Enrich The trenches.

Enrich tools

XLens

X open-sourced its algorithm — half a million lines of Rust, Scala, and JAX across 26 production services. Almost nobody will ever read it. We did. Enter your topic — get a content plan built on the real ranking weights.

26modules read
150+rules extracted
100%cited to file & line

Doing this yourself means wading through 26 services of production Rust, Scala, and JAX.
We hand-verified every number at its exact line — and every claim below links to the source so you can check us.

AI startups crypto fitness indie hacking design food sports music travel news

How the ranking system treats your specific niche.

What the algorithm actually pays for

Your post's score = predicted probability of each action × these exact weights, straight from home-mixer/params/param.rs.

Seven post archetypes, ordered by algorithmic value, adapted to your topic.

Your unfair advantages

How discovery actually works

The mechanics behind reaching people who don't follow you — this is where most creators lose.

Reach killers to avoid

These are the documented ways posts and accounts silently lose distribution.

Numbers to remember

The hard constants from the codebase, on one card.

48 hoursa post's entire algorithmic lifespan — then it's dropped from For You
8 likesminimum before a post enters the community-discovery index
~400 followersactive followers needed to be assigned to interest communities
8h half-lifehow fast your post's early-liker community signal decays
1,000 / 1,000follower & view ceilings for the guaranteed small-account boost
128 likesthe "high-traction" bucket where LLM safety & quality review kicks in
−234the report weight — one report outweighs ~468 likes
25k / cred 54the thresholds above which accounts skip automated demotion
7–30 dayshow long most automated reach-limit labels last before expiring
62% → 44%value of your 2nd and 3rd post in one feed load (diversity decay)
Honest fine print: weights shown are X's open-sourced defaults (synced Aug 2026). They're live-tunable server-side and the final score also depends on the model's per-user predictions. XLens tells you what the system rewards — it can't guarantee outcomes. Source: xai-org/x-algorithm.