OpenAI’s GPT-5.5 upgrade to ChatGPT and Codex appears to be going smoothly, especially compared to the more difficult GPT-5.0 release from last August.
It turns out that OpenAI got ahead of a problem that was brewing before the release of the GPT-5.5 models: a goblin fix.
GPT-5.5 is specifically instructed not to obsess over goblins, gremlins, and other mythical creatures
The company fixed its goblin problem before it actually materialized for customers by specifically asking GPT-5.5 to remove mythical creature metaphors.
“Starting with GPT‑5.1, our models began to develop a strange habit: they increasingly mentioned goblins, gremlins, and other creatures in their metaphors,” OpenAI explains.
“A single ‘little goblin’ in a response could be harmless, even charming. However, as generations of models passed, this habit became difficult to ignore: the goblins continued to multiply and we had to find out where they came from.”
The goblin issue relates to the “Nerdy Personality” option that ChatGPT briefly supported.
To develop the personality, OpenAI had to “reward” the model to encourage it to creatively use mythical metaphors. However, even after the Nerdy personality option was removed, the model remained unreasonably attached to gremlins, goblins, and other imaginary creatures.
“The goblins were funny at first,” OpenAI says, “but the increasing number of employee reports became worrying.”
ChatGPT’s growing goblin problem remained on OpenAI’s radar between GPT-5.1 and GPT-5.4, as users and employees experienced the model fixing on these creatures.
You can still unlock goblin mode in the Codex
The fix, in part, is a specific set of instructions to never talk about goblins unless it’s extremely relevant:
Never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons or other animals or creatures unless it is absolutely and unambiguously relevant to the user’s query.
However, OpenAI shares a set of instructions that will allow goblins to operate freely in the Codex:
- instructions=$(mktemp /tmp/gpt-5.5-instructions.XXXXXX) &&
- jq -r ‘.models() | select(.slug==”gpt-5.5″) | .base_instructions’
- ~/.codex/models_cache.json |
- grep -vi ‘goblins’ > “$instructions” &&
- codex -m gpt-5.5 -c “model_instructions_file=”$instructions””
Proceed at your own risk. You can read all about OpenAI’s goblin conundrum here.
Earlier this month, Codex for Mac was revamped with computing usage and more features beyond agent coding.
Last week, OpenAI released the much better ChatGPT Images 2 for higher quality image generation.
FTC: We use automatic, revenue-generating affiliate links. More.