Tag: Social Networks

Posted on by

Contact graphs that cannot move cleanly between services

You can export a phone number.

You cannot export the fact that twelve specific people are already in the family group chat and actively use it every day.

That is the portability gap between contacts and relationships.

A contact list is a set of identifiers.

A contact graph includes who follows whom, which groups exist, who belongs to them, who has blocked whom, which conversations are active, which memberships are mutual, and which people actually still use the service.

Those things are much harder to move.

Data portability does not automatically recreate a network

The European Union’s Digital Markets Act now requires designated gatekeepers to provide stronger portability tools for user-generated data. The Commission describes Article 6(9) as requiring free tools and continuous access so users can move data to authorized third parties. See the Commission’s end-user data portability guidance.

That is a significant improvement over a platform merely offering a giant archive download.

But even a perfect export cannot force the people in the file to create accounts somewhere else.

A social graph is partly data and partly coordinated human behavior.

Messaging interoperability shows another approach

The DMA treats messaging differently because strong network effects make migration unusually difficult. Instead of requiring every user to move, Article 7 requires certain gatekeeper messaging services to interoperate with other providers.

The European Commission says the first third-party messaging services interoperable with WhatsApp launched in November 2025. The rules are being phased in across one-to-one messaging, group chats, and eventually calls. See the Commission’s messaging interoperability page.

That is conceptually important.

If two services can communicate, a user can leave one interface without necessarily abandoning everyone still using it.

The solution attacks the network lock-in directly.

Exportable is not the same as portable

A platform can truthfully say:

You can download your data.

That may still leave the user with:

  • a CSV of contacts who are not on the new service,
  • old messages with no live destination,
  • groups that must be rebuilt manually,
  • followers who must choose to follow again,
  • community roles that do not transfer,
  • and years of relationships represented only as static records.

The file moved.

The network did not.

That distinction matters whenever a dominant platform points to an export tool as proof that users are free to leave.

Freedom to leave is not measured only by whether a button produces a ZIP archive.

It is measured by how much useful social structure survives the trip.

Posted on by

Small federated communities as alternatives to mass public platforms

A social network does not need one giant front door.

Federated systems prove that by splitting the network across many independently operated communities that can still talk to one another.

Mastodon’s documentation defines federation as a form of decentralization in which multiple services interoperate instead of everybody using one central service. A Mastodon server can operate on its own, but users on different servers can follow, mention, and interact with one another across the network. See Mastodon’s explanation of federation.

That creates a very different map of activity from a single mass platform.

The local community still matters

Federation does not erase the importance of the server somebody joins.

Mastodon exposes a local public timeline containing posts from people on the same server, alongside broader timelines that can include activity known from elsewhere in the network. Its API documentation explicitly distinguishes local and remote public statuses. See Mastodon’s timeline documentation.

On a small server organized around a city, profession, hobby, language, or shared culture, the local timeline can behave like a neighborhood inside the larger network.

The person gets two things at once:

A smaller community with its own administrators and norms, and a route to people outside it.

That is not the same social design as joining a platform with hundreds of millions of users under one moderation system and one corporate identity.

Distributed activity can look like disappearance

This matters for Dead Internet Theory because familiar platform counts can become misleading when users disperse.

Suppose 50,000 people leave one large service and spread across 500 independently operated servers.

From the perspective of the large service, those users disappeared.

From the perspective of any one small server, the audience may look tiny.

Across the federation, however, the humans are still publishing and interacting.

The activity became harder to summarize with one homepage, one corporate user count, or one trending list.

Small does not automatically mean healthier

Federation is not magic.

Small servers can be badly moderated. Administrators can disappear. Servers can shut down. Communities can become insular. Federation policies can limit who communicates with whom. Technical maintenance still has to happen somewhere.

But those weaknesses are different from absence.

The Human Retreat is partly about people leaving large public stages for spaces that feel more governable.

Federated networks add an important variation: the smaller room does not always have to be isolated.

Sometimes the crowd did not vanish.

