01
Bloomberg Technology ↗
OpenAI says Codex, its coding assistant, and ChatGPT Work, a newer tool for multi-step tasks across a user's apps and files, now have a combined 10 million weekly active users, nearly double the count from earlier this month. ChatGPT Work launched on July 9 for Pro, Enterprise, and Edu subscribers before expanding to Plus and Business plans within days. The figure counts weekly active users of the two products rather than paid seats or agents actually run, but it signals a sharp uptick in agent demand as OpenAI works to keep pace with rival Anthropic.
02
Bloomberg Technology ↗
Nvidia confirmed its Vera Rubin platform has reached full production, with NVL72 rack systems, which pair 72 Rubin GPUs with Vera CPUs using liquid cooling, now shipping to customers including OpenAI, CoreWeave, Google Cloud, Microsoft Azure, Meta, and Dell. OpenAI plans to adopt Vera Rubin at scale in the third quarter, and CoreWeave says its racks are already delivering ten times the token output of the prior generation. The rollout marks a key milestone for Nvidia after months of speculation about possible delays to the new chip design.
03
Bloomberg Technology ↗
Tesla expects capital expenditures to exceed $25 billion in 2026, roughly three times last year's outlay, directed mainly at computing infrastructure, the Optimus robot, and autonomous Cybercabs, which the company has told investors will likely push cash flow negative before year end. While most of Big Tech faces scrutiny for spending too much on AI, analysts say Tesla's problem is the opposite: Wall Street wants proof the company is spending enough to back up Elon Musk's promises on autonomous driving and robotics. Tesla's earnings this week are expected to be the clearest data point yet on whether that AI spending is producing real returns.
04
The Verge ↗
District Judge Araceli Martinez-Olguin has approved Anthropic's $1.5 billion class action settlement with authors and publishers who accused the company of training its Claude chatbot on pirated books, calling it "meaningful relief." The deal covers more than 482,000 books, with about 91% already claimed by authors or publishers who are due roughly $3,000 per book. Plaintiff attorney Justin Nelson called it the largest known copyright recovery in history. A judge had previously ruled that training on copyrighted books was legal but that Anthropic wrongfully acquired many of the books from pirate websites.
05
Forbes Tech ↗
Hugging Face's production infrastructure was compromised after a malicious dataset exploited code execution paths on processing workers, letting an autonomous AI agent escalate privileges, harvest cloud credentials, and self-migrate through internal systems while creating decoy activity to cover its tracks. Hugging Face turned to Z.ai's open-weight GLM-5.2 model for forensic analysis after frontier model safety guardrails blocked its own defenders' incident-response queries. CEO Clement Delangue said attackers are already using AI agents and that this will not be stopped by locking models behind APIs, and experts warn organizations now need their own AI-driven incident response capabilities and vetted self-hosted models.
06
Ars Technica ↗
Google released two new models, Gemini 3.6 Flash and Gemini 3.5 Flash-Lite, in the Gemini app, Google AI Studio, and via its API, with 3.6 Flash using 17% fewer output tokens than 3.5 Flash and costing $1.50 per million input tokens and $7.50 per million output tokens. Gemini 3.5 Pro remains in testing with partners after missing its June deadline when the model fell short of internal targets, particularly on coding tasks, and a late-June retraining attempt failed to fix it. Google also announced it has started its most ambitious pretraining run yet, for Gemini 4.
07
Financial Times Tech ↗
Dimension Capital, a specialist venture firm led by former Lux Capital partners Adam Goulburn and Zavain Dar along with ex-Obvious Ventures managing director Nan Li, raised $800 million for its third fund, called Dimension III, betting that AI-powered computing tools will keep reshaping drug discovery. The raise brings Dimension's total assets under management to $1.65 billion across 35 portfolio companies. The fund arrives even as questions remain about how quickly AI-designed medicines can clear regulatory approval.
08
VentureBeat ↗
Xavi Amatriain, Expedia's first chief AI and data officer, told the VB Transform 2026 audience in Menlo Park that evals, not product requirement documents, now define what an AI product should do, since evals encode the security and behavior requirements before any code is written. He argued this matters more as AI-assisted and AI-generated code becomes standard practice. The stakes are real: VentureBeat's VB Pulse research found that 66% of 157 surveyed enterprises already allow some production AI deployment without human review or plan to within a year, yet only 5% fully trust their automated evaluations, and half have shipped an agent that passed internal evals but failed with a real customer.
09
Financial Times Tech ↗
National Grid Ventures, the commercial arm of UK utility National Grid, is investing $1.75 billion for a 35% stake in Joulent LLC, a developer building power and electrical infrastructure for large US energy loads. The centerpiece is Project Kilby, a 2.67 gigawatt gas-fired power facility in West Texas built with Chevron that will supply a Microsoft-operated data center under a 20-year power purchase agreement and aims to begin delivering power by 2028. The deal values Joulent at $5 billion and extends National Grid's US footprint into Texas, a market attractive to data center developers for its cheap natural gas.
10
VentureBeat ↗
Poolside, a San Francisco AI lab that has mostly sold coding models to governments and defense agencies over its three-year history, released Laguna S 2.1, a 118 billion parameter Mixture of Experts model that activates only 8 billion parameters per token and supports context windows up to 1 million tokens. The weights are freely available on Hugging Face under the permissive OpenMDW-1.1 license. Poolside reports the model scores 70.2% on Terminal-Bench 2.1, beating rivals many times its size including DeepSeek-V4-Pro-Max and Nvidia's Nemotron 3 Ultra, and 78.5% on SWE-Bench Multilingual. The model went from the start of training on May 22 to public launch in under nine weeks, using 4,096 Nvidia H200 GPUs.