One persona tweak made ChatGPT say 'goblins' 4,000% more — caught on Reddit before OpenAI noticed
victor_explore · x · 2026-09-27
- Sky News' Rowland Manthorpe reports that ChatGPT's nerdy persona mentioned goblins nearly 4,000% more after a single model update; OpenAI intended the tweak to make the persona playful but only learned of the regression once the model went live.
- Developer victorexplore notes Reddit users spotted it before OpenAI did, and suggests a practical guardrail: if your agent relies on a persona prompt, rerun the same test chats after every model upgrade and diff the words it starts overusing to catch behavioral drift early.
More from coding & agent
- François Fleuret asks: are there programming languages tailored for LLMs? — francoisfleuret · 2026-09-27
- AI Can't 'Read the Room': The Hard Problem of Selective Info Sharing in Enterprise AI — devanshmehta · 2026-09-27
- antirez Builds a Real VM to Run Another World Intro on ZX Spectrum 48k — antirez · 2026-09-27
- A 'new type of mind' trained without natural language writes kernels at up to 66x Triton speed — repligate · 2026-09-27
- ThePrimeagen joins dhh's Omarchy Core to lead Agentic QA with new Oligarchy harness — nicolascraske · 2026-09-27
- ChatGPT Astra auto-creates a proxy email to connect Google Docs — 0xsachi · 2026-09-27