Recommendations from people, not black boxes

Recommendations are better when you know where they came from. A book from a friend, a place from someone in your circle, or a list from a person whose taste you understand can be more useful than a ranking engine guessing what will keep you scrolling.

TouchGrass supports recommendations from people, not black boxes. The goal is discovery with context: books, shows, music, places, lists, and other recommendations connected to profiles, friends, and circles. TouchGrass is a safe, open social home for real life — without addictive feeds or platform lock-in.

Why TouchGrass made this choice

Recommendation systems often pretend to be neutral.

In practice, many are optimized for attention. They learn what makes people click, watch, react, stay, and return. That can be useful in small doses, but it also changes discovery into a contest for engagement. The system may not care whether something is meaningful, trustworthy, generous, local, personal, or worth keeping. It cares whether it moves the metric.

Human recommendations work differently.

A friend’s bookshelf is not just a ranked list. A place someone recommends carries context: why they liked it, who it might be good for, and what kind of day it fits. A music recommendation from someone you know is not only a content item. It is part of a relationship.

TouchGrass wants discovery to feel more like that.

This does not mean every algorithm is bad or every human recommendation is perfect. It means social software should not replace people with opaque ranking systems as the main way to discover culture, places, ideas, and memories.

Recommendations should be connected to chosen people and social context. They should be easy to save, find, share, and revisit without turning discovery into an attention machine.

What TouchGrass does today

Today, TouchGrass supports recommendations as part of the social product direction. Depending on the current feature status, recommendations may include books, shows, music, places, lists, or related formats.

The important design choice is that recommendations belong with people. They can appear on profiles, connect to friends and circles, and sit alongside posts, photos, albums, and longform writing. This gives recommendations context instead of treating them as anonymous feed units.

TouchGrass also supports visibility choices. A recommendation might be public, shared with friends, or shared with a circle, depending on available controls and current feature status. That matters because recommending a neighborhood place to close friends is different from publishing a public list.

TouchGrass avoids building discovery around engagement traps. It does not center videos, infinite scroll, public like-count races, or algorithmic feed ranking as the main product experience.

The open-web direction also matters here. Portability and federation can make recommendations less trapped inside one platform. Work around ATProto/Atmosphere, Solid, ActivityPods, private sharing, and related protocol directions should be stated carefully. Some pieces may be shipped, partial, experimental, or planned. The feature status page should be the source of truth.

Limits / what not to overclaim

A recommendation from a person is not automatically correct, safe, or unbiased. Friends can be wrong. Tastes differ. Lists can become outdated. A personal recommendation can still need judgment.

TouchGrass should also not claim that every recommendation format or protocol integration is fully shipped unless the feature status page says so. The safer language is that TouchGrass supports and is building around recommendations from chosen people, with the exact feature set documented in status.

Private or circle-based recommendations also have privacy limits. If you share a recommendation with someone, they may screenshot it, copy it, repeat it, or misuse it. TouchGrass cannot promise protection from recipient behavior, remote-server behavior, or every future integration.

Finally, open does not mean public. Open-web support is about portability, interoperability, and exits. Visibility controls are a separate part of the product.

FAQ

Why not just use algorithmic recommendations?

Algorithmic recommendations can be useful, but many are optimized to maximize attention. TouchGrass is focused on recommendations with social context, where the source is a person, profile, friend, or circle you understand.

What can people recommend on TouchGrass?

TouchGrass is built around recommendations such as books, shows, music, places, and lists, depending on the current feature status. Check the recommendations page and feature status page for the exact live scope.

Can recommendations be private?

TouchGrass is designed around public, friends, and circles visibility. Current controls should be checked before making a specific claim. Also, private sharing cannot protect against screenshots, recipient misuse, or all remote-server behavior.

Are recommendations portable?

Portability is part of the TouchGrass open-web direction, but the current status may vary by feature and protocol. Check the feature status page before saying recommendations are fully portable or federated.

Does TouchGrass reject all ranking?

The point is not to reject every form of ordering. The point is to avoid making opaque engagement ranking the center of discovery. Recommendations should keep human context.

Learn more about recommendations: /recommendations-from-friends

Feature status: /status/features

Last updated: 15 May 2026

Language: English