Microsoft puts numbers on the AI era: Azure at $33.2B, and $10B from OpenAI
For the first time Microsoft disclosed Azure's actual revenue — $33.2 billion — alongside a number that reframes the industry: $10 billion in revenue recorded from OpenAI. Ten straight record quarters, and the AI cycle now has a visible line item inside it. Source: TechCrunch · OpenAI revenue
Anthropic's IPO: VCs win, Amazon wins much more
The S-1 made VCs rich, but the structure of the deal means Amazon's returns dwarf them — and Anthropic is already monetizing the moment with tools that show enterprises what their AI spend is actually buying. Post-IPO Anthropic is moving on revenue and transparency at once. Source: TechCrunch · Spend tools
Google drops Gemini 4 Carbon
'Gemini 4 Carbon' is the top story on Hacker News this morning (791 points) — Google's latest model landing to a warm technical reception. Watch the benchmark follow-through over the next week. Source: Hacker News
Chrome's AI Autofill will dig through your Gmail to fill out forms
Google's AI-powered Autofill now reads your Gmail to complete forms — convenient, and a quiet expansion of what the browser is permitted to touch. Consent defaults will decide whether this lands as helpful or creepy. Source: TechCrunch
Wiz's CTO: the 'Vulnpocalypse' is coming for aging code
Wiz's CTO warns the internet's aging codebases are the next mass-exploitation frontier — AI is making vulnerability discovery cheap, and the patch side hasn't kept up. Source: TechCrunch
🏠 Local models: Nvidia's open transcription weights, and the quiet day around them
nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B (187 likes) is Nvidia's open-weights push into speech — a 30B-A4B mixture-of-experts transcription model, with GGUF and 4-bit MLX community builds already appearing for home hardware. Otherwise the new-model board is quiet since yesterday: Baidu's Unlimited-OCR-2 keeps climbing (324 likes, 8.4k downloads) and ccore's Qwen3.5-4B-3D-Coder holds at 227 likes — no fresh qualifiers, and the section refuses to pad. Source: Hugging Face