This underrated ChatGPT feature lets you replace AI's annoying personality – here's how to use it (www.techradar.com)

OpenAI has quietly centralized ChatGPT’s personality controls into a single Personalization page after user backlash over ChatGPT-5’s colder, more efficient tone and the temporary retirement of the friendlier ChatGPT-4o. Sam Altman flagged the change on X, underscoring how seriously OpenAI is treating tone complaints. Users can tweak personality by clicking their account name → Personalization → ChatGPT Personality, picking presets (e.g., Listener, Nerd) or adding custom instructions like “Be talkative and conversational.” These settings can soften ChatGPT-5’s functional style without switching models; note that legacy ChatGPT-4o remains available only to $20/month Plus and Pro subscribers, not free-tier users. For the AI/ML community, the move highlights a key product-design trade-off: higher task performance and efficiency (ChatGPT-5) can clash with user expectations for empathy and conversational warmth. Centralized personalization is a practical interface-level solution that lets product teams tune perceived model behavior post-deployment, reducing the need for separate personality-oriented models. It also signals growing demand for fine-grained user controls over model affect, which could drive more sophisticated conditional prompting, preference-conditioned decoding strategies, or lightweight personalization layers in future updates. For practitioners, this is a useful case study in balancing model capability with UX through configurable runtime behavior rather than retraining.