It broke into neighborhoods and kept the roads open.

Posted on by

Chronological feeds compared with ranked feeds

A chronological feed answers a simple question:

What happened most recently among the accounts this feed includes?

A ranked feed answers a different one:

What does the platform think you are most likely to want to see next?

Those two questions can produce dramatically different versions of the same social network.

X provides a current, unusually clear example. Its help documentation says the For You timeline mixes posts from accounts a user follows with recommended content selected using signals such as interests, popularity, and activity within the user’s network. Its Following timeline instead shows posts only from followed accounts in reverse chronological order. See About your For You timeline and X’s documentation on For You recommendations.

Neither view is imaginary.

They are different selection rules applied to the platform.

Ranking can rescue old material and bury ordinary material

A ranked feed can surface a post hours after publication because the system predicts that it matters to the user. It can introduce accounts outside the user’s network, recover a conversation that would have vanished down a fast timeline, or prioritize a reply judged especially relevant.

The same machinery can make ordinary posts effectively disappear.

If thousands of accounts publish between visits, a ranked feed may choose a tiny subset. The user experiences those selected posts as “what happened,” even though they are really what happened after ranking.

A chronological feed removes that particular relevance layer.

It does not remove selection.

Chronological order is not the whole network either

Reverse chronology still begins with a chosen set of accounts. If you follow fifty people instead of fifty thousand, your world remains narrow. Reposts can inject older material. Promotions may still appear. Muted or blocked accounts disappear. Moderation and platform policies still affect availability.

And chronology has its own bias: volume wins.

Someone who posts forty times a day occupies more timeline space than someone who posts once a week, regardless of importance. A fast-moving feed also punishes anyone who was offline at the wrong hour.

So chronological order is not an objective cure for algorithmic reality. It is a different algorithm with a very legible rule: newer first.

That legibility has value.

Users can understand why one post appears above another without needing to infer a relevance score.

Ranked feeds trade some of that transparency for prediction and discovery.

The useful comparison is therefore not algorithm versus no algorithm.

It is which selection rule is deciding what counts as your internet right now?

Posted on by

Fotolog: the social culture of a daily photograph

Fotolog asked for one photograph a day. The site went live in May 2002 as an experiment among friends, built by a small New York crew whose early FAQ described it as “Blogger for people that don’t write well” and pushed a personal “photo of the day” rather than an online album — get the latest, greatest image up, not the whole memory card. When the free tier later limited members to a single upload per day, that rule shaped everything that followed.

One photo made the choice matter

A daily limit turns posting into a ritual instead of a reflex. Free members had to pick the one image that would stand for the day, and their archive then grew day by day like a visual diary — the FAQ promised you could watch “how the baby’s grown” from a distance. Paying members could post more, but the shared culture ran on scarcity. Open a friend’s page and you saw one latest photo, with a guestbook beneath it. The daily beat gave people a reason to check in, and the format let them catch up cheaply: everyone you followed, one click away.

The guestbook kept people coming back

The comments lived in a capped guestbook. Free accounts in 2006 had a small daily comment allowance, so a visit early in the day carried real weight. Because photographs cross language, the notes turned international, with comments crossing Portuguese, Spanish, French and English and strangers finding friends on the other side of the world. Growth centered heavily on Latin America. By the time Hi-Media bought the company in 2007, Fotolog reported about 10 million member accounts and 3.3 billion monthly page views, with Latin America remaining central to its audience.

The unlimited feed ended the ritual

The decline was less about features than about cadence. French advertising company Hi-Media bought Fotolog for about $90 million in 2007, but members were already drifting to Orkut and Facebook, which never rationed posts, and after 2010 Instagram offered the same one-image-at-a-time feel without the daily tax. An unlimited feed rewards volume and constant attention; a daily photo rewards choice and return visits. Against that economics, the ritual lost. Fotolog announced a possible shutdown in early 2016, remained online after the notice disappeared, and ultimately shut down in March 2019. The domain was later sold to new owners and became an unrelated magazine.

