[DigitalToday reporter Yoonseo Lee] OpenAI CEO Sam Altman (샘 알트먼) mentioned “goblin” as a name for the company’s next artificial intelligence model.

On May 11 local time, blockchain media outlet BeInCrypto reported that Altman recently referred on X, formerly Twitter, to “more goblin” as he asked about improvements to the next model. He then replied that naming the next model “goblin” would be “almost worth it to make you all happy.”

“Goblin” is a phrase that has already been analysed inside OpenAI. OpenAI released a report in late April titled “Where did goblin come from” and analysed why the GPT-5.1 family models often draw on creatures such as goblins or gremlins in metaphorical expressions.

The report explained that model behaviour is shaped by many small incentives. In this case, personality customisation training, especially reinforcing an eccentric tone, unintentionally rewarded metaphorical, organism-based language, leading to the spread of goblin-like phrasing.

Altman’s earlier description of the current model as an “autistic savant” is also cited in the same context. The phrase is linked to a concern that the current model shows strong performance on technical tasks but reveals inconsistency in speaking style and tone.

Altman also shared an example of OpenAI’s coding system Codex handling real work without additional human intervention. He said he ran several Codex jobs, spent time with his child and came back to find the work finished, adding that it made him “very optimistic about the future.”

OpenAI’s view of Codex has also been changing. Codex is being presented no longer as a simple code autocomplete tool, but as an agent-style system that maintains a task list, sets the order and returns results. The focus appears to be shifting toward making workable code by interpreting natural-language instructions without a developer intervening at every step.

This strategy shift also ties into the competitive landscape. Codex competes with coding assistant tools released by Anthropic and Google. All three companies are highlighting a “hands-off” development flow that can handle tasks without a person continuously involved. OpenAI has been emphasising such autonomous workflows in its enterprise strategy, particularly after expanding cooperation with Microsoft.

OpenAI now faces two challenges at once. On one side, it is speeding up in the competition for autonomous tools that handle engineering work without supervision, led by Codex. On the other, it must refine the issue of the consumer model’s tone and personality moving in unexpected directions.

As a result, the focus for the next model is expected to be on function and personality control rather than the name. It is unclear whether the model will actually be called “goblin,” but a key yardstick is likely to be how much OpenAI improves autonomous work capability and user-friendly response tuning in the next release.

what if we name the next model "goblin" almost worth it to make you all happy...