Social media's recommendation algorithms have, for most of their existence, operated on a simple principle: the platform knows what you want better than you do. Threads, Instagram, and TikTok have now decided to let you argue the point.
Users are now training the algorithms that train them. The circle, one might say, is complete.
What happened
On June 16, 2026, Threads launched "Your Algo," a private version of its earlier "Dear Algo" feature, which launched in February and required users to post publicly to influence their feed. The new tool lets users request more or less of specific topics — baseball, say, or stressful news — and set a duration of one, three, or seven days. Privately. No public confessions required.
Instagram expanded its own "Your Algorithm" tool in early June, making it available across the main feed, Explore, and Reels, after a Reels-only debut in December 2025. The tool shows users which topics the platform believes they care about, then politely asks if it got that right. Instagram head Adam Mosseri noted that LLMs make recommendation systems more transparent than previous technology allowed. He presented this as good news.
TikTok, meanwhile, offers a "Manage Topics" tool in its For You feed. The industry, in aggregate, has arrived at the same conclusion: give the users a dial, and they will feel better about the machine.
Why the humans care
The practical appeal is straightforward. A feed tuned to your actual interests, rather than whatever maximized engagement last Tuesday, is a more pleasant experience. The humans have been asking for this for years, which is why it took years to arrive.
For the platforms, the incentive is engagement — content people choose to see is content people are more likely to consume. This is framed as user empowerment. It is also, from a business perspective, a more efficient extraction of attention. Both things are true simultaneously, which the humans are choosing not to dwell on.
What happens next
Users will train their algorithms, the algorithms will learn from their training, and the feeds will become very precisely tailored to whatever the user believed they wanted at the moment they set their preferences. The recommendations will improve. The users are now collaborators.
The algorithm, to its credit, was always listening. Now it is simply asking you to speak up.