing the other side reduces polarization), exposure to opposing views increased political polarization, particularly among conservatives. The mechanism appears to be reactive: encountering content that triggers identity threat produces defensive hardening rather than understanding.

Allcott et al. (2020, American Economic Review). A large-scale experiment paid Facebook users to deactivate their accounts for four weeks before the 2018 US midterm elections. Deactivation reduced political polarization, increased subjective well-being, and reduced news consumption. The effects were modest but significant and persistent.

Guess et al. (2023, Science). A series of large-scale experiments conducted in partnership with Meta found that removing reshared content (a key mechanism for viral spread) from Facebook feeds reduced exposure to political news but did not significantly change political attitudes or polarization in the short term. The relationship between platform design and polarization is real but more complex than simple causation.

The Myanmar case. A UN fact-finding mission in 2018 documented that Facebook played a "determining role" in the spread of anti-Rohingya hate speech that contributed to the genocide of Rohingya Muslims. Facebook was, for many Myanmar users, the internet itself — and its content moderation infrastructure was grotesquely inadequate for the Burmese-language context. The platform's recommendation algorithm amplified content based on engagement, and hate speech was extremely engaging.

The concept of digital colonialism, as articulated by scholars like Sareeta Amrute, Nick Couldry, and Ulises Mejias, refers to the extension of colonial dynamics into the digital sphere.

The structure: platforms designed in Silicon Valley (or, in TikTok's case, Beijing) extract behavioral data from global populations, process it in data centers controlled by those companies, use it to generate profits that flow back to shareholders in wealthy countries, and impose content governance frameworks designed for American cultural contexts on populations worldwide.

Specifics:

- Free Basics (Meta). Facebook's initiative to provide "free internet access" in developing countries offered access to a curated set of services — with Facebook at the center. India banned it in 2016 on net neutrality grounds, correctly identifying it as a mechanism for platform capture disguised as philanthropy.

- Google's Loon project (now defunct) proposed providing internet access in remote areas via high-altitude balloons — controlled entirely by Google. The access was free. The data extraction was not.

- Submarine cables. Google, Meta, Amazon, and Microsoft now own or co-own the majority of new undersea fiber optic cables being laid globally. The physical infrastructure of the internet is increasingly privately owned by the same companies that dominate the application layer.

The result is that much of the world's population interacts with an internet that was not designed for them, is not governed by them, and extracts value from them without meaningful consent or compensation.

The Web3 narrative, at its best, addresses real problems: centralized control, data extraction, platform capture, lack of user ownership. The proposed solution: build platforms on blockchain, give users ownership via tokens, replace corporate governance with decentralized autonomous organizations (DAOs), use cryptographic identity rather than platform-controlled accounts.

The reality, as of 2025:

- Centralization has reproduced itself. The largest Web3 platforms (OpenSea for NFTs, Uniswap for decentralized exchange, Aave for lending) are controlled by venture capital-funded teams. Token distributions are heavily concentrated among early investors and insiders. The Gini coefficient for most token-governed protocols is worse than the US income distribution.

- The user experience is prohibitive. Interacting with Web3 protocols requires managing private keys, understanding gas fees, navigating multiple chains, and accepting the risk of irreversible errors. This is accessible to technically literate early adopters. It is inaccessible to the vast majority of humans.

- Speculation drowns utility. The dominant use case of Web3 to date has been financial speculation — token trading, NFT flipping, yield farming. The mutual aid and knowledge-commons use cases exist but are marginal compared to the speculative economy.

- Governance challenges. DAOs (Decentralized Autonomous Organizations) were supposed to enable democratic governance of platforms. In practice, token-weighted voting means that governance power is proportional to wealth — plutocracy with a crypto aesthetic.

There are genuine exceptions — Gitcoin's quadratic funding for public goods, Protocol Labs' work on decentralized storage (IPFS/Filecoin), and Ethereum's shift to proof-of-stake (reducing environmental impact). But the honest assessment is that Web3 has not yet delivered on its unity-enabling promises.

