What X Won't Show You (And Why That's OK)

The X algorithm repo is open source — but it's not complete. The README explicitly states that some things are withheld:

Deliberately Withheld

1. 1.LLM prompts used by content-classification systems — the prompts that drive automated labeling are not published
2. 2.Some abuse-detection rules — specific rules that would allow bad actors to evade detection are not included

Why This Is Good

X is transparent about *why* these are withheld: to reduce gaming. If every abuse-detection rule were public, bad actors could engineer around them. By withholding the specifics, the system is harder to game.

What This Means for You

• The tutorial is honest about its limits. We can teach you what's in the repo — and we're upfront about what isn't.
• You don't need the withheld info to grow. The ranking weights, sourcing systems, filters, and adjustments are all public. That's more than enough to work with.
• Trust is the moat. By being the source that says "here's what we know, here's what we don't, and here's the file that proves it" — we build more trust than sites that claim to know everything.

The Full List of What IS in the Repo

30+ services, mostly Rust:

• home-mixer/ — the core ranking pipeline
• thunder/ — in-network sourcing
• phoenix/ — out-of-network ML retrieval
• simclusters/ — community detection
• visibility-filtering/ — trust/safety
• under-the-hood/ — transparency tool
• botmaker/ — bot detection
• scarecrow/ — rule enforcement
• user-cred-v2/ — account credibility
• agatha/ — (classification system)
• clip/ — content understanding
• vm-ranker/ — value model ranking
• And more

This is more than enough to understand how the feed works. The parts that are withheld are the parts that would help bad actors — and they don't help you grow legitimately.