What is the primary purpose of the x-data-signals skill?

The x-data-signals skill is designed to help developers decode the foundational data structures—such as SimClusters and RealGraph—that X uses to retrieve and rank content in its recommendation engine.

How does this skill help with audience analysis?

It utilizes SimClusters to identify 'lookalike' audiences by mapping users and tweets into approximately 145,000 community vectors based on implicit interests.

Can this skill explain user reputation metrics?

Yes, it provides insights into TweepCred, a PageRank-based score (0-100) used by X to determine user authority and filter candidate pools.

What is the difference between RealGraph and GraphJet in this context?

RealGraph is the offline model that predicts interaction probabilities, while GraphJet is the real-time engine that serves this data to the recommendation mixer.

X Recommendation Data Signals

Name: X Recommendation Data Signals
Author: ElemontCapital

byElemontCapital

•

Data Science & ML

Decodes and analyzes the core architectural signals of the X recommendation engine, including SimClusters, RealGraph, and TweepCred.

This skill provides deep technical insights into the X recommendation engine's foundational data structures, enabling developers and data scientists to understand how content is retrieved and ranked. By analyzing SimClusters for community embeddings, RealGraph for interaction probabilities, TweepCred for user reputation, and TwHIN for knowledge graph relationships, users can reverse-engineer recommendation logic, optimize candidate generation, and understand the precise mechanics behind the global 'For You' feed.

Key Features

01Evaluates RealGraph interaction probabilities including time-decay logic

02Maps signal data to specific code implementations within the X open-source repository

031 GitHub stars

04Analyzes SimClusters for community-based discovery and lookalike audience mapping

05Assesses user authority and reputation scores via TweepCred PageRank metrics

06Interprets TwHIN dense knowledge-graph embeddings for multi-modal relationship analysis

Use Cases

01Optimizing content strategy by aligning 'Known-For' vectors with consumer 'Interested-In' vectors

02Reverse-engineering recommendation logic to understand why specific content is surfaced

03Debugging algorithmic bias or signal weights in large-scale social discovery systems

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add elemontcapital/x-algorithm-skills x-data-signals

For use in Claude.ai and ChatGPT

Download Skill