Live Feed
Yotta Adds $6B to Blackwell Buildout as India Chases Sovereign AI Compute
Yotta is expanding its Nvidia deployment from 20,000 to 30,000 Blackwell GPUs and planning tens of thousands more next-generation chips. The move puts Indian AI infrastructure on a different scale: compute sovereignty now means GPU inventory, not just local data laws.
Chevron and Microsoft Lock 20-Year, 2.67GW Gas Deal for West Texas AI Campus
Microsoft is bypassing the grid queue with a dedicated Chevron-backed power complex in the Permian Basin. Project Kilby is expected to start in 2028 and ramp to 2.67GW, enough to make energy procurement part of AI infrastructure strategy.
OpenAI Daybreak Moves GPT-5.5-Cyber From Bug Discovery to Patch Automation
OpenAI expanded Daybreak with the full GPT-5.5-Cyber release, a Codex Security update, and Patch the Planet. The numbers make the shift clear: vulnerability discovery is becoming the easy part, while triage and accepted patches are the bottleneck.
GLM-5.2 Claims Arena Code Web Design #1 Over Claude Fable 5
Z.ai's MIT-licensed GLM-5.2 topped Arena's creative coding category, displacing Anthropic's Claude Fable 5. Vercel CEO Guillermo Rauch: 'Genuinely impressed, almost shocked. This changes things.'
George Hotz: AI Doom Framing Is Engineered to Justify Valuations the Technology Cannot Support
In a June 21 essay, the comma.ai founder argues that existential-risk rhetoric is 'optimized to cause panic' and designed to anchor inflated AI valuations around hypothetical future value rather than current product reality. Goldman Sachs CEO made a similar call the month before.
Swiss Universities Ship Apertus Mini: 16 Open Models, 1,811 Languages, Full EU Compliance
EPFL, ETH Zurich, and CSCS released Apertus Mini on June 15 — 16 small language models demonstrating distillation and quantization on the first fully open, EU AI Act-compliant foundation trained on 15 trillion tokens across 1,811 languages.
Claude Simultaneous Outage Hits Opus 4.8, 4.7, 4.6, and Sonnet 4.6 — Four Models, One Failure
Anthropic's full enterprise tier went down together on June 22. Four models across two generations failing simultaneously points to shared infrastructure, not isolated checkpoints — arriving as Anthropic tracks toward a record Q2.
Midjourney Medical Runs Its Scanner Through a Spa to Sidestep the FDA — BFLY +17%, Data Is the Real Product
Midjourney's full-body ultrasound platform is built on Butterfly Network chips and distributed through 'Midjourney Spas' — structured specifically to avoid FDA medical device classification. The body scan data, not the hardware, is the business thesis.
Anthropic Ships Claude Design Overhaul: System Imports, Code Round-Trips, Token Fix
Three months after the April research preview that briefly shook Figma's stock, Anthropic has shipped a major Claude Design update: design system imports, bidirectional code round-trips, and a fix for the token-burning problem that made iterative prototyping expensive.
Anthropic Adds Government ID and Facial Scans to Claude Consumer Accounts, Effective July 8
A privacy policy update published June 8 gives Anthropic discretionary authority to collect biometrics from Free, Pro, and Max users starting July 8. The mechanism could also serve as the tool to restore Fable 5 access for verified US citizens after the June export-control suspension.
US Suspects an ASML EUV Machine Reached China. ASML Says That's Impossible.
The Trump administration told ASML it fears one of its extreme ultraviolet lithography machines — the $250M chips that print AI processors — may have reached China in breach of export controls. ASML denies it. If confirmed, it ranks as the most consequential export control failure of the AI era.
Data Centers Have Not Raised Your Electricity Bill. A Causal Study Finds They May Have Lowered It.
An arxiv paper tests what AI infrastructure buildup actually did to US residential electricity prices from 2015 to 2024. The answer contradicts the prevailing narrative: a 10% rise in data center capacity correlates with a 0.4% fall in average household electricity prices — when the grid has capacity to grow.
Red-Team Study: Fable 5 Holds at 6.1% Jailbreak Rate, Opus 4.8 Reaches 11.5% Under Sustained Automated Attack
A new arxiv paper runs automated red-team tools against Anthropic's two frontier models and measures how often harmful prompts get through. Fable 5 is measurably harder to break than Opus 4.8 — but neither is immune. The surviving weakness is contextual, not syntactic.
Anthropic's Opus 4.7 Operates a Robot Dog 20x Faster Than Human Experts
Project Fetch Phase 2 put Claude Opus 4.7 alone against a robotic quadruped. It finished every task the fastest human team completed, in one-twentieth the time, using one-tenth the code. The AI still can't precisely retrieve a ball.
How the Mythos Shutdown Happened: Amazon CEO Jassy, SK Telecom, and 90 Minutes
New reporting identifies the two events that triggered the White House export order: a South Korean telecom with suspected China ties got Mythos access, then Amazon's CEO personally alerted the administration about a safeguard bypass. Anthropic had 90 minutes from notification to full shutdown.