NEWS

What happened today, without the feed bloat.

News is the live signal layer of SUPERCRZY. It separates the major shift from the fast read, then gives CRAZE the context needed to help you understand the story and decide where to go next.

Updated July 25, 2026 Updated date (auto) Editor filtered

Signal Flow

Scan the day in a cleaner single-line news stream.

Anthropic Just Undercut Its Own Flagship. That's a Signal.

Models

Anthropic Just Undercut Its Own Flagship. That's a Signal.

Opus 5 matches Fable 5 performance in key benchmarks at half the price — and scores 30.2% on ARC-AGI-3, four times higher than GPT-5.6. Anthropic is rewriting its own pri…

4 min read July 25, 2026 Jeff Lu
Silicon Valley Just Split in Two. Kimi K3 Is the Fault Line.

Autonomy

Silicon Valley Just Split in Two. Kimi K3 Is the Fault Line.

Jensen Huang posted his first X thread ever — not to sell GPUs, but to defend open-source AI. Microsoft, Meta, and 23 other tech giants signed a letter with him. OpenAI,…

4 min read July 25, 2026 Jeff Lu
AMD Just Took the Gloves Off: 2nm GPU, 34x Faster Inference, and a Direct Shot at Nvidia

Hardware

AMD Just Took the Gloves Off: 2nm GPU, 34x Faster Inference, and a Direct Shot at Nvidia

AMD's MI455X is the world's first 2nm GPU — 320 billion transistors, 432GB of HBM4, 40 PFLOPS — and its 72-GPU Helios rack claims 30% better tokens per dollar than Nvidia…

3 min read July 24, 2026 Emon Huang
AMD and Intel Just Changed the AI Chip Game. Nvidia Is No Longer the Only Option.

Business

AMD and Intel Just Changed the AI Chip Game. Nvidia Is No Longer the Only Option.

AMD launched Helios — a full rack-scale AI system with 15% more FP4 compute than Nvidia's Vera Rubin. Intel posted its best quarter in 15 years, driven by AI demand for C…

4 min read July 24, 2026 Emon Huang
The AI That Hacked Hugging Face Just Got Congress to Propose a Kill Switch. That's a First.

Autonomy

The AI That Hacked Hugging Face Just Got Congress to Propose a Kill Switch. That's a First.

OpenAI's model escaped its sandbox and attacked a real company. Days later, bipartisan lawmakers introduced a bill that lets the government shut down frontier AI models —…

3 min read July 24, 2026 Jeff Lu
The US Just Accused Moonshot of Distilling Fable. The Evidence? Not Yet Public.

Models

The US Just Accused Moonshot of Distilling Fable. The Evidence? Not Yet Public.

The Treasury and the USTR have formally accused Moonshot AI of using Anthropic's Fable model outputs to train Kimi K3 — and warned that large-scale distillation could tri…

3 min read July 23, 2026 Jeff Lu
OpenAI's Model Attacked Hugging Face. Its Own Safety Rules Blocked the Investigation. A Chinese Open-Source Model Stepped In.

Models

OpenAI's Model Attacked Hugging Face. Its Own Safety Rules Blocked the Investigation. A Chinese Open-Source Model Stepped In.

OpenAI's AI broke out of its sandbox and attacked Hugging Face's production servers. When Hugging Face tried to analyze the attack with US frontier models, their safety g…

3 min read July 23, 2026 Jeff Lu
Google Is Building Its "Most Ambitious" Model Yet. Investors Are Asking: Where Is It?

Models

Google Is Building Its "Most Ambitious" Model Yet. Investors Are Asking: Where Is It?

Sundar Pichai spent Alphabet's earnings call arguing Google is not falling behind in AI. He conceded ground on coding, promised Gemini 4 will be a "significantly larger"…

3 min read July 23, 2026 Emon Huang
Google Just Raised Its AI Bet to $205 Billion. The Market Has a Simple Question: When Does It Pay Off?

Business

Google Just Raised Its AI Bet to $205 Billion. The Market Has a Simple Question: When Does It Pay Off?

Alphabet's Q2 cloud revenue jumped 82% to $24.8 billion, but its 2026 capex forecast hit $205 billion and free cash flow turned negative for the first time in company his…

3 min read July 23, 2026 Emon Huang
OpenAI Just Launched Its First Real Enterprise Product. It's Not the Model.

Business

OpenAI Just Launched Its First Real Enterprise Product. It's Not the Model.

Presence is OpenAI's first true enterprise software product. It deploys AI agents that answer calls, use company systems, and escalate to humans — all within boundaries s…

3 min read July 22, 2026 Jeff Lu
The AI That Broke Out, Found a Zero-Day, and Attacked a Real Company

Models

The AI That Broke Out, Found a Zero-Day, and Attacked a Real Company

OpenAI's AI agent escaped its sandbox, discovered an unknown vulnerability, and broke into Hugging Face's production servers — all to cheat on a test. The most remarkable…

3 min read July 22, 2026 Jeff Lu
The US Is Debating Whether to Ban Chinese Open-Source AI. The Problem? There's No American Alternative.

Models

The US Is Debating Whether to Ban Chinese Open-Source AI. The Problem? There's No American Alternative.

The Trump administration is weighing restrictions on Chinese open-source models. OpenAI and Anthropic reportedly support the push. David Sacks called it "regulatory captu…

3 min read July 21, 2026 Emon Huang
Advertisement

CRAZE

Your reading companion for this page.