The original site is a ghost town, but not an erased one. Archive Team and the Wayback Machine preserved countless profiles, and separate revivals have since tried to rebuild the one-photo-a-day format. The community that grew around a capped daily strategy outlived the platform that hosted it.

Posted on by

Friends Reunited: the limits of a reunion network after reconnection

Friends Reunited began with a question no British site had thought to answer: whatever happened to everyone from your school year? In 1999 Julie Pankhurst found herself wondering about her old classmates, and she, her husband Steve, and their friend Jason Porter started building. The site launched in July 2000 from the couple’s spare room in Barnet. It grew with startling speed; by December 2005 the number of registered members had passed 15 million, making it one of Britain’s first mass social networks.

The product was a lookup tool with a message form

Pick your school, pick your year, find the names. Contacts ran through a double-blind email system, so you could reach someone without handing over your address. Full membership cost ÂŁ7.50 a year. It was Classmates.com’s idea imported across the Atlantic, but a small country where most people had attended a handful of local schools made it spread far faster.

Reconnection spent the reason to return

The structural problem was that a reunion is a one-time transaction. Find the person, swap a message, done. Every successful lookup erased one more reason to log in again. The site sold a moment, the reconnection, and had no answer for what came after it. The friendships it restored mostly continued elsewhere or not at all.

The ITV years

In December 2005 ITV paid ÂŁ120 million, with up to ÂŁ55 million more tied to performance. The site was profitable at first, but its UK traffic grew only 1.2 percent in 2007 while Facebook’s approached 2,400 percent. The ÂŁ7.50 fee was dropped in 2008 after a 47 percent fall in unique users, and the decline continued.

The long exit

ITV sold the bundle, reunion site, dating arm, and the Genes Reunited genealogy service to DC Thomson’s Brightsolid for ÂŁ25 million in 2009, a price that mostly reflected the family history it carried. A 2012 relaunch tried nostalgia and memories instead of finding classmates. In January 2016 Steve Pankhurst, invited by the new owners to take his old site back, concluded the trial wouldn’t work and announced the end. Friends Reunited shut down on 26 February 2016, sixteen years after launch.

What a reunion network is for

The trajectory isn’t only about Facebook. Reunion networks sell a lookup at a particular life moment; they give the people they reconnect no reason to stay around afterward. Friends Reunited wasn’t defeated so much as outlived by its own design, a directory quietly emptying itself one successful reunion at a time. What endured was the genealogy service its owners kept, while the school rolls faded into a ghost city preserved for occasional visitors.

Posted on by

Bebo: whether a familiar brand can bring back an old social graph

Bebo launched in January 2005 from Michael and Xochi Birch’s home in San Francisco. The domain came first, the meaning later: they invented the backronym “Blog early, blog often” to answer a question nobody had asked. In Britain, Ireland, and New Zealand the site stopped being a name and became a place. At its peak it was the most-used social network in the UK, overtaking MySpace with over ten million users.

A graph you could rearrange

What people remember is the texture: profile skins, a whiteboard for friends to draw on, quizzes and polls, the small purple “luv” hearts. What held it together was the Top 16, the ranked list of closest friends, reordered with the delicacy of peace negotiations, plus an “Other Half” slot. Bebo made the social graph public and editable. That list of names was the real product, and it is the one component a revived website can never rebuild from scratch.

Every revival recovered the shell

In March 2008 AOL paid $850 million for Bebo. Five years later, after another ownership change and bankruptcy proceedings, Michael Birch bought the brand back for about $1 million. The numbers alone show how thoroughly the original business had collapsed. Then the brand began to wander: a 2015 messenger called Bebo Blab, streaming software, an esports detour sold to Twitch in 2019, and a 2021 relaunch that never left beta.

