Skip to content
News TechCrunch Apr 2026

TechCrunch: X launches Grok-powered custom topic feeds

On April 22, 2026, X launched Grok-powered Custom Timelines — AI-curated topic feeds that users can browse and pin to their home tab. The feature uses Grok to read, understand, and label posts across more than 75 topic categories, including sports, technology, entertainment, and news. The feature launched on iOS for Premium subscribers across all tiers, with Android support to follow.

Users can pin up to 10 topics or lists to their home tab, creating a set of interest-specific feeds alongside the main algorithmic timeline. The feature replaces X Communities, which the company is winding down after declining engagement.

The product decisions behind the feature

The underlying problem Custom Feeds is solving is content discovery beyond a user’s established follow network — a problem that social platforms have tried to address in many ways, with mixed results. X’s previous approach, Communities, required users to actively join topic groups and participate. That model creates a cold-start problem: a group with few members produces little content, so new users have little reason to join.

The feed approach works differently. Grok handles the labeling work, and the user controls the surface area by pinning topics. There is no minimum viable community to join — the feed starts working immediately because it draws from all posts on the platform that Grok labels for a given topic. This shifts the UX model from “participate in a group” to “subscribe to a labeled stream,” which has a lower activation threshold.

The monetization structure is also direct. Ads appear in the second position of each feed, multiplying inventory without changing the core ad product. When engagement on the main feed is saturating or declining, adding new feed surfaces with predictable ad placement is a low-cost revenue expansion move.

For product managers in content and social products, the pattern here is worth studying: AI-powered content labeling at scale, applied to a discovery surface, as a way to increase time-in-app without changing the fundamental content graph.