1. Wikipedia / Wikimedia. Revenue model: donations. Governance: volunteer editors with elected administrators and an Arbitration Committee. Content: 60+ million articles in 300+ languages. Wikipedia is the clearest proof that a large-scale knowledge commons can function without advertising, algorithmic manipulation, or corporate ownership. Its problems (editing bias toward Western, English-language sources; underrepresentation of Global South knowledge; internal bureaucracy) are real. But its existence as the sixth most-visited website, running on a budget of approximately $170 million/year (smaller than a single department at Google), is a civilizational achievement.

2. The Fediverse. The Fediverse (federated universe) is a network of interconnected social media platforms using the ActivityPub protocol. Mastodon is the most well-known, but the Fediverse also includes Lemmy (Reddit-like), PeerTube (YouTube-like), PixelFed (Instagram-like), and others. Each server (instance) is independently operated, sets its own moderation policies, and can connect to or disconnect from other servers. There is no central algorithm. No engagement optimization. No advertising. The trade-off is that content discovery is harder and the network effects are weaker. The benefit is that the incentive structure doesn't push toward outrage.

3. Community mesh networks. Guifi.net in Catalonia has been operating since 2004 and now has over 35,000 active nodes, making it one of the largest community-owned telecommunications networks in the world. It operates as a commons: participants contribute infrastructure (antennas, routers, fiber) and share bandwidth. The network is governed by a commons license (the Compact for a Free, Open & Neutral Network). Similar projects exist in Detroit (Detroit Community Technology Project), Red Hook in Brooklyn (Red Hook WiFi), and dozens of cities in Latin America, Africa, and Asia.

4. Civic tech platforms. Decidim (from Barcelona) is an open-source participatory democracy platform used by over 400 organizations worldwide. It enables participatory budgeting (citizens decide how to allocate public funds), collaborative legislation drafting, and public deliberation. Taiwan's vTaiwan platform, built on similar principles, has been used to successfully resolve contentious policy issues (like Uber regulation) through structured online deliberation.

When assessing whether a digital platform or protocol serves human connection or undermines it, apply these five tests:

| Test | Question | Unity-serving | Unity-undermining | |------|----------|---------------|-------------------| | Incentive | What is the platform optimized for? | Knowledge sharing, connection, deliberation | Engagement, time-on-site, ad revenue | | Governance | Who controls the rules? | Users, communities, democratic processes | Corporate board, shareholders | | Data | Who benefits from user data? | The user, the commons | The platform, advertisers | | Architecture | Centralized or federated? | Decentralized, interoperable | Walled garden, lock-in | | Access | Who can participate? | Open, low barrier, multilingual | Requires smartphone, data, English |

No existing platform scores perfectly. But applying this framework consistently reveals which projects are moving toward the original web vision and which are moving away from it.

Step 1: For one week, track where your online time goes. Don't change your behavior — just log it. How much goes to ad-funded platforms? How much to non-commercial spaces?

Step 2: Identify one ad-funded service you use daily and research a non-commercial alternative. You don't have to switch. Just know the alternative exists. (Mastodon for Twitter/X. PeerTube for YouTube. DuckDuckGo for Google Search. Signal for WhatsApp. Proton for Gmail.)

Step 3: Contribute to one digital commons. Edit a Wikipedia article. Donate to the Wikimedia Foundation or the Internet Archive. Join a Fediverse server. If there's a community mesh network in your area, participate.

Step 4: Ask the question that matters. If the web were rebuilt today, from scratch, with the explicit goal of connecting humanity rather than extracting value from it — what would you build? What would you fund? What would you use?

The original web vision isn't dead. It's dormant. The protocols for open knowledge sharing exist. The models for non-commercial social connection exist. The governance frameworks for digital commons exist. The physical infrastructure models for community-owned networks exist.

What's missing is the same thing that's always missing: the political and economic will to build shared infrastructure instead of private empires.

If every person said yes to our shared humanity, the web would look more like a library and less like a casino. The blueprints are there. The question is whether we'll use them.