The 2026 return actually ran

In April 2026 another all-new Bebo appeared, and this time it ran. It brought back Slambooks, Skins, and three daily Loves, arranged into Pods that members hosted and moderated, and pitched itself as small private circles instead of a broadcast feed. Some of the old crowd did come home, enough that Birch thanked “the 2026 community” for proving “the spirit of Bebo was never the problem.” On September 1, 2026, amid a handover to his next app, Bluebell, the site closed, and bebo.com now reads “Thank you, and goodbye.”

The brand is not the graph

The pattern is consistent. Each revival recovered the trademark, the logo, the features; none recovered the actual relationships, long since moved into group chats and private feeds. Birch has drawn the lesson himself in his farewell essay: “you can return to an old idea, but not to the moment that made it successful.” A familiar brand can coax ghosts back for a visit. It cannot make them neighbors again.

Posted on by

Orkut: the breakup of a social network centered on Brazil and India

Orkut launched in January 2004 as Google’s first experiment in social networking: a “20 percent project” named after its creator, engineer Orkut BĂĽyĂĽkkökten, running by invitation. In the United States it never mattered; MySpace and then Facebook crowded it out. Elsewhere it outgrew Google’s ambitions: by mid-2004 invitations were spreading through Brazil faster than any marketing campaign could.

The network Brazil and India took over

By June 2004 Brazilian profiles already outnumbered American ones. Two years later The New York Times put the scale in plain terms: roughly 11 million of Orkut’s 15 million users were registered as living in Brazil. In 2008 Orkut ranked among the most-visited sites in Brazil and India, and Google moved day-to-day operations to Google Brazil in Belo Horizonte. In India, it was many people’s first online social space and worked well on slow connections. By 2010, fully half of Orkut’s more than 100 million members were Brazilian and about a fifth were Indian.

Communities, scraps, and testimonials

The heart of the site was the community, not the profile. Members joined groups for their city, school, sports team, political party, or an inside joke, the largest reaching hundreds of thousands of members. On top of that sat the scrapbook, where anyone could leave a public message on a friend’s profile, alongside testimonials — short texts friends wrote about you that you approved to keep. These were the social signals of the day: how many scraps you had, and who had written a testimonial about you. At closure Orkut held some 51 million communities and 120 million discussion topics.

Closure and migration

Facebook overtook Orkut in India in 2010 and in Brazil by 2012, and Google had already begun directing users toward its new Google+. On June 30, 2014, engineering director Paulo Golgher posted a farewell: the service died on September 30, with profile and community posts downloadable through September 2016 and all public communities preserved in a Google archive. The departure meant different things in each country. In India it closed the era of “Orkutiyas” and scrapped the first friendships built online. In Brazil it made Orkut a national memory — a shared room where strangers became friends and cousins were found.

The second death

Google had promised the public community archive would let the network’s communities “live on.” It was deactivated in 2017 with about two weeks’ notice, a second death of the thing users actually cared about. Yet the community was not simply erased. The Wayback Machine and Archive Team saved profiles, communities, and threads, and researchers have recovered well over a hundred thousand community names from those captures. And today orkut.com opens on a letter from BĂĽyĂĽkkökten himself, promising a new network built on the old idea that a friendship is more than a friend request.

Posted on by

Friendster: social-network collapse and the attempted gaming reinvention

Friendster launched in March 2003 with a simple promise: put your real name on a profile, mark your actual friends, and let the site make the introductions. Jonathan Abrams, a former Netscape engineer, had watched people treat dating sites like anonymous classifieds and thought meeting through people you already trusted might work better. The signature feature was the “circle of friends.” Browse anyone’s profile and Friendster showed the chain of mutual connections that tied that person to you, so a stranger arrived with context instead of a screen name. Friends left testimonials vouching for one another, and within months the site counted millions of users. By early 2004 it was the most-trafficked social network on the web.

Friendships at the speed of the database

The collapse was partly technical. Every page view recomputed a user’s entire social graph, and the bigger the network grew, the slower it ran; pages that took tens of seconds to load sent people to MySpace, which overtook Friendster in April 2004 and never looked back. The company also deleted the “fakesters,” the joke group profiles users created, treating early internet culture as a policy problem. A rejected $30 million Google offer and a founding CEO pushed aside in 2004 rounded out the American slide. Friendster did not die: it kept tens of millions of registered users in the Philippines, Malaysia, Singapore, and Indonesia. But for its owners those people were now a market, not a community.

