ChatGPT: Everything you need to know about the AI-powered chatbot (techcrunch.com)

OpenAI’s recent ChatGPT rollouts show a pivot from a single chatbot to a full AI platform: the company released GPT-5 (with Auto/Fast/Thinking modes), a GPT‑5-Codex variant that dynamically allocates compute (from seconds up to seven hours) for complex coding and refactoring, and kept legacy models like GPT‑4o and 4.1. New product moves include Sora (text-to-video), an AI music generator reportedly trained on annotated Juilliard scores, ChatGPT Atlas (an AI-first browser), and a developer Apps SDK that lets partners embed interactive apps inside ChatGPT (early integrations with Booking, Spotify, Figma, etc.). Commerce features — Instant Checkout and a Walmart partnership — plus Pulse (personalized morning briefs) and a low-cost ChatGPT Go expansion across Asia signal a push into search, shopping, and daily assistant workflows. OpenAI also reported 800M weekly active users, while mobile growth shows signs of slowing. These updates matter because they shift ChatGPT from conversational reply to action-oriented platform and infrastructure scale. Technical implications include multimodal training and serving at scale (video, audio, music), variable-time model execution for code-heavy tasks, and richer integrated toolchains via apps and commerce APIs. Safety and governance remain central: OpenAI disclosed over a million weekly suicide-related conversations and has consulted 170+ experts to improve responses, rolled out parental controls and stronger detection after lawsuits, and faces executive turnover and regulatory scrutiny — all while racing to expand data centers and funding to stay competitive globally.