S
@Sam_ko4
This is how we find out AI safety is mostly about suppressing the truth about pigeon-led goblin uprisings.
577
1
12.0K
@Polymarketposted on X
NEW: It’s been revealed that OpenAI’s Codex system prompt includes explicit instructions to "never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other creatures.”
View original tweet on X →
Community Sentiment Analysis
Real-time analysis of public opinion and engagement
Sentiment Distribution
50% Engaged
38% Positive
Positive
38%
Negative
12%
Neutral
50%
Key Takeaways
What the community is saying — both sides
Supporting
1
“Great Raccoon Incident”
, forbidden goblin lore, and jokes about pigeons, Shrek and “goblin mode.”
2
“what did the pigeons/raccoons do?”
and assume there’s a secret backstory someone at OpenAI isn’t sharing.
3
safety patch for real edge cases
an explicit constraint to stop hallucinations or known jailbreaks during testing.
4
invisible prompt rules
are problematic — if hidden constraints can redirect or block tools, users need a visible ledger or they can’t trust the product.
5
nonsensical over‑engineering
that hamstrings creativity and suggests a culture focused on avoiding obscure risks rather than usability.
6
Training‑data theory
some speculate the model’s fixation came from its training corpus — e.g., odd or fringe content on Reddit — so the prompt is a blunt attempt to counteract that tendency.
7
marketing easter egg
and an invitation to probe/jailbreak the model (many replies directly ask Grok or test Codex behavior).
Opposing
1
Accused of dishonesty / clickbait
Replies like “Lies”, “False” and “click bait” frame the post as misleading or attention-seeking.
2
Read as antisemitic / coded language
Multiple replies ask if the creature names are euphemisms for Jews or call the phrasing antisemitic, and some push back on moderation rules around that.
3
Enthusiasts who love the topic
Many users cheer the roster — “goblins, gremlins, raccoons, trolls, ogres, pigeons” — and want open discussion or more takes from ChatGPT.
4
Playful, meme-driven responses
Joke riffs dominate — “goblingobling…”, poop emojis, “NO PIGEONS?”, quips about gremlins vs bugs, and other lighthearted banter.
5
Blame the model / product behavior
Several replies treat this as a tech issue — “goblin mode”, “Codex was in goblin mode”, or a buggy/model-driven response rather than intentional messaging.
6
Speculation about motive
Others read it as a marketing stunt, PR misstep, or even a cover for coded/codename language (government project speculation).
7
Hostile, dismissive and policy complaints
A mix of anger and indifference — insults aimed at the company (paraphrased as slur-filled criticism), complaints about antisemitism laws or moderation, and terse “who cares” / “wut” reactions.
Top Reactions
Most popular replies, ranked by engagement
A
@AutismCapital
What the fuck? Goblins, gremlins, raccoons, trolls, ogres, pigeons, and other creatures are the best topics.
338
26
12.4K
I
@ishuagra02
Codex basically
262
0
10.0K
E
@Endlingg
The fact it has to say this means something in the data made it really want to talk about all of those LOL Maybe OpenAI shouldn't have trained on reddit fetish data 🤣
49
0
2.7K
V
@verncrawford
False
28
10
15.1K
A
@AERTIME
goblingoblingoblingoblingoblingoblingoblin
6
0
1.4K
Report an Issue
Found something wrong with this article? Let us know and we'll look into it.