Full Summary
This Saturday morning, OpenAI is making significant strides in both healthcare and user automation, with multiple sources confirming key new features. Both Dailyhunt and Memeburn report that OpenAI is expanding its healthcare focus, with researcher Karan Singhal leading the charge. A new free model, GPT-5.5 Instant, now outperforms GPT-4o and physician responses in company evaluations, and has reduced inaccurate health responses by 71% in two months. This is particularly relevant as over 230 million people already use ChatGPT weekly for health questions. The Decoder and Tech Times both highlight OpenAI's new "Record & Replay" feature for its Codex macOS app. This allows users to demonstrate a workflow once, and Codex transforms it into a reusable "skill" that can be repeated independently. This feature, available to paid ChatGPT subscribers outside the EU, UK, and Switzerland, shipped on June 18th. It creates a human-readable "SKILL.md" file that can be inspected, edited, and shared, making automation more accessible. In other news, Nokiapoweruser confirms that ChatGPT's voice input now supports over 70 languages and allows for natural mixing of multiple languages without manual switching. Additionally, The Decoder and DigitBin announce ChatGPT's new "Scheduled Tasks" for Plus, Pro, Business, and Enterprise users. This feature adds a "Scheduled" page in the sidebar to manage active tasks, allowing ChatGPT to proactively run reminders and monitoring jobs. Meanwhile, ETV Bharat reports that Google's Gemini Live voice assistant now remembers past chat details, aligning its voice mode with the existing text memory feature, though it's currently US-only and in English. Finally, Midjourney, known for its image generator, is launching "Midjourney Medical" with a full-body ultrasound scanner, the Midjourney Scanner, that creates detailed 3D body maps in about one minute. This could change how we monitor our internal health.