← X Coverage
1984 — X Posts
@TheOfficial1984 (1984) is a technically focused account whose threads on automated content moderation are preserved in this site's Censorship photo cluster. The bullets below are claims attributed to their X posts; the technical descriptions of published research systems are plausible on their face, but the operational claims about who acts on the outputs are the poster's and are not adopted here.
Claims from X posts
- The ADL allegedly assigns a "social credit score" on X using machine-learning models built at UC Berkeley's D-Lab — the Online Hate Index V3 (OHI V3) (extended thread capture).
- Technical slides describe transcribed text fed into a fine-tuned transformer-based neural network (BERT / RoBERTa architectures), with semantic analysis for implicit bias and dog-whistles, contextual scoring, and thresholding at 0.9 confidence triggering alerts.
- The thread further alleges the system can identify users by voice from X Spaces audio, that it flags content, and that "Israeli intelligence removes it" — claims about voice ID and foreign-service action that are unverified and not adopted by this site.
- Related posts allege X moderation teams working with CHEQ and AU10TIX (described in the posts as Israeli-intelligence-linked firms) and academic labs to monitor, label, and suppress content in real time, and discuss "labeling bias" in hate-speech classification models (thread captures).
Investigative value
Automated flagging is directly relevant to how quickly investigation material disappeared from platforms — the reason these screenshots are preserved under Censorship. The checkable core (published classifier research) should be separated from the unverifiable claims layered on top of it.