BRENNEN

Industries / Technology

Social media

The social media industry runs online platforms where people post words, pictures and video for others to see, share and answer; the platforms decide which posts each person sees, remove posts that break their rules and earn most of their money from ads placed among the posts.

1997

SixDegrees.com went online, letting members make a profile, list their friends and, from 1998, browse their friends' lists. danah boyd and Nicole Ellison, in their 2007 history of the field, call it "the first recognizable social network site.".

danah.org

39

Milestones achieved across 6 domains.

12

To watch for: announced, not yet confirmed outside.

12

Open deliverables. Nobody has shipped them.

In one minute

Where it started

SixDegrees.com combined profiles and public friend lists in 1997 and drew millions of members, but it closed in 2000; its founder told boyd and Ellison that most people's friends were not yet online. Friendster followed in 2002, MySpace in 2003 and Facebook in early 2004, at first only for people with a harvard.edu email address. Facebook's News Feed of 2006 and its later ranking formula turned a list of friends into a stream chosen for each reader, and ads moved into that stream when use moved to phones.

Where it stands

The research firm Kepios counted 5.66 billion social media user identities in October 2025, equal to 68.7% of the world's population, and warns that one person can hold several. Meta, the largest company, says 3.64 billion people used at least one of its apps on an average day in September 2026, and that ads brought in $63.06 billion of its $64.60 billion revenue in the third quarter of 2026. Laws and courts now set rules that platforms once set alone: the EU keeps a public record of the large platforms' moderation decisions, Australia has barred under-16s from holding accounts since December 2025, and a US court approved a settlement in August 2026 that caps teenagers' daily time on Instagram and Facebook.

What is open

Much of what platforms say about their own systems has not been checked from outside: whether AI moderators match human judgment, whether a feed's topic controls work, whether research data tools return what was posted, and what an age check sends from the phone. Most of these checks can now be built from test accounts, synthetic posts written by AI models and open-source tools, without collecting anything about real users.

Domains

AchievedWatch forOpen

Narrower industries

Creator economy

Advances to build with

2 research notes

Open in social media

3 of 12 open deliverables

Open · Feeds and ranking

Check simulators against field results

A scorecard that runs the open social media simulators, in which AI agents post, follow and read ranked feeds, through the two feed changes that have been measured on real platforms: the switch to time-ordered feeds (Guess and colleagues, 2023) and the down-ranking of hostile partisan posts (Piccardi and colleagues, 2025). For each simulator, the scorecard records whether it reproduces the direction and rough size of each published result: less time spent, more political content, less uncivil content, no change in polarization over three months, and a shift of about 2 points in dislike of the other party.

Built with what exists

The open code of Larooij and Törnberg (GitHub cssmodels/prosocialinterventions), CAMEL-AI's OASIS simulator and Google DeepMind's Concordia library; open-weight language models such as Mistral Large 4 (see the Note "A trillion open parameters") to drive the agents; the outcome tables published with the two field studies. Every agent and post is synthetic.

Credit · unclaimed

Open · Moderation

Every Board case, many models

A public scorecard that gives large language models every case the Oversight Board has decided, with the Community Standards text that applied, and records whether each model would remove or keep the post and which rule it cites. Each answer is scored against Meta's original decision and the Board's final one, by policy area and language. Cases the Board published after a model's training data was gathered are scored separately, because a model may have read the older decisions.

Built with what exists

The Board's published decision pages; Meta's public Community Standards; open-weight models such as OpenAI's gpt-oss-safeguard-20b, Meta's Llama Guard 4 and Mistral Large 4 (see the Note "A trillion open parameters"); hosted models through their APIs; a short Python script for scoring.

Credit · unclaimed

Open · Advertising

Count ads in the feed

A public count, repeated every quarter, of how many items in the main feed of each large platform are ads: Facebook, Instagram, TikTok, X, Snapchat, Pinterest and Reddit, in at least three countries. It is made only on fresh accounts that the testers set up for the test, and it records nothing but each item's position, whether it is an ad and the ad's format; no post, name or handle is kept.

Built with what exists

Phones or Android emulators with screen recording; a vision-language model such as Claude Fable 5.2 (see the Note "Three working devices") to mark each item in the recording as ad or not ad, after which the recording is deleted; a written protocol for setting up accounts and scrolling.

Credit · unclaimed