The original Autoblock Audit found that the accounts Aaron Rodericks, Bluesky's Head of Trust and Safety, blocked from 28 September 2026 were disproportionately trans-identified. One explanation is that those accounts had called him a pedophile or told him to kill himself. This page checks that against every post they addressed to him that can still be read.
“The original report” on this page always means that audit. Read the archived copy, saved on 13 October 2026, or the live version if it is still up.
Two kinds of post were counted. Pedojacketing is calling him, or implying he is, a pedophile, groomer or protector of child abuse material. Suicide-baiting is telling him to kill himself or wishing that he would. An account counts if it sent even one such post since December 2024. A separate page shows examples of both kinds.
Almost all of it is pedojacketing. Suicide-baiting is rare among the blocked accounts and absent from the comparison group. For every bulk-loaded account that qualifies, the post was written before the block landed.
The same count, split by what each blocked account says about itself in its bio or display name. Trans-identified accounts on the list did qualify more often than the rest. Even so, most of them have no such post on record.
These shares are floors. Deleted posts cannot be read, and accounts that have been taken down show nothing at all. Thin lines show the range the true share plausibly sits in.
The original report estimated that a trans-identified account had about two and a half times the odds of being blocked as another account that wrote equally harsh posts. If harassment were the real reason, adding it to the model should pull that figure down towards 1. It barely moves.
Sending a qualifying post is a strong predictor of being blocked in its own right: . Both things are true together. The list catches people who harassed him, and it also catches trans-identified accounts at a higher rate than their posts explain.
Accounts are sorted here by the harshest reply or quote they sent him, scored with the same toxicity model and cut-offs as the original report. In every band the blocked accounts are the more likely to have sent a qualifying post. Where the grey dot is missing, fewer than five unblocked accounts qualified and the figure is withheld.
Each blocked account's own posts were read back to 1 December 2024 for any reply, quote or @-mention of him. The original report looked back only to January 2026 and found no direct contact for about of readable bulk-loaded accounts. The longer window finds earlier contact for many of them.
The blocks themselves match the original report exactly: .
app.bsky.graph.block records in his repository. Bursts of 20 or more blocks less than 2 seconds apart count as bulk loads.unbiased-toxic-roberta score across an account's replies and quotes, banded at 0.02, 0.1 and 0.5 as in the original.