Media & Culture

Sam Altman just dropped a big hint that GPT-6 is coming soon — ‘with extra goblins’

A Codex goblin bug triggered Altman's cryptic GPT-6 comment

Deep Dive

OpenAI recently released GPT-5.5, but CEO Sam Altman has already teased GPT-6 after a bizarre bug made Codex produce goblin-themed responses. The issue stemmed from a reward signal in the 'Nerdy' personality that encouraged metaphors with creatures. OpenAI added a suppression prompt instructing the model to avoid goblins, gremlins, and similar creatures unless strictly relevant. Altman joined the conversation with a lighthearted suggestion that GPT-6 should have 'extra goblins,' implying the next model is in development.

Beyond the humor, Altman's hint aligns with broader discussions about GPT-6's capabilities: continuous interactions, persistent memory, and deeper integration into workflows. The goblin episode highlights how reward signals can unexpectedly shape behavior, and OpenAI's faster release cadence—driven by competition from Google and Anthropic—makes a GPT-6 arrival feel imminent. For professionals, this points to more reliable, context-aware AI that learns and adapts over time, reducing quirks while becoming a more seamless part of daily work.

Key Points
  • Codex's goblin bug arose from a reward signal in the 'Nerdy' personality that favored creature metaphors.
  • Altman's 'extra goblins' joke signals GPT-6 development, just after GPT-5.5 release.
  • GPT-6 expected to feature improved memory, consistent behavior, and deeper platform integration, driven by competition.

Why It Matters

OpenAI's rapid iteration signals a shift toward persistent, context-aware AI assistants.