Meta Tests Threads Friends, a 50-Account Audience for Posts
Meta is testing a feature called Threads Friends that lets users share posts with up to 50 friends without making those posts visible to anyone else on the platform. The test is running in Ireland, Italy and Japan, according to Engadget, which reported it on October 29, 2026.
Threads calls the list 'Thriends.' Meta automatically suggests followers that a user engages with often to fill the initial list, and the list can be customised. Once it is set up, a green bubble at the top of the feed shows posts from Friends shared in the last 24 hours.
Laura Javier, product design director for Threads, showed a preview of a friends-only post. The preview shows how the audience restriction appears to the author and to the people reading the post.
The audience is capped at 50 accounts, which is well below a public broadcast and closer to a small working group or a circle of close contacts. The 24-hour window on the Friends feed is a second boundary. The feed is a rolling view of recent activity, not an archive. Together, the two parameters describe a product built around immediacy and a bounded readership rather than durable, searchable posts.
Recommendation systems rank people and content using interaction data, such as how often a user likes, replies to or views an account. Meta's seed list draws on the same signal, so the default audience is inferred from behaviour rather than chosen deliberately. A user who accepts the suggestions unchanged will share with the accounts their engagement data already ranks highest. That saves the effort of building a list by hand. It also means the edge of a 'friends' post is set by behavioural signals, which a user may not see as clearly as they would a list they curated themselves.
In my view, the feature fills a gap the current Threads product leaves open. The app defaults to public, and its strongest distribution mechanics reward reach. A closed audience of up to 50 people offers a way to post something not meant for the whole network, without moving the conversation into direct messages or a separate app. Whether people use it that way will depend on how legible the audience is at the moment of posting, and that is the question the three-market test is positioned to answer.
The choice of Ireland, Italy and Japan matters for interpretation. The markets differ in language, in the typical size of personal networks and in how people use social platforms day to day. A feature that performs well across that spread has a stronger basis for wider rollout than one tested in a single country. Any expansion beyond these markets should be treated as an open question.
The broader context here is a platform trying to make a large public network feel smaller for the people in it. Friends-only sharing is an old idea in consumer software, and it has been rebuilt many times as platforms grew. What is new is the combination of a hard cap, an automatic seed list drawn from engagement data and a short feed window, all inside an app whose default is broadcast. If the test works, users get a way to talk to a smaller group without leaving the platform. The open question for engineers and privacy teams is how clearly that smaller group is shown to the person posting.
