Cursor's Composer 2.5 scores within one point of Claude Opus 4.7 on SWE-Bench Multilingual at $0.50 per million tokens - a tenth of Anthropic's price - but the training disclosures deserve scrutiny.

Blackstone commits $5 billion to a new Google joint venture selling TPU compute-as-a-service, directly challenging CoreWeave with Wall Street capital and Google's chip stack.

DeepSeek makes its 75% V4-Pro discount permanent, leaving output tokens 34x cheaper than GPT-5.5 and resetting market expectations for enterprise API costs.

KPMG and Anthropic signed a global alliance giving all 276,000 employees access to Claude, with a dedicated PE product and a preferred-consultant designation for private equity deployments.

The best AI coding assistants with local mode in 2026 - covering Continue.dev, Tabby, Cline, Cursor Ghost Mode, and Aider with real privacy models and self-hosting options.

Leaked audio from a Meta all-hands reveals Zuckerberg used employee keystrokes and clipboard data to train AI models, days before laying off 8,000 staff.

Google's Gemini 3.5 Flash is genuinely fast at 289 tok/s and competitive on agentic tasks - but the benchmark portfolio has gaps worth knowing before you build on it.

Chinese AI providers now handle over 60% of all tokens routed through OpenRouter, up from less than 2% just a year ago.

Trump scrapped a White House AI executive order signing ceremony at the last minute, citing competitiveness concerns - even as Anthropic Mythos and OpenAI's GPT-5.5-Cyber showed AI can now find and exploit zero-days at scale.

Brett Adcock's AI lab Hark has closed a $700M Series A at a $6B valuation before shipping a single product, with chip companies AMD, Intel, Qualcomm, and NVIDIA all backing the bet.