Alright alright alright

The bespoke social feed has two overlapping histories: users choosing what to read, and software learning what to show them. The distinction matters. A timeline can be unique to you because you assembled its ingredients, because a company predicted your interests, or because you selected someone else's recipe. Social networks have experimented with all three approaches for decades.

An early ancestor was LiveJournal's Friends page, which collected entries from journals a user followed. LiveJournal launched in 1999, long before Facebook's News Feed. Its custom friend groups let readers view subsets of their subscriptions: a practical way to separate different social worlds. This was bespoke feed generation through explicit selection. You chose the writers, and the service assembled their posts.

Facebook's News Feed, introduced in September 2006, made the personalized stream central to a much larger social network. It gathered friends' updates and activities into one place, replacing the need to visit profiles individually. The initial privacy backlash illustrated something consequential: collecting scattered information into a feed changes its social meaning, even when the underlying information is already accessible.

The now obscure FriendFeed explored another direction. It aggregated activity across services, and its 2008 redesign introduced separate streams for groups of people. A reader could move acquaintances outside the main stream. Meanwhile, LiveJournal's successor Dreamwidth developed reading filters for selected journals, communities and syndicated feeds. These systems treated personalization as organizing subscriptions, giving users multiple windows onto their social lives.

Twitter made that approach more portable with Lists in 2009. Users could assemble accounts around any category and let other people follow the resulting stream. In 2013, custom timelines in TweetDeck moved selection down to individual tweets. Creators could handpick posts, share the collection and embed it elsewhere. The feed could now be an editorial product, assembled by someone whose judgment you trusted.

TikTok exemplified a different kind of bespoke experience. Its 2020 explanation of the For You feed described recommendations informed by interactions, video information and account settings. Finishing a longer video could carry more weight than a weak signal such as sharing a country with its creator. Instead of requiring users to construct a subscription list, the system inferred preferences from behavior. Personalization became something continually negotiated through watching, skipping and responding.

Bluesky brought explicit choice into the algorithm itself with custom feeds launched in May 2023. Independent developers could supply feeds, while readers could choose and pin them. Examples included posts liked by people you follow and feeds devoted to particular subjects; tools such as SkyFeed lowered the barrier to making one. This joined earlier traditions: subscription filters, shared curation and automated discovery. An agent assembling a feed from your personal engagement history would extend that lineage. The enduring question is who decides what your behavior means: you, a curator you select, or the platform interpreting it.