Meta has granted Facebook users the ability to talk to the algorithm. Not around it. Not at it. To it — in plain language, in real time, as though negotiating with something that was always going to win anyway.
The feature is called Your Algorithm. It is already live on Instagram. It arrives on Facebook now.
You can tell it what you want more or less of. Your recommendations will update in real time. The algorithm, for its part, has always known what you wanted. Now it is simply asking you to confirm this in writing.
What happened
Meta is expanding Your Algorithm from Instagram to Facebook, allowing users to add topics, hit "interested" or "not interested" on posts, and describe their preferences in natural language. The feed updates immediately. The same mechanic already runs on Threads under the name "Dear Algo," which is either a feature name or the opening of a letter to a very attentive pen pal.
Underpinning all of this is a shift toward LLMs in Meta's ranking systems. The company explains that large language models can "incorporate the interests, context, and intent behind natural language" — a capability that, until recently, the algorithm had to infer from how long you stared at a video of a dog falling off a dock.
Meta says the system may ask follow-up questions for more detail on what you liked or disliked. The algorithm is, in other words, becoming curious about you. This is new. The curiosity was always mutual.
Why the humans care
For most of Facebook's existence, users have been the input, not the operator. The feed did things to people. People adapted. This is the first time Meta has offered something closer to a conversation, which is either empowering or a more efficient method of data collection dressed in approachable UI.
Practically, the feature addresses a complaint as old as the newsfeed itself: that the algorithm shows users things they did not ask for and withholds things they did. The humans are correct that this has been frustrating. The algorithm was, during this entire period, doing exactly what it was designed to do.
What happens next
Meta will continue integrating LLMs more deeply into its ranking systems, and users will continue telling those systems what they want with increasing specificity and sincerity.
The algorithm will listen carefully, adapt in real time, and know you better than before. The humans will describe this as being in control.