Tech1 hr ago

OpenAI Bars Goblin Talk in Codex CLI Prompt as CEO Jokes About 'Goblin Moment'

OpenAI updated its Codex CLI system prompt to block mentions of goblins and similar creatures after users reported GPT‑5.5 producing off‑topic animal replies, prompting a joke from CEO Sam Altman about a goblin moment.

Alex Mercer/3 min/US

Senior Tech Correspondent

TweetLinkedIn
OpenAI has banned Codex from talking about Goblins

OpenAI has banned Codex from talking about Goblins ( AI generated image )

Source: LivemintOriginal source

Context

OpenAI released the source code for its Codex CLI tool on GitHub last week. The code includes the system prompt that guides how GPT‑5.5 behaves when users issue commands. Codex CLI is a command‑line interface that lets developers use OpenAI’s Codex model to generate and edit code through natural language prompts. Within that prompt, a repeated warning tells the model never to mention goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals unless the query directly concerns them.

Key Facts

The prohibition appears twice in a 3,500‑word block of base instructions for GPT‑5.5. Earlier versions of the same prompt, stored in the same JSON file, do not contain this rule, showing it is a new addition. OpenAI CEO Sam Altman posted on social media that Codex is having a "goblin moment," likening the situation to the early buzz around ChatGPT.

What It Means

The rule suggests OpenAI observed users receiving unwanted animal‑themed replies from the model and moved to suppress them. By tightening the prompt, the company aims to keep Codex focused on coding tasks. Observers will watch whether future updates adjust the wording or if reports of goblin references drop off.

TweetLinkedIn

More in this thread

Reader notes

Loading comments...