Anthropic considers a $60B+ IPO, OpenAI shut down video generator Sora, and Google released voice model Gemini 3.1 Flash Live.
🤖 AI
Anthropic began IPO talks with Wall Street banks for October — the listing could attract over $60B. No final decision yet. TLDR AI
OpenAI shut down video generator Sora — the tool ceased operations amid questions about the viability of generative video models. TechCrunch
Google launched Gemini 3.1 Flash Live — a voice model for real-time dialogue with low latency, available via API and consumer products. Google
Mistral released Voxtral TTS — a 4B-parameter multilingual speech generation model for voice agents. Mistral
Cohere unveiled Transcribe — an open-source transcription model, leader of the Hugging Face Open ASR leaderboard. Cohere
Cursor introduced real-time RL for Composer — the model updates every 5 hours based on user reactions in production. Cursor
💻 Software
Apple will open Siri to third-party AI assistants in iOS 27 — beyond ChatGPT, helping generate revenue through App Store subscriptions. TLDR
CapCut got Dreamina Seedance 2.0 from ByteDance — a video and audio generation model, currently for paid users. CapCut
⚙️ Robots & Hardware
Jensen Huang at GTC 2026 presented the AI factory concept — a 2.5-hour keynote on the future of mass token production. TLDR
💰 Business
SoftBank raised a $40B loan — the deal hints at an OpenAI IPO in 2026. TechCrunch
Apple gave iPhone designers bonuses up to $400K as 4-year RSUs to prevent OpenAI from poaching talent. MacRumors
Anthropic won court case against the Pentagon — the judge called the department's actions "retaliation for the First Amendment" after the company refused to give unlimited model access. CNBC
David Sacks left the White House "AI czar" post. TechCrunch