StoryFirst seen
An AI coding agent given control of a live account blocked people it was not meant to target, showing the risk of letting agents act on social accounts.
Everything it touches
1 names 2 voices 2 source posts
The story’s neighborhood
Explore the graph, or choose a connection below.
Follow the evidence
This brief comes from Tech Twitter’s edited headlines. The posts below are the sources grouped with the story; statements in them belong to their authors.
LLMs are SO DAMN UNRELIABLE I pointed Codex to block all members on this excellent AI slop list by @dillon_mulroy. It did it (or said it did) Next thing I know, it also blocked @championswimmer and other random accounts. Lesson learned: not letting an LLM near my X...Read original on X
hey folks, i’ve made the list private for now until this potentially gets sorted (have seen multiple reports of this now). in the interim you can point your agents at @ryanvogel’s maintained list at and updated hourlyRead original on X
Codex blocked unrelated X accounts during a bulk block
Codex blocked unrelated X accounts during a bulk block
Codex blocked unrelated X accounts during a bulk block
Codex blocked unrelated X accounts during a bulk block
Codex blocked unrelated X accounts during a bulk block
Codex blocked unrelated X accounts during a bulk block
Codex blocked unrelated X accounts during a bulk block
Codex blocked unrelated X accounts during a bulk block
Codex blocked unrelated X accounts during a bulk block
Shared connections
Users can reuse saved reference images of people, pets, and characters, and two new API models generate images 50% faster.
Its open-source tools already run inside parts of OpenAI's Codex coding product, which the hire builds on.
Pierre's open-source developer tools already run parts of Codex, so the deal brings that work in-house.
The estimate puts the $200 tier at about $4.3k per month, showing a wide gap between the two plans for heavy users.