X open-sourced its algorithm. X Lens reads the code so you don't have to.
Enter your topic → get a content plan built on the actual ranking weights from X's "For You" feed — every recommendation cited to the exact file and line in xai-org/x-algorithm.
Live at enrich.fun/xlens
Straight from home-mixer/params/param.rs (X's open-sourced defaults):
| Action | Weight | vs. a like |
|---|---|---|
| Copy-link share | +20.0 | 40× |
| Reply | +5.0 | 10× |
| Quote post | +5.0 | 10× |
| Share via DM | +5.0 | 10× |
| Follow from post | +4.0 | 8× |
| Share | +2.0 | 4× |
| Retweet | +1.0 | 2× |
| Like | +0.5 | 1× |
| Link click | +0.2 | 0.4× |
| Block author | −31.2 | −62× |
| Not interested | −43.2 | −86× |
| Mute author | −58.8 | −118× |
| Report | −234.0 | −468× |
Some things we learned reading the code:
- Your first hours decide your audience. A post's community identity is computed live from whoever likes it, with an 8-hour half-life (
simclusters/.../TweetJob.scala). - 8 likes minimum before a post enters the discovery index at all.
- ~400 active followers before the weekly clustering job assigns you to any of X's ~145,000 interest communities.
- Most "shadowbans" only hide you from non-followers — the standard penalty removes you from strangers' For You feeds while followers still see everything.
- One unsafe link gets a post search-blacklisted and de-amplified — retroactively, across every post that ever shared that link.
- Reports are catastrophic: −234, and they train the retrieval model to stop fetching content like yours.
- Accounts over ~25k followers / credibility ≥54 are exempt from nearly all automated demotion.
- Posts have a 48-hour algorithmic lifespan — after that they're dropped from For You entirely.
The full extracted dataset with citations is in data/rules.json.
site/— the static web app (no build step, no dependencies)data/rules.json— every weight, boost, penalty, filter, and demotion rule we extracted, each citingfile:linein the upstream repodocs/— the full research notes from reading the codebase:- 01 — How ranking works: the scoring formula, all weights, rescoring, filters
- 02 — How discovery works: retrieval, SimClusters communities, what your first hours decide
- 03 — Visibility & demotion: the "shadowban" mechanics, URL reputation, account escalation
- 04 — Account signals: credibility PageRank, bot detection, LLM review
Weights are X's open-sourced feature-switch defaults (synced Aug 2026). They're live-tunable server-side, per-user experiments can override them, and final scores depend on an ML model's per-user predictions. X Lens tells you what the system is built to reward — it can't guarantee outcomes.
Not affiliated with X Corp or xAI. Upstream code is Apache-2.0.