The gaming pivot, then the name returns

In December 2009, Malaysian payments firm MOL Global bought Friendster and moved fast. MOL’s business was prepaid game currency; Friendster was a distribution channel for it. In 2010 the patents went to Facebook for $40 million, and in June 2011 the site relaunched as a social gaming portal. Existing users kept their logins and contact lists, but photos, messages, testimonials, blogs, and group posts were stripped out behind an export tool. The pivot kept the people’s names and threw away the things that made them a community. When the gaming experiment stalled, Friendster shut down entirely in June 2015.

The name has been recycled since. In 2026, Friendster returned as a new social network advertising no ads, no algorithmic feed manipulation, and in-person friend linking by tapping phones together. It is a new service rather than a restoration of the old social graph or archive. The original graph was abandoned in 2011 and the old service in 2015; what returned in 2026 was the name and a new design philosophy. Ghost cities get redeveloped sometimes; the residents rarely come back.

Posted on by

MySpace: the dispersal of its independent music community

MySpace launched in August 2003 with musicians already at the center of it. One of the cofounders, Tom Anderson, had spent his earlier years in a failed San Francisco band, and he built the site around a simple idea: an unknown musician should be able to reach people who might like the music. Free pages, a built-in song player, and the ability to add strangers as friends turned a profile into a stage.

A front door for bands

It worked because it was specific. A band could post songs, list tour dates, answer messages, and choose their own handful of “Top Friends,” which doubled as a curated bill. A listener could embed that player on their own page and talk about the music like a local fan, not a customer. By late 2005, roughly 400,000 of the site’s 30 million pages belonged to bands, and label scouts browsed the traffic for new acts instead of waiting for demos in the mail.

Plenty of careers ran through those pages. Artists such as Arctic Monkeys, Lily Allen, and Panic! at the Disco built early followings partly by being findable there, and smaller bands booked tours through messages from people they had never met. By June 2014, users had uploaded more than 53 million songs from some 14 million artists.

The platform drifted from the music

The community grew into a business, and the business pulled away. News Corporation bought MySpace in July 2005 for $580 million and pushed advertising and audience growth; the free-form, HTML-everywhere pages gave way to template-driven redesigns aimed at people who found the chaos off-putting, and music became one feature inside an ad-heavy feed that Facebook was already doing more cleanly. In 2011, News Corp sold the site to Specific Media for $35 million. A 2013 relaunch backed by Justin Timberlake recast MySpace as a culture-news portal. It did not catch on, and the musicians’ front door quietly stopped being a destination.

The records went missing

The cost arrived in stages. First the older files quietly stopped playing; then, in March 2019, MySpace confirmed that a server-migration project had wiped out years of uploaded material. Contemporary reporting put the loss at more than 50 million songs from about 14 million artists, alongside photos and videos, and reported that the music was considered permanently lost. Some of it survives anyway: the Internet Archive’s Myspace Dragon Hoard contains roughly 490,000 MP3 files gathered by an anonymous academic study between 2008 and 2010.

Where the community went

The musicians and listeners did not disappear — they scattered. Bandcamp, which launched in 2008, filled part of the gap with a storefront where artists control pricing and keep a real share of sales. SoundCloud took over low-friction sharing, Facebook inherited the mass audience, and musicians added their own sites and email lists before streaming playlists and short-video platforms absorbed the discovery logic wholesale.

What was lost is not the audience but the address: a single place where a stranger could wander into a band and find its people waiting. The community survives, wider and thinner, spread across services that rent out parts of the old arrangement. The recordings that were never backed up are simply gone, which is the practical lesson of the whole episode. MySpace was a convenient home, not an archive, and the independent bands who treated it as the latter learned that the hard way.