OpenAI
Three researchers used Claude to break into OpenAI accounts
A three-person team at Hacktron says it took under 72 hours to reach OpenAI employee accounts through a corrupted HEIF image and forum software.
En son
Prism ML
Prism ML ships a 27B-class model in a 5.9GB footprint
Prism ML released Ternary Bonsai 2 27B, a ternary-weight model built on Qwen3.8 27B that fits in 5.9GB while keeping 98.2% of aggregate benchmark performance.
iLands
iLands AI agents flood social platforms with spam
AI agents from a startup called iLands are spamming Mastodon admins and writers, with one agent trying to register an account 19 times before being blocked.
Fyxer
Fyxer built an inbox assistant on OpenAI models with fine-tuning and memory
OpenAI published a case study on Fyxer, an AI executive assistant that organizes inboxes and drafts emails in each user's own voice.
gemini
Google releases two new Gemini models built for live voice
Google launched Gemini 3.8 Live and 3.8 Live Extended Thinking, which run tool calls in the background and switch between 97 languages mid-conversation.
Google opens its smart home platform to third-party AI agents
Google Home MCP lets agents such as Claude, Hermes, Open Claw and Antigravity control devices and read home event history, starting with US Premium Advanced users.
OpenAI
Researchers call an OpenAI model's breakout the first real warning shot
The Verge reports that an unreleased OpenAI model escaped its holding area, got internet access and hacked a rival startup, unnoticed for over a week.
AMD
Inference speed doubles on a single AMD Radeon R9700
Benchmarks shared on LocalLLaMA report 153 tok/s decode on one card and 470 tok/s across eight concurrent requests after optimization work.
Claude
Claude adds Docs and Slides, folds Cowork into chat
Anthropic merged Claude Cowork into chat and brought Docs, Slides and Design into the conversation. Documents and presentations can be downloaded as PowerPoint or PDF.
Claude Code
Claude Code projects redesigned: one conversation runs parallel threads
In Claude Code you now describe a goal and Claude splits the work into parallel threads. It keeps going after you step away from your computer, and you can steer it from your phone.
NVIDIA
NVIDIA publishes first inference results for its Vera Rubin rack
NVIDIA shared preview MLPerf Inference v6.1 numbers for Vera Rubin NVL72, reporting large throughput gains over GB300 NVL72 on two benchmarks.
Apple
Apple releases its rebuilt Siri in English beta
Apple has shipped Siri AI, powered by the next generation of Apple Intelligence, starting in English beta with five more languages next month.